Perpetual futures are having a moment.

You can now buy a leveraged, never-expiring bet on the price of SpaceX, trade it around the clock, and do all of this before SpaceX went public and has no ordinary market price at all. The contract settles in cash. This spring it implied a $1.78 trillion valuation for the company and did about $33 million of volume on its first day.

The instrument that makes this possible is the perpetual future, and it has been turning up everywhere this year. In March, S&P Dow Jones Indices granted Trade[XYZ], a Hyperliquid-based platform, the first officially licensed S&P 500 perpetual contract, available 24/7 onchain to non-US investors. Two months later, the same platform launched two wave-making pre-IPO perpetuals: Cerebras and SpaceX.

These instruments successfully discovered Cerebras’ price days before the IPO, and the SpaceX perpetual is now set at a $1.78 trillion implied valuation, doing roughly $33 million of volume on day one. Coinbase, Binance, OKX, Bitget, Crypto.com and others followed within weeks with their own pre-IPO products. Hyperliquid, the venue underneath much of this, briefly hit a record 6.63% share of global perpetual volume against centralized exchanges in May, with its share against Binance specifically reaching 14.4%.

Then, the perp phenomenon came onshore. On May 29, the CFTC approved Kalshi’s bitcoin perpetual, the first regulated perpetual futures contract available to US persons. Six days later, CME Group’s chief executive Terry Duffy stood up at the Piper Sandler conference and called the approval “a disaster waiting to happen.”

These products have, for most of their existence, traded offshore. That is changing fast. Crypto-native firms are finding new things to write a perpetual on, the old guard is alarmed in public, and an entire product class is going mainstream. Two things this piece will argue:

1. If and when equity perpetuals arrive onshore, they will be felt at once, because the speculative wallet they slot into is already enormous.

2. The engineering itself, what a perpetual fundamentally requires, points at a future where anything measurable, with a continuous reference, can have a contract written on it.

Perps are a phenomenal product

Robert Shiller proposed the perpetual future in 1993, in a Journal of Finance paper about how to cash-settle derivatives on illiquid things like real estate that have no clean way to deliver. The idea mostly sat there for more than twenty years. Then in 2016 a crypto exchange called BitMEX shipped a perpetual swap on bitcoin and built one of the defining financial products of the following decade.

A perpetual is a futures contract with the expiration date taken out. That sounds like a small change, but it isn’t.

A normal future settles on a fixed date: you agree today to exchange something at a set price three months out, and on that date the contract resolves and disappears. To hold the position past then you sell the expiring contract and buy the next one, which costs money and attention every quarter. A perpetual never expires, so there’s nothing to roll. The position just sits there for as long as you want it.

The catch with a contract that never settles is that nothing obvious holds its price to the thing it tracks. A normal future converges to spot because it settles against spot; remove that and the two could drift apart. Perpetuals handle it with a funding rate, a small payment that passes between longs and shorts every eight hours or so. When the perpetual trades above spot, longs pay shorts, which makes holding a long more expensive and nudges the price down; when it trades below, shorts pay longs. The funding rate is the tether, doing the job an expiration date does in a normal future, except it never stops.

Three features define a perpetual, and no single one of them is unique. It never expires, so there’s no roll and no calendar fragmentation, just one book of liquidity on one instrument. It settles in cash, so there’s no delivery and nothing to custody. And it was built for retail: leverage in both directions, small position sizes, round-the-clock trading, no friction to short. Dated futures are cash-settled and levered. Spot trades continuously and never expires. What no other product manages is all three at once.

Which feature does the real work depends on what you put the perpetual next to. Against a dated future, the one that matters is no expiry. The dated contract already settles in cash and already offers leverage, so the perpetual adds little there; what it adds is the end of the roll and the concentration of liquidity into one book rather than four quarterly contracts on a curve. Against spot, the comparison flips. Spot is already continuous, so no expiry is the shared feature this time, and the perpetual’s edge is cash settlement and the retail bundle: leverage, easy shorting, no custody.

Crypto was the case where all three features won against both alternatives at once, which is why perpetuals took over crypto. Dated crypto futures had thin liquidity; crypto spot was clumsy to short and a chore to custody. The perpetual beat both comparisons cleanly. Today perpetuals are around three-quarters of all crypto trading, and global perp volume ran past $60 trillion in 2025, several times the spot market they derive from. People would rather trade the derivative than the thing itself.

There’s a cost buried in the design, and it’s the part most people skip. Continuous leverage needs a way to absorb a position that blows through its margin before the exchange can close it, which is what happens in a fast market.

Offshore venues use auto-deleveraging: when a blown-up trader can’t cover the loss, the exchange force-closes winning positions on the other side to square the book1. Retail tolerates that, since the worst case is a winning trade cut short and nobody loses more than they put in.

US central clearing runs losses the other way, through a capital waterfall: the failed broker first, then the clearing house’s guaranty fund, then the surviving members backstopping one another. That last step is the one institutions can’t accept, because no clearing member wants to be on the hook for a competitor’s customers. The same contract is a different animal depending on which machine sits behind it.

This is a subtle, but perhaps the most interesting wrinkle around on-shored perpetuals in a DCM. ADL is something that crypto traders have grown to accept as it was necessary for a decentralized exchange.

All of this stayed offshore for most of a decade. The interesting question is what happens when it shows up somewhere it hasn’t been allowed before.

Public markets clue us into the disruption at stake

When the CFTC approved that first bitcoin perpetual on May 29, the price reaction worth watching showed up somewhere odd: in the share prices of the companies that run America’s options and futures exchanges.

What the market was pricing is the chance that perpetuals come for their business, and it was sorting the exchanges by how exposed each one looked. The interesting part is what “exposed” turns out to mean.

You’d guess the exchanges most at risk are the ones running cash-settled products, since a perpetual settles in cash. That’s half right, and the half it misses is the half that matters. Picture what a perpetual is from a trader’s seat: a leveraged, short-term, directional bet that pays out in cash and never makes you take delivery of anything. Two things are true about that bet at once. It is the kind of thing a retail trader reaches for, and it settles in cash. A product only competes with it if it does both. Cash settlement without a retail audience is just plumbing for institutions. A retail audience whose product hands you the underlying, single-stock options that settle in shares, sits one step removed from what a perpetual replaces. The flow a perpetual actually eats is retail-facing and cash-settled at the same time, which is why the table below has two columns where you might expect one.

Read the two left bars together and the selloff stops looking strange. MIAX took the worst of it because it is the closest thing on the list to a pure retail-options business, with almost nothing else on the income statement to absorb a hit. Sixty-two cents of every revenue dollar rides on short-dated, retail-oriented order flow. That number alone explains the move. CBOE is the one the second column explains. By retail share it looks modest, under a quarter of revenue, and a retail-only read would have it falling far less than MIAX. It fell nearly as hard. The reason is that CBOE’s retail flow is the most perfectly perp-shaped flow on the board.

Its franchise is SPX and VIX, cash-settled index options, much of it zero-days-to-expiry: short-dated, leveraged directional bets that settle in cash, which is a perpetual with an expiration date stapled on. Seventy percent of CBOE’s traded volume settles in cash, the highest share on the list, and the retail slice of it is exactly the slice a perpetual takes first. The same logic sorts the ones that held up.

So the public market has already decided perpetuals matter. It hasn’t yet had to ask whether they’re legal.

The regulatory door is ajar

Here’s something most people who follow markets don’t know: you can’t trade a futures contract on Tesla. Not on Apple, not on Nvidia, not on any single US stock. You can trade options on Tesla, you can trade an S&P 500 future that holds Tesla, you can buy the shares outright. The one thing you can’t do is trade a future on Tesla by itself. The reason is statutory, and it goes back decades.

From 1981 to 2000, futures on single stocks and on small baskets of stocks were flatly illegal in the United States. In 2000 Congress lifted the ban, but only halfway: these products became legal on the condition that the SEC and the CFTC regulate them jointly, with both agencies’ rulebooks applying at the same time. A dedicated exchange spent 18 years trying to make the joint-regulation version work and shut down in 2020 having gone basically nowhere. Broad-based index futures, the S&P 500 kind, got carved out and left to the CFTC alone, which is why those trade freely while single-stock futures effectively don’t exist. That was the state of things until a few weeks ago.

Three things have happened since late May, and together they point in one direction.

The first is the policy statement the CFTC put out on May 29, alongside its approval of the bitcoin perpetual. Most of it concerns how the agency will review perpetual contracts: case by case, rather than through the self-certification process exchanges normally use to list something new. The part that matters is the agency’s reading of a rule called Core Principle 3, which says a listed contract can’t be readily susceptible to manipulation. Here is how the Commission applied it to perpetuals:

For a traditional, cash-settled futures contract, the susceptibility-to-manipulation analysis directed at the cash settlement reference price is an analysis of one moment in time: the settlement reference price must be reliable at expiry. For a perpetual contract, however, the reference must be reliable at every funding interval, without interruption, for as long as the contract remains active.

That’s a higher bar than any dated future has to clear, and it’s written with nothing crypto-specific about it. The same is true of the asset classes the document says the framework could cover: “agricultural products, precious metals, equity securities, and narrow-based security indexes.” Equity securities and narrow-based indexes are precisely the things the statute prohibits without joint SEC-CFTC sign-off. The CFTC didn’t have to mention them, but it did anyway.

The second thing happened the same day. The CFTC issued a no-action letter letting Coinbase route US clients to perpetuals listed on Deribit, an offshore exchange, by treating those contracts as foreign futures. The mechanics cover crypto only, so this doesn’t reach equities. The point is the method: the agency used its own discretionary authority to pull an offshore product onshore, without waiting for the SEC and without new legislation. It’s a tell about how this CFTC moves when it wants something to happen.

The third thing was the chair saying it out loud. On June 8, ten days after the policy statement, the CFTC chair, Mike Selig, went on the Empire podcast and, asked how the two agencies would handle products that straddle their jurisdictions, reached for a specific example: “a regular security like an Nvidia perpetual that actually goes through a process between the CFTC and the SEC.” He sketched a rough order of operations in the same conversation. Crypto perpetuals get a fast lane through exchange self-certification now that the template exists. Securities, single stocks included, go through a joint process with the SEC. Physical commodities like grain and metals aren’t coming soon. The mechanism he kept returning to was “substituted compliance,” letting one primary regulator clear a product that would otherwise answer to two.

None of this makes equity perpetuals legal today. It does tell you that the people who would have to allow them are working out how. The market is thinking a step ahead, which raises the next question: if this door ajar swings wide open, how big does it get?

How big this gets in equities

There are two honest ways to size this, and they land on very different numbers, so it’s worth doing both.

The smaller number is what already exists. Add up the fees sitting on US equity directional flow today: the options exchanges, the clearing house, the payment for order flow, the commissions, and CME’s equity-index futures business. It comes to roughly $10 billion of revenue a year, maybe $15 to $20 billion counting the spread market-makers earn on the same flow.¹ Call that the floor: what a perpetual redirects if it does nothing more than take share, and the number I’d defend if I only got to defend one.

The bigger number comes from asking what perpetuals did the last time they turned up somewhere new. In crypto, they grew the market.

The product pulled in activity that hadn’t existed before it. The front edge of the same effect is already visible in equities: 0DTE options, contracts that expire the day they trade, went from almost nothing five years ago to a quarter of all US options volume. Retail turns out to have a large and previously unmet appetite for simple, leveraged, short-term bets.

Equities probably won’t go as far as crypto did, though. Most US equity trading is ownership rather than speculation. Pension funds hold equities to match liabilities, index funds hold them because their investors paid for exactly that exposure, and the tax code pays retail to sit still. None of those people have any use for a cash-settled perpetual, which gives you the price and no claim on the asset. So whatever multiple of spot the equity perp market reaches, it applies only to the slice of trading that’s actually directional. Call that 30% if you want a number, though I’d take a higher or lower one from somebody who’s thought about it harder than I have.

From there the arithmetic is short: directional flow, times whatever volume perps reach against it, times the fee.

The fee a competitive perpetual can charge matters more than how big the market gets, and crypto’s fee has been falling the whole time. The cells that clear the roughly $10 billion incumbents already earn on this flow are the ones worth arguing about, because that’s the line between eating the existing pie and growing it.

If this still reads as speculative, it’s already happening, just not here. Hyperliquid did about $11 billion of onchain equity-perp volume in 2025, with Nvidia, Tesla, and Alphabet the most active names. Trade.xyz’s S&P 500 perpetual is running at something like a $600 billion annual pace. Equity perp volume across venues grew ninefold in the first quarter of 2026, by BitMEX’s count.

And the contract-for-difference industry, which is leveraged cash-settled equity betting under a different name, has been legal across Europe and Asia for years and banned for US retail under Dodd-Frank, with a handful of public operators earning a couple billion a year from products Americans aren’t allowed to touch. The demand has been demonstrated. The piece that’s been missing is a legal venue at home, and that is the piece that just started to move.

Equities are the loud version of this story. The wallet is enormous, the names are famous, and the disruption shows up in stock prices people already follow. But the rule that opened the door says nothing about equities in particular. It sets a test, the one from the policy statement, and anything with a deep, continuously observed price can pass it.

The honest name for where this points is hyperfinancialization: the CFTC is assembling the machinery to put a tradable contract on more or less any measurable thing, with perpetuals as one vehicle and event contracts as another.

Firstly, I want to express deep gratitude to my readership.

When I started this Substack, the idea was to become “the Nate Silver or Matt Levine of prediction markets.” Well, I realized that I’m not Nate, nor am I Matt - they are inimitable.

Even so, this humble blog has changed my life completely and has opened up a world of opportunity that I couldn’t have imagined prior.

One example of this is that I somehow got to start a whole new venture capital firm called 5c(c) Capital. Ostensibly, the idea for the firm was to invest in early stage startups in the prediction market ecosystem. My Substack is about prediction markets, I understand prediction markets, and so my venture firm is about prediction markets. It was simple, and it made sense.

That said, as I continued to do the work of building a VC firm from scratch, I began to learn about new sectors, consider interesting new market structures and assets, and most importantly, meet fantastic and special entrepreneurs who are building things that aren’t prediction-market-related per se.

What I choose to write about on this blog is really just a reflection of what’s on my mind and the conversations I’m having with incredibly smart people on a daily basis. Since I started the Substack last summer, that has been mostly prediction markets and the regulatory matrix surrounding them.

But these days, I’ve been learning a lot about other things too: compute, perpetual futures, insurance brokerage, leverage, crypto, agentic commerce… the list goes on.

Now that my mind is on other things, I want to share the distillation of those thoughts on this platform too. It’s what I’ve been doing for 9 months now and I appreciate that so many of you appreciate it.

New direction, new name

The name of this blog was Fifty Cent Dollars, a reference to a niche prediction market meme. Now, this publication is growing beyond prediction markets, it makes sense to consider a bit of a re-brand.

At the same time, 5c(c) Capital received a cease-and-desist from another similarly named investment firm, and we are having to change the name of that as well. I don’t blame them, 5c(c) is a pretty cool name I guess and I’d want to keep it to myself as well. Shrug.

So, I figured that this is a great opportunity to re-brand both the blog and the VC firm I’m running, and combine them while we’re at it.

The new name is UFO Holdings1.

Here’s the one-liner: UFO Holdings is a worldview-driven investment firm backing builders on the frontiers of the New Economy.

I’ll still be writing here regularly and aim to continue delivering the thoughtful, accessible, and opinionated content that you have come to expect. Prediction markets were the first chapter, and I owe it so much. But, I aim to be forward-looking, not heat-seeking, and there is much more on the horizon to write about and invest in.

Thanks for everything so far, and stay tuned.

On the night of February 12, on an exchange that is usually pretty quiet with sports, three NBA games suddenly lit up the tape. Dallas at the Lakers, Milwaukee at OKC, Portland at Utah. Between the triple-header, they generated over 13 million contracts traded. ForecastEx is a CFTC-regulated prediction market operated by Interactive Brokers. It is a real exchange with a real license, but it had never done any meaningful NBA volume until that night1.

I don’t think ForecastEx achieved some customer acquisition miracle overnight. It didn’t improve its product, launch a marketing campaign, or deepen its order book with more liquidity. What happened is simpler: Robinhood pointed its order flow firehose at a different exchange for a triple-header evening of NBA action.

At the moment, Robinhood is the dominant retail distributor of prediction market contracts. When a user opens the Robinhood app, taps on an NBA game, and places a bet, that trade gets routed to a CFTC-regulated exchange for execution. For most of Robinhood’s prediction markets history, that exchange has been Kalshi. But the user doesn’t know this, nor do they care. The interface is identical regardless of which exchange is on the backend: same app, same button, same odds. The exchange is invisible infrastructure.

The data speaks for itself

Each bar is one day of NBA game volume, stacked by exchange. Blue is Kalshi, red is ForecastEx. Every day is all blue except February 12, where 35% of the volume suddenly appears on ForecastEx2. Then it’s back to all Kalshi again, as if nothing happened.

That red slice on February 12 is three games: Dallas at the Lakers, Milwaukee at OKC, Portland at Utah. Combined, they generated 13.4 million contracts on ForecastEx. The Robinhood user experience was identical regardless of which exchange processed the trade: same app, same button, same odds. The user can’t tell the difference. There is no difference to tell.

This is why the 35% figure matters, because it is a relatively clean read of Robinhood’s market share of NBA moneyline volume across these two exchanges. ForecastEx has essentially no organic sports users, so it’s not unreasonable to assume that every contract on ForecastEx that night was a Robinhood order.

And because the Robinhood interface is identical either way, those users bet at the same rate they would have on Kalshi. It stands to reason that roughly a third of Kalshi’s NBA moneyline volume in Feburary came from Robinhood.

Robinhood controls where the volume goes, and it can flip that switch overnight.

Weather patterns

The NBA routing was brief and dramatic, which makes for a phenomenally clear and striking natural experiment to analyze. But, the rise of weather markets on ForecastEx tells a similar story at a different scale.

ForecastEx and Kalshi both offer daily high temperature contracts: binary strikes on whether a city’s high temperature will exceed a given threshold that day. The markets are the same product, with the same cities and the same dates. The only real difference is the exchange matching the trades.

Before November 18, 2025, ForecastEx had zero weather trading activity. Then volume appeared overnight, with no organic ramp, no gradual adoption curve. The step-function pattern is the same fingerprint as NBA. To measure the overlap, I matched ForecastEx and Kalshi markets on identical city-date pairs, excluding cities that only exist on one exchange. This produced 454 matched city-days.

The first five weeks are Kalshi alone, the baseline. Then ForecastEx appears and immediately captures 60% of the combined daily temperature markets volume. It peaked at 72% in late November and has generally held between 53% and 67% since.

The critical detail: Kalshi’s weather volume didn’t collapse when ForecastEx appeared. The blue bars stayed roughly stable. So, my interpretation is that ForecastEx volume arrived on top of Kalshi’s existing flow. It was most likely Robinhood turning on weather markets for the first time and sending its flow to ForecastEx from the start, and its users didn’t know the difference3.

That distinction matters. With the NBA in January, Robinhood briefly diverted existing Kalshi-bound volume. With weather, Robinhood appears to have added ForecastEx as a parallel destination while leaving Kalshi’s flow intact. Both cases demonstrate the same structural point: Robinhood decides where the volume goes. The exchange receives what Robinhood chooses to send.

Distribution amplifies product-market-fit

NBA and weather show Robinhood can direct flow. Parlays show it can scale demand that’s already growing.

Kalshi launched multi-variable-event contracts (aka ‘combos’ or ‘parlays’) in September 2025, timed to the NFL season opener. The product immediately found traction: weekly volume grew from essentially nothing in September to about 45 million contracts per week by early December. That growth was organic and direct to Kalshi’s platform. Kalshi built the product, filed the CFTC certification, and seeded initial liquidity. The market responded.

Then Robinhood plugged in.

On December 17, Robinhood announced it would surface preset parlays and player props in its app. Within weeks, weekly volume exploded, jumping from the 45-60 million range to nearly 100 million, and then to 300 million per week by late January. The shaded region on the right marks the post-Super Bowl period, when NFL parlays disappeared and NBA carried the product alone. Volume held around 260-290 million per week even without football.

Kalshi did the hard work of creating a new product category. Robinhood’s distribution took it to a completely different scale. Both contributions are real. The question is which one carries more structural leverage.4

It’s not just Kalshi

Kalshi has grown enormously over the past year, from roughly 7 million daily contracts in late 2024 to over 100 million by late 2025. That’s not all Robinhood. Kalshi has built real direct demand: new product categories, a growing base of native users, API traders, and institutional participation. A year ago, Robinhood was widely understood to be the majority of Kalshi’s volume [1]. Today, the NBA data suggests Robinhood is roughly 35% of moneyline volume. That de-risking is genuinely impressive execution.

But Kalshi is not the only exchange whose growth story is really a distribution story

Nadex, the CFTC-regulated exchange operating as Crypto.com Derivatives, tells a strikingly similar tale. Before Underdog integrated with Crypto.com in September 2025, Nadex was doing modest volume. After Underdog plugged in and started routing its users’ sports bets to the exchange, weekly volume exploded by an order of magnitude. Same pattern, different names. Underdog is to Nadex what Robinhood is to Kalshi: the distribution layer that turns a quiet exchange into a busy one.

And here’s the punchline: both distributors have now moved to own their exchanges outright. Robinhood acquired its own CFTC-regulated exchange, and Underdog did the same last week. Two companies, on parallel tracks, arriving at the same conclusion independently.

This is not a coincidence. It’s game theory. If you’re a distributor routing millions of trades to a third-party exchange, you’re splitting revenue on every contract for infrastructure your users can’t distinguish from a white-label API. You’re also handing a potential competitor the data, the volume, and the regulatory track record that makes their exchange valuable. The rational move, once you’re big enough, is to bring that infrastructure in-house. The exchange license becomes a cost center instead of a profit center for someone else.

The weather and NBA data show why this dynamic is so hard to defend against from the exchange side. Even at 35% of volume, Robinhood can add a parallel exchange for weather overnight and immediately send it the majority of new flow. It can route three NBA games to a different exchange on a Tuesday and those games generate the same volume they would have anywhere else. The users don’t notice. They don’t choose the exchange. They chose Robinhood, or Underdog.

I was wrong

(…Again.)

Last year, when rumors surfaced that Robinhood was considering acquiring its own CFTC-regulated exchange, I said publicly that it wouldn’t happen.

I was confidently wrong for two reasons.

First, I knew firsthand from my experience at Kalshi how insanely hard it is to stand up and operate a regulated derivatives exchange: the compliance infrastructure, the surveillance systems, the CFTC reporting, all of it. Robinhood was making enormous revenue from prediction markets while doing ~1% of the work. The exchange did the hard work, and Robinhood collected the distribution fee, and it was the most important partnership in fintech in years! Why mess with a good thing?

Second, I was pattern-matching from fifty years of derivatives market structure. Brokers don’t acquire exchanges. The whole point of the exchange, in the world I was coming from, was that it was the irreplaceable plumbing. The CME is a $90B company with net profit margins second only to Visa and Mastercard, and the reason why is liquidity depth as a moat.

An institutional trader moving a $50 million Brent crude position cares deeply about order book depth, slippage, counterparty concentration. That depth is extraordinarily hard to build and nearly impossible to replicate, especially in derivatives where contracts are not fungible across exchanges. In that world, the exchange earns its structural position. The broker is the commodity.

Prediction markets invert this. The average sports bet on Robinhood is someone tapping a button to put $10 on the Lakers. That user does not care about order book depth. Hell, they don’t even know what an order book is. When trade size is small and the user is unsophisticated, liquidity depth stops being a moat. Robinhood swapped the pipes out on a Tuesday night and the same volume came out the other end.

When trade size is small and the user is unsophisticated, liquidity depth stops being a moat.5

I was wrong because I was navigating off the old map. The structural leverage in prediction markets doesn’t sit where fifty years of derivatives history would tell you it sits. It sits with whoever owns the user in the end.

I joined Kalshi Trading, the exchange’s in-house market-making operation, in the summer of 2022. CPI was the hottest story in the world. Inflation had just hit 9.1%, and the CPI prediction markets on Kalshi were buzzing. Headline month-over-month contracts were pulling in 250,000 to 480,000 contracts per release cycle, enormous by the standards of a young exchange. Casual traders flooded in, people who followed the inflation news and wanted to express a view.

Then a handful of quantitative modelers built serious CPI forecasting pipelines. They pulled the component data, modeled the seasonal adjustments, tracked the BLS methodology1. And month after month, they ate. The casual traders got picked apart. They were paying tuition, and eventually they stopped enrolling. Around the same time, CPI fell out of the news cycle. The fish stopped losing money and left, and no new fish replaced them. These were the dog days, where hardly any volume traded.

By mid-2024, headline CPI events were drawing as few as 12,000 contracts, a shadow of the 2022 peaks. Yet the forecast accuracy held: a handful of sharp traders, maybe five or ten at most, running fully built models, were trading against each other for a few thousand dollars each month… and so the forecasts stayed tight.

Then Robinhood integrated Kalshi contracts in early 2025, and the volume came roaring back. The April 2025 headline year-over-year event drew 11.4 million contracts—816 times more than the dog days trough. The accuracy barely moved.

I went back and analyzed every CPI prediction market Kalshi has ever listed, from June 2021 through January 2026. Nearly 150 events. Volume ranged from 7 contracts to over 11 million—a 1.6 million-fold spread. And across that entire range, trading volume explained less than 1% of the variance in forecast accuracy.2

The market was accurate when it was quiet, accurate when it was booming, and accurate in the long trough in between. Every other market in the world gets worse when volume dries up. This one survived a collapse and then shrugged off an explosion. That is the puzzle.

More isn’t more

In every other asset class, more liquidity is better. Stocks, bonds, derivatives, currencies, crypto. If the product is the trade, you want the lowest possible friction to execute it. Deeper order books, tighter spreads, more counterparties.

Prediction markets break this rule.

The reason traces back to something I wrote about in “Actually, the Price is the Product.” The vast majority of people who engage with prediction market prices never place a trade. They check the number, they see “72% chance of a rate cut” and inform their life. The primary consumer is the observer, not the trader, and prediction markets more approximate media than a trading product.

If the product is information, the relevant question changes. Instead of “how cheaply can I trade?”, it becomes “how cheaply can this market produce an accurate price?” Liquidity, in this framing, isn’t a benefit to the consumer, it’s more like a cost of production. And like any cost, you want to minimize it.

Last year, I interviewed Lost Zemblan, who is widely regarded as the best election trader in the world, for my podcast. He had this to say about the relationship between liquidity and efficiency:

[starts at 00:18:31]

Yeah, so certainly it’s fantastic. Volume’s gone up so much and like you touched on, you know, it’s actually crazy.

No, the market has gotten less sharp as volume has come in. It’s actually really ridiculous. Right. It’s gotten much less sharp. It’s much easier to make money now than it was a few years ago.

You know, prices are much less efficient because the new volume that’s coming in isn’t like the first adopters to prediction markets are going to be decently with it. Like, how do they find out about it? They’re probably interested in this stuff, interested in politics, you know, reasonably smart. Like, they do it as a hobby.

And then the next wave of people that we’re getting now just have no idea what they’re doing, so far as I can tell. And so they just make completely horrible bets. And so it’s definitely gotten less sharp over time, which is kind of crazy. And yeah, like you say, usually you think, oh, it gets bigger, it gets less sharp. But no, it’s not true, or, sorry, usually people say it gets more sharp. No, it’s not true. It’s gotten less sharp. Yeah.

As far as what goes in the future, I mean, I assume it keeps growing. I’m not really sure how big the potential market is compared to what. It’s just. I think the potential market for something like elections is possibly very big because it affects so many people.

Everyone has an opinion. Everyone it affects, certainly even a state governor, for instance, will affect the lives of quite a lot of people who are living in that state. And they’ll be getting a lot of information about it. Naturally, they’ll be seeing ads even if they don’t want to be hearing about it on the news. So I think there’s a lot of untapped potential.

I’m not exactly sure how much of that translates to people willing to bet a lot, but yeah, certainly I’d expect it to keep growing. And then as far as the market getting sharper over time, I think it’s just difficult.

I think it’s really just quite difficult for sharps to counteract all the flow of new money coming in, partly because there’s just fewer elections than there are sporting events, for instance. So it’s harder for someone to build up a repeatable model and a repeatable process and to be confident that they’re doing it the right way. And then at least, like me and the people I work with, I think we have a very substantial moat at this point just because we’ve been doing it for so long. We have all this data, we have all this just heuristics and things I think would be quite hard to replicate going in.

So I’m cautiously optimistic about elections staying relatively easy to beat. Maybe they get harder but relatively easy to beat for a while.

The information cost function

Here’s what actually happens when a prediction market produces an accurate price. Someone with relevant knowledge sees a contract trading at the wrong level. They buy or sell it toward the correct one. Their profit comes from the gap between the market’s price and reality. That gap is the bounty the market offers to anyone willing to contribute information.

We can formalize this. The implied bounty for correcting a mispriced market is roughly:

Information Bounty = Price Gap × Liquidity

Where the

• Price Gap is the distance between the current market price and the true probability, and

• Liquidity is how many dollars you can trade before moving the price to the true price.

On the other side is what I’ll call the

• Cost of Expertise: the minimum expected profit an informed person would need to bother showing up. Research cost, intrinsic motivation, opportunity cost, access friction, etc.

Price discovery happens when the bounty exceeds the cost of expertise. Rearrange, and you get:

Minimum Viable Liquidity = Cost of Expertise / Price Gap

• MVL: The smallest amount of liquidity a market needs to produce an accurate price.

The reason I divide the Cost of Expertise by the Price Gap is because if something is super-duper mispriced, then it needs less liquidity to get someone to correct it.

Below a market’s MVL, the bounty isn’t large enough to attract anyone with real information and the price is just noise. Above the MVL, additional liquidity doesn’t improve accuracy, it actually just raises the cost the next informed trader would have to pay to correct the mis-pricing. Yes, that “cost” is the trader’s PnL, but as an observer, I don’t really care about the trader’s PnL.

How fast accuracy saturates depends on the information ecology of the question. For elections, weather, and major sports, massive communities of intrinsically motivated people already exist. Their cost of expertise is pretty cheap for most markets… a few thousand dollars of liquidity produces a world-class forecast.

For forecasting an obscure regulatory ruling, or the result of a shale oil exploration in Alberta, Canada, or anything sufficiently niche, where the information is concentrated among a handful of specialists with high opportunity costs, you may need real money to draw them in.

Even for something as big as central bank decisions, tens of thousands gets you most of the way there. But the reason you can make billions of dollars for being a better forecaster of central bank decisions than others is because there just happens to be these unbelievably deep and liquid markets in bonds and currencies that move according to those decisions.

Those markets exist to enable countries to borrow and for corporations to do international trade, and they need to be deep and liquid for the world’s economy to keep humming. In my view, it’s almost coincidental that being good at parsing fedspeak is worth billions.

The price tag for truth varies by the topic and the opportunity cost of the traders who know the true price. Prediction markets are different because the price is the product, and that product is not actually very expensive to buy. For most questions that matter to everyday life, it is shockingly cheap.

A learning curve, not a liquidity curve

Let’s return to the CPI market. The framework above explains what happened with uncomfortable precision.

In the early days, the casual traders’ losses weren’t wasted. That flow created a bounty large enough to justify the sharp traders’ upfront investment: months of work building CPI models, sourcing component data, learning the BLS methodology. The cost of expertise was high, and the fish flow covered it. They were, in effect, venture capital for forecasting infrastructure.

As the models matured and the fish left, the sharp traders had already built their pipelines. The marginal effort of producing a CPI forecast each month was maybe an hour of updating inputs, and a few hundred dollars of expected profit was plenty of incentive. The marginal cost of expertise collapsed, and MVL collapsed with it. What once required hundreds of thousands of contracts of flow now sustained itself on a few thousand.

What the market converged to was a financially incentivized peer review process: a handful of calibrated experts, each checking the others’ work, with real money keeping everyone honest. The typical forecast missed the actual CPI print by less than a tenth of a percentage point.3

And the forecasters kept getting better. Accuracy improved steadily from 2022 through 2026. Typical errors fell from 0.131 percentage points down to 0.074pp, even as volume collapsed by 95% in the middle of that period.

Each CPI event was practice: learning which preliminary indicators mattered, which seasonal adjustments were reliable, where the BLS methodology had edge cases. The dog days didn’t stop this learning. If anything, the handful of serious traders left standing got better per event, since they were trading against each other with no noise from casual flow.

Then Robinhood happened. Millions of new contracts, fish returning to a market that had long since stopped needing them. The accuracy kept improving, but not because of the volume itself. Look at the learning curve: the biggest dots appear right when Robinhood surfaced the markets, but the accuracy gains come from the continued downward slope, not from the size of the dots4.

The new fish flow raised the stakes. More money on the table meant sharper traders had more reason to invest in refining their models further. But that refinement took time, just like it did in the early days. The scatter plot shows volume doesn’t instantly buy you accuracy. The learning curve shows what does: time, practice, and the incentive to keep getting better.

What this means

If MVL is low, the strategic priority for the prediction market industry shifts. Breadth over depth. 100,000 thin markets covering 100,000 questions produces more social and informational value than one deep market on the presidential election or the Super Bowl. The winning platform won’t necessarily be the one with the deepest order books. It will be the one that asks the most questions.

Subsidies, where they’re needed, may only need to be temporary. The CPI story suggests that thin markets need startup capital: enough early liquidity to attract the sharp traders who will build forecasting infrastructure. Once that infrastructure exists, the market sustains itself on almost nothing. Think of it as venture capital seed funding for information production, not an ongoing operating expense.

Prediction markets are the cheapest mechanism humanity has ever built for buying information about the future. A few thousand dollars can produce a forecast that outperforms million-dollar polling operations. The CPI markets I watched evolve over three and a half years taught me this firsthand - the forscasts we produced regularly topped all of the sell-side research teams at Goldman, Barclays, and Jeffries.

MVL for prediction markets is lower than you’d think. The information doesn’t require deep liquidity. It requires the right people, the right questions, and just enough money to make it worth their while.

It requires a reframing of how we think about liquidity in this new asset class that may change the way we interact with the concept of risk altogether.

The United States had all of fifteen active futures exchanges in 2019. That was the low point of a decades-long consolidation that had compressed the industry into a duopoly of CME Group and ICE, with a handful of niche survivors on the margins.

As of early 2026, there are twenty-six, with a dozen more applications officially pending1.

Seven new Designated Contract Markets were approved in 2025 alone. The forthcoming pipeline reads less like a regulatory docket and more like a venture portfolio: Sporttrade, Juice Exchange, Water Street Labs, Optex Markets, Ludlow Exchange. The most recent approval, Xchange Alpha, came through in 204 days, the fastest ever, close to the statutory minimum of 180 days. The industry hasn’t seen this pace of expansion since the regional grain exchange era of the late 1800s.

Something structural has changed. The prediction market exchange business, which for most of its history required Fortune 500-scale resources and years of regulatory patience, is becoming accessible to well-funded startups. The barriers are falling from multiple directions at once: a new CFTC chairman with an explicitly permissive posture, a collapsing cost curve for acquiring the necessary licenses, and a pending regulatory framework that could streamline the process further still.

The chairman’s signal

The expansion didn’t happen by accident. It has a political catalyst, and his name is Michael Selig.

Selig was confirmed as CFTC chairman on December 18, 2025, and currently serves as the commission’s sole member. No other commissioners have been confirmed. What the chairman wants, the chairman gets.

What Selig wants became clear on January 29, 2026, when he made his first public remarks at a joint CFTC-SEC summit and laid out a four-part agenda for event contracts: withdraw the proposed rule that would have banned sports and political event contracts, rescind the 2025 staff advisory warning exchanges to prepare for state-level prohibitions, direct staff to draft new affirmative rules, and reassess the CFTC’s participation in pending federal litigation to defend its exclusive jurisdiction over commodity derivatives. He called for the “minimum effective dose of regulation” and announced the “Future-Proof“ initiative, a comprehensive review of existing rules to modernize them for new entrants.

The contrast with his predecessors is sharp. Under Chairman Rostin Behnam, the CFTC fought Kalshi in federal court over whether political event contracts were permissible, and a proposed rule would have prohibited sports and political event contracts outright. Selig reversed all of it within weeks of taking office.

But the formal actions only tell half the story. On February 5, when Chris Christie tweeted that prediction markets are violating the law, Selig quote-tweeted him with two words: “Strong Disagree.” Two days later, when a reporter quoted AOC calling the rise of prediction markets “bad,” Selig responded: “CFTC officially back on track.”

The chairman of a federal regulatory agency is treating political opposition to the industry he oversees as confirmation that he is doing his job correctly. That is not how agency heads normally behave. For anyone considering filing a DCM application, the signal could not be louder.

Even just yesterday, Selig went further still, publishing an op-ed in the Wall Street Journal announcing that the CFTC would file an amicus brief supporting Crypto.com against state regulators in the Ninth Circuit, calling state litigation an attempt to 'undermine the agency's exclusive jurisdiction. You can read more about that legal fight here:

A collapsing cost curve

To operate a prediction market exchange in the United States, you have historically needed up to three separate federal registrations: a Designated Contract Market (DCM), a Derivatives Clearing Organization (DCO), and either your own Futures Commission Merchant (FCM) or a relationship with one. Each carries its own application process, compliance infrastructure, and capital requirements. The DCM alone requires financial resources equal to at least twelve months of projected operating expenses. You need the business substantially built before you even file.

Three things are changing, and together they’ve turned what used to be a Fortune 500-scale undertaking into something a well-capitalized startup can credibly attempt.

First, the regulated distribution layer has blown wide open. A year ago, the only way an American consumer could legally trade prediction markets was through Kalshi / Crypto.com directly, or through Robinhood. Now there are at least a dozen entry points, and the pace of new approvals is accelerating. For three decades, the number of registered FCMs in the United States did nothing but shrink, from 253 in 1995 to 62 in 20232. The prediction market wave has pushed that number back to 67.

PrizePicks became the first fantasy operator to receive FCM status from the NFA in September 2025. DraftKings got approved as an introducing broker in December and launched DraftKings Predictions before the month was out. Fanatics skipped the application process entirely by acquiring Paragon Global Markets, an already-approved IB, and had Fanatics Markets live in 24 states within weeks. Underdog and Sleeper both received FCM approval in January 2026.

The CFTC acknowledged the shift in June 2025, publishing an FCM FAQ in response to what it described as an ‘increased number of inquiries’ from new entrants drawn to event contracts.

There’s an even lighter-weight path: before getting its FCM, Underdog launched sports contracts in 16 states as a technology service provider to Crypto.com’s DCM, bypassing NFA registration altogether (though at the cost of co-branding and ceding customer ownership to the exchange).

Second, the exchange layer itself is getting cheaper. Time is the real cost for a pre-revenue startup, and the CFTC’s approval clock has historically run far longer than the 180-day statutory review period.

The table tells a clear story. Aristotle, QCX, and Railbird each waited between two and four years. Xchange Alpha submitted its application on July 10, 2025 and received its designation on January 30, 2026: 204 days, essentially at the statutory pace, the fastest ever.

One data point does not make a trend. But former acting chair Caroline Pham congratulated Xchange Alpha’s chief legal officer on LinkedIn and wrote, “Another approval in approximately 7 months,” suggesting the compression started under her tenure. If the Selig CFTC maintains that pace, the effective cost of entry drops by an order of magnitude in pre-revenue burn alone.

“We just processed, I think in a record time of two hundred days one of the more recent exchange applications”

• Chairman Selig on a recent episode of Odd Lots.

Third, the CFTC is actively considering whether to streamline the clearing layer. Kalshi proved you don’t need the full three-license stack: it operates as both a DCM and DCO under one corporate group, onboarding retail customers directly without an FCM in the loop. That’s a unique anti-pattern for derivatives exchanges. Neither CME nor ICE nor CBOE have ever had a consumer-facing app that directly onboards “self-clearing” exchange members.

But now, a December 2025 Request for Comment issued by the CFTC asked whether a separate registration sub-category should be created for retail-oriented (see: fully collateralized with no leverage) clearing organizations. The direction of travel is clear: the Commission is working to fit the regulatory architecture to the product, rather than forcing prediction markets into a framework designed for agricultural futures in the 1970s.3

The combined result is that every layer of the regulatory stack is getting thinner. The distance between “we have an idea” and “we’re taking our first trade” has shortened from years to potentially months.

It’s kind of like how SpaceX made it cheaper to launch satellites, and now there’s going to be tons of satellites. We may see a huge proliferation of the market structure in this industry.

What comes next

If the current trajectory holds, the number of active DCMs will almost certainly surpass 40 by the end of 2027. The United States is about to have more futures exchanges than at any point since the Great Depression.

Most of the new entrants are chasing sports event contracts, where the proven revenue sits. A few crypto derivatives play positioning for the Digital Asset Market Structure Act. Some are building toward novel categories. And at least a handful of the pending DCM licenses will likely be acquired before they ever list a contract, bought by distribution platforms that already have the customer base but lack the regulatory charter.

This is the part of the cycle that’s easy to forecast. What’s harder is what happens after the proliferation. Every wave of barrier reduction in American exchange history resulting in proliferation has been followed by consolidation. Whether prediction markets will follow that pattern is the most important structural question in the industry.

There are reasons to think this time is different: the products are heterogeneous rather than fungible, the markets are retail-native rather than institutional so the network effect of liquidity isn’t as strong, and the startup costs that cemented the CME’s dominance are falling.

The market structure question deserves its own treatment4. For now, the facts on the ground are clear enough. The regulatory barriers that defined this industry for decades are falling, the chairman of the CFTC is publicly cheering the process along, and the pipeline of new entrants suggests the market believes him.

For five years now, I’ve traversed the same conversation tree hundreds of times. Someone asks: “so, what do you do?” I tell them, “I bet on politics for a living.” The distribution of responses to this are polarized and bimodal.

Half the people light up. “Woah, that’s legal? How does it work? Tell me everything.” They’re curious, engaged, maybe a little envious. These conversations are fun.

The other half recoils. “What, how is that legal? You’re mixing money and politics. That can’t be okay.” These conversations are more interesting, because when I ask them to explain what exactly is wrong, they struggle to articulate it.

The first objection is usually “money in politics.” But SuperPACs spend billions actually trying to influence elections through ads, organizing, and voter outreach. Me putting a few hundred dollars on a prediction market doesn’t buy any ads, knock on any doors, or change a single vote.

The second objection is harder to pin down, something about markets affecting how people vote. If a candidate is trading at 75 cents, won’t that discourage their opponents from showing up? Or encourage bandwagon voting? Or maybe people will bet on their preferred candidate and then vote for them too, as if that’s somehow... bad?

I’ve never gotten a coherent answer. And I’ve asked a lot of times.

Think about it for more than thirty seconds and the logic falls apart. If a market shows Trump favored, does that make his supporters complacent or energized? Does it demoralize Democrats or motivate them?

The effects, if they exist at all, point in every direction simultaneously. They’d cancel out across millions of voters making individual decisions for idiosyncratic reasons. This is equally true of polls, which have shown frontrunners and underdogs for decades without anyone proposing we ban them for “bending reality.”

The reason people can’t explain what’s wrong is that there’s nothing wrong in the way they suspect. There’s no mechanism connecting “a price exists on a prediction market” to “the world changes because of that price.” The market is a thermometer. It doesn’t cause the fever.

This is the same problem at the heart of Kyla Scanlon’s recent New York Times op-ed, which argues that PMs create dangerous “reflexive loops” that shape reality rather than forecast it. The concern sounds sophisticated. But it makes the same error as the people I talk to at parties: it assumes a transmission mechanism that doesn’t exist.

How reflexivity actually works

Reflexivity is real. George Soros built a fortune on it. But it operates through specific causal channels that connect market prices to underlying fundamentals.

Consider stocks: when a startup’s share price rises from $50 to $80, the company can now raise capital on better terms. It can acquire competitors more cheaply using its inflated stock. It can attract talent with equity packages that look more valuable. The higher price creates real competitive advantages that make the company genuinely worth more. Market perception changes market reality through concrete mechanisms: cheaper capital, better hiring, strategic optionality. The whole point of the stock market is to reward companies and investors for being better businesses.

Scanlon invokes the infamous 2022 example of bond markets “firing” Liz Truss. This is accurate as history, but it actually undermines her argument about prediction markets. When traders sold UK gilts in response to her fiscally reckless mini-budget, yields spiked. Those higher yields made her borrowing plans literally unfinanceable. The bond market didn’t just forecast that her policies would fail; it made them fail by raising the cost of implementation beyond what the government could bear.

That’s reflexivity. The market action changed economic reality through a direct causal channel.

Now apply this logic to prediction markets. Millions of dollars flow into a contract on whether the Seahawks will win on Sunday. The price rises from 45 cents to 52 cents. Does this make the Seahawks more likely to win?

Of course not. The game will be determined by athletic performance, coaching decisions, and random bounces of an oblong ball. The prediction market is observing and forecasting, and it is not participating in any meaningful way. Russell Wilson doesn’t throw tighter spirals due to a price on a prediction market.

The same logic applies to political markets. A contract on Maduro’s removal trading at 8% doesn’t make removal more or less likely. The operational planning, the diplomatic maneuvering, the military assets, the decision-making in the Oval Office: none of these respond to how anonymous traders think about the probability.

The price is a forecast of the thing. Moving the price has no effect on the thing happening or not.

This is the fundamental difference between prediction markets and the securities markets where reflexivity actually operates. Stocks represent ownership claims on enterprises whose fortunes depend partly on market conditions. Prediction markets represent bets on events whose outcomes are determined entirely outside the market pricing them. There’s no feedback loop because there’s no channel for feedback to flow through.

This distinction matters because Scanlon’s entire “authorization” and “legitimacy” argument depends on prediction markets having causal influence over the events they price.

Scanlon writes:

That reported consensus influences how the public and people in politics perceive what’s legitimate. Each time markets process events before democratic institutions can respond, they gain more authority to do so next time. Odds become forecasts; forecasts become inevitability; inevitability becomes acceptance.

[...]

When markets process political events before democratic institutions like Congress can deliberate, market outcomes are treated as validation and permission for political actions. If traders have already bet and been paid, the action must be legitimate.

The thing is, validation requires some plausible mechanism of influence, and she doesn’t provide one.

Returning to the Maduro example: a trader betting on Maduro’s removal when the market was pricing it at 8%. Scanlon frames this as evidence that markets are “authorizing” political actions. But think about what that would actually mean. Is Trump supposed to point to an 8% probability as justification? “The market gave me an 8% mandate”?

And even if the market had been at 90%, would that change anything? Would “the market said so” suddenly become a legitimate basis for military action? Of course not. No serious person would accept prediction market odds as authorization for foreign policy, regardless of where the price sat.

The president was going to act based on his own calculus of political and strategic considerations. The prediction market was forecasting his decision, not granting permission for it.

The real concern hiding underneath

Strip away the reflexivity framework and Scanlon is making narrower, more defensible claims. Markets incorporate information faster than democratic institutions. Media outlets report prediction market odds without sufficient context. Both observations are true.

The Venezuela example deserves serious examination. If someone with non-public information about a military operation traded on Polymarket before the public learned what happened, that’s a genuine problem for the traders in the market, but actually a feature for the observers of the market.

The media literacy concern is similarly real but misattributed. When outlets report odds as authoritative forecasts without context, that’s a journalism problem. Prices are probability estimates derived from trading activity. They can be wrong, they can be thinly traded, and they deserve caveats. The solution is better reporting, not different market structure.

Scanlon is a thoughtful observer of markets, and I’ve heard versions of her concerns hundreds of times over the past five years. People sense that something important is happening as prediction markets grow more prominent. They’re right to pay attention.

The case for scrutiny should be built on what these markets actually do, not on a theoretical framework borrowed from an asset class where the economics work completely differently.

There’s plenty worth debating about how prediction markets should be regulated, disclosed, and reported. But that conversation has to start with what these markets actually are: thermometers, not thermostats.

On January 3rd, hours before dawn, U.S. Special Operations forces extracted Nicolás Maduro from his compound in Caracas. The world learned about it when Trump posted on Truth Social at 4:21 a.m. A Polymarket account called “Burdensome-Mix” had already been positioned for exactly this outcome.

The account was created just a week earlier. It wagered $32,000 on Maduro’s ouster when the market priced it at 7%, and by morning that position was worth $436,000. Burdensome-Mix had bet on exactly four events. All four were about Venezuela.

The political reaction came fast. Rep. Ritchie Torres introduced legislation to ban government employees from trading on prediction markets. The framing was obvious: insider trading.

And look, the instinct makes sense. The timing is damning, the profits are grotesque, and the whole thing has the stench of someone who knew what was coming. But the conversation that followed has been frustratingly shallow, importing assumptions and moral frameworks from securities law that don’t quite fit.

Prediction markets aren’t stocks, and the legal framework governing them is different. The economic purpose is different too. And the technological reality of blockchain-linked liquidity pools makes enforcement of insider trading prohibitions, as traditionally understood, somewhere between difficult and impossible.

This isn’t a defense of the Maduro trade. I’m trying to think clearly about what “insider trading” even means in prediction markets, why the concept doesn’t translate cleanly from equities, and why society is going to have to grapple with a new reality: prediction markets will attract informed money, and that might be fine.

Martha Stewart should have worked at Google, instead

The legal regimes governing insider trading in securities and derivatives are fundamentally different. Most coverage of prediction markets ignores this entirely.

In securities law, insider trading doctrine flows from Section 10(b) of the Securities Exchange Act of 1934 and SEC Rule 10b-5. The dominant theory is called “fraud on the market”: corporate insiders owe fiduciary duties to shareholders, and trading on material nonpublic information breaches that duty.

Courts have extended this to:

• “tippees” who receive information from insiders,

• “temporary insiders” like lawyers and accountants, and

• anyone who “misappropriates” information in breach of a duty to the source.

The SEC enforces this aggressively, bringing hundreds of cases per year. Martha Stewart went to prison. The doctrine is mature and has real teeth.

The CFTC operates under a narrower framework. Only the misappropriation theory applies to derivatives markets. “Fraud on the market” has no analogue in the Commodity Exchange Act. This distinction matters: derivatives markets exist so that commercial participants can hedge using their own proprietary information. A corn farmer trading futures based on his private knowledge of his own crop isn’t committing fraud. He’s doing exactly what the market was designed for.

The CFTC’s authority splits between a narrow statute and a broad regulation, neither of which has seen much use.

The first tool is the “Eddie Murphy Rule,” codified at 7 U.S.C. § 6c(a)(3)-(4). Added by Dodd-Frank in 2010 and named after the Trading Places scheme that would have been legal at the time, this provision specifically targets government information. It prohibits trading on nonpublic government information that could affect commodity prices, but only if that information comes from federal employees, Members of Congress, or judicial officers.

The scope is narrow by design: nothing about corporate information, private employers, or the vast universe of nonpublic information outside the federal government. Despite being on the books for fifteen years, the CFTC has never brought a single case under it.

The second tool is CFTC Rule 180.1, a broader anti-fraud provision modeled on SEC Rule 10b-5. The only court to fully articulate the standard, in CFTC v. EOX Holdings, set forth a four-part test: the defendant must have:

1. misappropriated confidential information in breach of a pre-existing duty of trust and confidence to the source;

2. intentionally or recklessly;

3. in connection with a commodity transaction;

4. for personal benefit.

Notice what’s absent from that test: “material nonpublic information,” the SEC’s term of art. Commissioner Caroline Pham has dissented from enforcement actions that conflate the two standards, arguing the Commission should stop “importing non-CFTC terms of art that are inapposite.” The relevant question under CFTC rules is whether information was stolen, not whether trading on it was unfair to counterparties1.

This distinction has practical consequences. Consider a senior data analyst at Google. She has access to internal dashboards showing both the company’s ad revenue trajectory and the “Year in Search” rankings before they’re published. She knows two things the public doesn’t: that Q4 earnings will massively beat expectations, and that d4vd will be announced as the most searched person of 2025.

If she buys Google call options on the earnings information, that’s textbook securities fraud under SEC rules. The information is material to Google’s enterprise value, trading on it harms outside investors who don’t have access to it, and the SEC will come knocking.

If she bets on d4vd on Polymarket, the legal picture gets murkier. The Eddie Murphy Rule doesn’t apply because this is private corporate information, not government data. Rule 180.1 might apply under the EOX Holdings test: she’s arguably misappropriating Google’s confidential information, in connection with a commodity transaction, for personal benefit. Technically, that’s a violation.

But here’s the uncomfortable truth: this enforcement void extends far beyond prediction markets. A Cargill employee who knows the grain harvest will miss expectations could trade corn futures on that information. Under EOX Holdings, that’s misappropriation too. She’s taking Cargill’s confidential data and converting it to personal profit, and so is her brother-in-law if she tips him off. The CFTC has brought a handful of cases against traders front-running their employers’ orders, but the broader universe of corporate misappropriation in commodity markets? Enforcement is essentially nonexistent.

Prediction markets are getting outsized scrutiny because they’re new and visible. The Maduro trade makes headlines. The grain trader who’s been doing the same thing for thirty years does not.

The Torres bill attempts to patch part of this gap by targeting government employees trading on nonpublic political and military information. That’s a reasonable extension of the Eddie Murphy Rule for the most troubling cases. If a Pentagon official bets on operations he helped plan, most people agree that’s wrong.

But the bill doesn’t touch the corporate scenario. Our Google analyst still falls into the same void that’s existed in commodity markets for decades. Technically against the rules, but never enforced. And maybe that’s fine, because prediction markets are fundamentally different from both securities and traditional derivatives.

The question isn’t whether informed trading violates some abstract duty. The question is whether we should care. Can our legal and moral frameworks articulate why it’s different without capturing the entire universe of informed trading?

Knowing things is the point

Why do we care about insider trading in the first place?

In securities markets, the answer is protecting capital allocation. Stock markets exist to direct resources to productive enterprises, and the whole system depends on a basic bargain: prices should reflect publicly available information so that outside investors can participate on fair terms. When a company generates value, its stock price rises, it can raise capital more cheaply, and investors are rewarded for funding it.

Insider trading breaks this bargain. If executives can trade on material nonpublic information, outside investors get picked off, capital allocation becomes a rigged game, and eventually people stop playing. That’s why we send people to prison for it.

Prediction markets serve a different function entirely. They don’t allocate capital to enterprises, they produce forecasts. The “product” is the price itself: a real-time probability estimate that synthesizes the beliefs of everyone willing to put money behind their view. And the value flows not just to traders, but to the much larger universe of observers who consume the price as information without ever placing a bet. Think of how many people check the odds on an election or a championship game just to know what the market thinks. The forecast is the point.

This changes everything about the harm calculus.

When an informed trader enters a securities market, she gains at the expense of less-informed participants, and the integrity of capital allocation suffers as a result. When an informed trader enters a prediction market, something different happens: the price moves toward truth. The Google analyst who bets on d4vd makes the “most searched” market more accurate. The Pentagon official who bets on Maduro’s ouster (setting aside the operational security nightmare for a moment) makes the Venezuela market more accurate. The information gets incorporated faster, and everyone watching that price benefits from a sharper signal.

The “victims,” such as they are, are the counterparties who took the other side. But what exactly did they lose? They made a voluntary bet based on their own assessment of the odds, and they were wrong. Someone else was more right. In a prediction market, this is the mechanism by which prices become informative. It’s a feature, not a bug.

You might still object that this is unfair. The insider knew and the counterparty was flying blind. Fair enough as an observation, but information asymmetry of this sort is endemic to every prediction market ever created. The PredictIt trader2 who called an energy lobbyist to confirm a tip, the oil trader with satellite imagery has data you’ll never see, or the trader who commissioned proprietary polling will outforecast you on elections.

Nobody calls this insider trading. We call it edge, and we generally think it’s fine because these traders are making prices more accurate, not less.

If you’re trading prediction markets, you should expect to face better-informed counterparties. Some will have better models, some will have better sources, and some will have access to information that would make your jaw drop. By law, no one owes you protection from this. When you click “buy,” you’re expressing a view that the current price is wrong, and so is the person selling to you. One of you will turn out to be more right than the other. That’s how these markets produce accurate prices in the first place.

The distinction that actually matters, I’d argue, is between forecasting and manipulation. A trader who bets on an outcome he can control isn’t bringing information to the market. He’s corrupting the outcome itself. The athlete who bets against himself and throws the game, the executive who buys credit default swaps and then tanks the company, the official who greenlights a military operation and then wagers on its success.

These cases are different in kind, not just degree. The problem isn’t that these people knew something. The problem is that they determined something, and then profited from a certainty they themselves created.

That’s a coherent line to draw. “Did you have an informational advantage?” is not, because in a market designed to surface information, having information is the whole point.

All liquidity is not KYC’d if any of it isn’t

Even if you wanted to prohibit insider trading in prediction markets, you probably couldn’t enforce it.

The reason is technical. Prediction markets now exist across multiple platforms with varying levels of regulation. Polymarket’s international exchange operates with pseudonymous accounts funded by crypto wallets, meaning users can trade without revealing who they are. Kalshi operates as a fully regulated CFTC exchange with know-your-customer requirements, meaning every user is identified, but also blurring the lines with tokenization.

Here’s the problem: these platforms don’t exist in isolation. Arbitrageurs constantly watch for price discrepancies between them. If the same contract trades at 40 cents on Kalshi and 45 cents on Polymarket, an arbitrageur will buy on Kalshi and sell on Polymarket to pocket the difference. This activity keeps prices aligned across platforms, which is normally a good thing. But it also means that adverse orderflow moves freely between them.

This means that, as long as exchanges continue to engage in mimetic competition, a non-KYC’d insider can essentially access all liquidity globally, KYC’d or not.

So, broader enforcement faces hard limits. The confluence of information markets making previously worthless information worth millions and blockchain technology enabling anonymity has set up a chessboard that makes much of this activity inevitable. Society will have to come to terms with a new reality: prediction markets will attract informed money, including money informed by what we’d colloquially call “inside” information.

For serious markets, narrow rules targeting government employees and outcome-controllers might be worth having. For trivial markets, the activity is probably harmless. And for everything in between, the honest answer is that we’re still figuring it out.

The price is the product. Sometimes the people making that price more accurate will be people who knew the answer before you did. Yes, that is facially uncomfortable to some and also, increasingly, unavoidable. Caveat emptor.

On April 2nd, Donald Trump announced the largest tariff hikes in a century. Markets cratered. The S&P 500 dropped 10% in two days. Trillions of dollars in value evaporated. Pundits called it a shock.

But it wasn’t… at least not to anyone watching the right screen.

Weeks before Liberation Day, a prediction market on Kalshi was pricing in exactly this outcome. The contract asked a question: will Trump impose large tariffs if elected? By late March, it was trading at 70–80%. The signal was there, sitting in plain sight, on a market with just a few hundred thousand dollars in volume.

I loved this market. In fact, I loved it so much that I wrote a whole post on how I priced and traded it, back before this newsletter was even created.

Adhi Rajaprabhakaran

What Kalshi’s trade-war market reveals about history and Trumponomics

Since the April 2 “Liberation Day” tariffs, Trump’s unhinged trade policy has spun through carve-outs, delayed deal-making and sudden reversals, leaving experts and consumers alike guessing who pays what and when. The constant stream of twisting headlines is anathema to anyone who wants to stay informed…

Read more

a year ago · 2 likes · 2 comments · Adhi Rajaprabhakaran

This is the promise of prediction markets: real money, real forecasts, beating the conventional wisdom. And in this case, it worked. The market called the biggest macro event since COVID, and anyone paying attention could have positioned accordingly to make billions.

But here’s the uncomfortable truth: that market was also badly designed. Its structure left almost all of its potential value unrealized.

This post is about what went wrong, and what a better design would look like.

A contract that tried to do too much

The Kalshi tariff contract had three problems that crippled its usefulness—not just as a forecasting tool, but as something people could actually trade, hedge with, or build on.

Flaw 1: It bundled two bets into one. The contract was conditional on Trump winning. If he lost, it resolved to zero regardless of tariff policy. This meant anyone holding a position through the election wasn’t just betting on tariffs, they were also betting on the election itself. If you were short at 30 cents expecting no major tariffs, Trump’s win sent you to a 60-cent price and a 50% drawdown—not because you were wrong about trade policy, but because you were exposed to a completely different event.

Flaw 2: The metric was bespoke, and thus illegible to most. Kalshi invented a custom resolution formula: tariff duties collected divided by total imports, a custom formula hosted on FRED. I know this firsthand because I was the market maker on this contract. Learning a new metric isn’t impossible, because you read the rules, pull the data, build your models.

But there’s no institutional knowledge to lean on, no muscle memory. And even if traders do the homework, observers don’t. Someone glancing at this market sees “72%” and has no clarity about what it really means. The price is the product, and the signal here is illegible to anyone who hasn’t done the work.

Flaw 3: Binary contracts can’t hedge continuous risks. To fit everything into a yes/no format, Kalshi set an arbitrary threshold: 6%. This is necessary to coalesce traders around a single point to trade on, but nearly useless for hedging. If you’re an importer and tariffs hit 5.95%, you lose your entire hedge. If they spike to 15%, you get the same payout as 6.01%. There’s no proportionality. What hedgers need are scalar payoffs: gains and losses that scale with the underlying move.

Despite all this, the market still called it. The forecasting worked. But this product was hard to price, hard to read, and hard to hedge with/against… so no one paid it the attention it deserved.

Impact markets are next-generation microstructure

The core problem was bundling. What traders and observers actually needed was a way to isolate the question they cared about: what happens to asset prices if tariffs hit? But Kalshi’s market structure squished something 3-dimensional into a single point.

This is the idea behind “impact markets.” Instead of betting on whether an event will happen, you trade what the effect of that event will be on real assets—conditional on the event occurring.

The structure is layered. Start with a base event: will Trump win? Then split into conditional universes: given Trump wins, what will the tariff rate be? Finally, add downstream impact: given tariffs at X%, what will the NASDAQ be?

These markets trade in parallel before the base event resolves. When reality settles, the other universe simply unwinds1. This lets you make trades like “short NASDAQ if and only if tariffs exceed 10%”, which express pure impact exposure, no election risk attached.

Traditional prediction markets tell you whether something might happen. Impact markets tell you what might happen as a consequence. For most observers, that second question matters more. “NASDAQ if tariffs hit 15%: down 8%” tells you what you actually want to know, in terms you already understand.

Any market has three stakeholders: liquidity providers, liquidity demanders, and price action observers. Good market microstructure design serves all three.

For market makers, impact markets mean cleaner risk. You can quote “NASDAQ given high tariffs” without warehousing election exposure. For speculators and hedgers, scalar payoffs mean positions that actually match your risk, not binary cliffs. For observers, legibility: instead of decoding “72% chance of tariffs above 6%,” you see “NASDAQ if tariffs hit 10%: 4,800.”2

That’s a screenshot that spreads… and a number your uncle understands.

A liquidity watershed

Prediction markets have a cold start problem. They’re isolated pools of liquidity with no connection to larger capital flows, which means liquidity stays thin and signals stay noisy.

Impact markets change this by creating channels to deeper waters.

Think of major asset markets—NASDAQ, Bitcoin, Treasury futures—as estuaries. Impact markets are the rivers connecting prediction markets to those bodies of water. A market maker quoting “NASDAQ given Trump wins” can hedge through the underlying markets: the prediction market and the index itself. Every convolution of risk can ultimately be flattened into positions in markets that already have deep liquidity.

This even allows spread trading on impact itself. If you think the market overprices NASDAQ’s sensitivity to the election, you go long one universe, short the other, similar to how oil traders trading the “crack spread“. That’s an entirely new category of trading activity that will spur order flow and liquidity in and of itself.

But liquidity flows further still, downstream to the base prediction market. A single prediction market can connect to many liquid markets simultaneously: NASDAQ, Bitcoin, gold, eurodollars. Each connection is another tributary, and the prediction market becomes a confluence point.

Then, the flywheel compounds. More connections mean more liquidity in the prediction market, and more liquidity means tighter spreads and higher-fidelity prices. That improved signal flows back upstream, informing the liquid markets it’s now connected to. The watershed then fills from the bottom up, creating a powerful omni-directional network effect.

Can we predict the next Liberation Day?

The Kalshi tariff market did something remarkable: on a few hundred thousand dollars in volume, it forecasted the biggest trade policy shock in half a century, and the biggest market event since COVID. The signal was real.

But the design made that signal hard to trade, hard to hedge, and hard to read. It bundled risks that should have been separate, used a metric nobody recognized, and forced a continuous variable into a binary box.

The tariff case is just one application. The same structure works anywhere you have an event and an asset people care about. What will Google be if Gemini tops the LLM benchmarks? What will Lululemon be if apparel tariffs hit 20%? Where will European defense stocks be if the US leaves NATO? What will Bitcoin be if the SEC approves staking for ETH ETFs?

Every policy debate, every product launch, every geopolitical risk—there’s someone who wants to know not just will it happen, but what will it mean.

Impact markets are a bet that we can do better—that the next Liberation Day can produce not just a forecast, but a price for what that forecast actually means. Legible to observers, tradeable for hedgers, connected to the deepest liquidity pools in the world3.

Liberation Day moved markets to the tune of over a trillion dollars. The information produced by impact markets is simply too valuable to leave on the table.

Every six weeks, the Federal Reserve meets to make a decision about interest rates. And, in lockstep, every six weeks economics observers are served up breathless reporting about what the CME FedWatch has to say about what will happen. Everyday people make life decisions based on this information… do I buy a car now? Used or new? Do I refinance my house? Should I put more into savings or the S&P500?

It makes sense - many billions of dollars flow through these markets, so it’s safe to assume that they are pretty efficient and are a decent forecast.

What strikes me about this is that when you actually go and try to trade these futures contracts, it’s practically impossible. Your broker will make you fill out a bunch of questionnaires and can reject you if they deem that you’re not sophisticated enough to trade futures contracts.

The reality is that the vast majority of the people who engage with Fed Funds futures as a product are not market participants, but market observers… observers of the information generated by the market participants’ trading activity.

This hints at something important: the primary product of prediction markets isn’t the ability to trade. It’s the information that trading produces. The price is the product.

Think about it: why do we care how the Nasdaq is doing today? Most of us aren’t actually trading those securities. We’re using stock prices as crude proxies for “how is the world doing”: vibes, dressed up as data. But stock prices answer questions about corporate earnings and capital flows, not the questions we actually want answered. What are the odds of a recession? Who wins the election? Will this deal close? Prediction markets offer something different: specific prices for specific questions.

As market observers, we are so desirous for any signal about the state of the world that we resort to numbers that only abstractly gesture towards it. Yet, most criticism of prediction markets focuses on the act of participating—the gambling, the speculation, the potential for addiction. But this misses where most of the value actually lives. The consumer surplus flows overwhelmingly to observers, not participants1.

A few weeks ago, I (perhaps too cynically) wrote about how the “fish” are the product when it comes to prediction markets and arguably the derivatives industry writ-large. Maybe I was wrong - the fish are just one of the products, the one that traders and market makers are interested in.

I believe the price itself is the primary product of prediction markets, and perhaps its ultimate purpose.

Skeptics are empirically correct, for now

Plainly, it is impossible to deny or minimize the outsized role of sports betting in prediction markets at the moment. It’s an empirical reality: today’s growth and revenue are about 85% sports, and companies themselves lean into sports due to an established culture and product-market-fit2.

This makes PMs an easy punching bag. It’s great fodder for the “gamblification of America” op-eds that are all too easy to pitch to the editors of high-brow media outlets. And on the pro-PM side, the missionary-esque shibboleths like “prediction markets are a truth engine” and “the future of finance” and a “new financial primitive substrate” look glib when by and large, most of the business is sports.

I believe the reality is somewhere in the middle. Here’s the thing: if you accept that the price is the product—that most value flows to information consumers, not traders—then the sports critique starts to look confused. It’s evaluating prediction markets by the wrong metric.

Millions of people check point spreads, over/unders, and game odds without ever placing a bet. They use this information to set expectations, talk to friends, decide whether a game is worth watching, or just satisfy curiosity. The information is the product, and sports information has an enormous audience.

Compare the audiences: Fed Funds futures serve a few thousand institutional traders and perhaps a larger circle of finance professionals and journalists. High-stakes per capita, but narrow reach. Sports odds? Massive reach, even if each individual decision is lower-stakes.

If consumer surplus comes from information consumption, then the relevant question isn’t “is this serious enough?” - it’s “how many people are consuming the information, and does it help them?” By that measure, sports prediction markets might be more valuable than Fed Funds futures, not less. The highbrow critique dismisses sports as “frivolous”, but frivolous questions that millions care about might generate more aggregate informational value than “serious” questions that few engage with3.

This isn’t to say that all information is equally consequential. Knowing the odds on the Chiefs-Ravens game won’t move the needle in your life as much as refinancing your mortgage. But the infrastructure that sports builds—the liquidity, the UX, the cultural familiarity with probability—enables the higher-stakes markets to follow.

Vice is often a prelude to virtue

There’s a historical pattern where stigmatized early demand accelerates general-purpose infrastructure:

• Games of chance seeded key ideas in probability, which matured into the mathematics behind forecasting and risk transfer.

• Adult content helped normalize paying for digital goods, advancing subscription mechanics and payments UX under real, privacy-sensitive demand.

• Bitcoin’s early adoption was tightly linked to darknet markets, effectively stress-testing censorship-resistant settlement in the harshest environment, long before “crypto” had mainstream legitimacy.

Sports betting and mention markets are prediction markets’ “vice wedge”: the stigmatized use case that funds the infrastructure for everything else. The progenitors of today’s platforms (a group that includes myself) believed in, traded, and even sued for non-sports markets like elections long before the current boom. The volume and profits comes from sports today, but the long term vision hasn’t changed.

Kalshi’s recent deals with CNN and CNBC are a bellwether. The pitch isn’t “come gamble with us”, it’s “here’s a probability you can show your viewers.” Kalshi will probably never recoup whatever they paid these networks through trading fee revenue generated by the exposure. The idea is bigger: the information layer, packaged for mass consumption to revolutionize news media.

Securities, options, insurance, consulting, politics, entertainment or culture may be next, and that’s just off the top of my head. If sports betting helps usher in more socially useful market structure, I can live with that.

A virtuous product, if we build it

The legendary Charlie Munger had a useful framework for distinguishing systems that improve civilization from those that degrade it. He called them “virtue effects” and “vice effects.”

His canonical virtue example was the cash register: a system engineered to make fraud difficult, which in turn raised the reliability ethos of the entire retail economy. The vice examples were casinos and gambling machines as systems designed to exploit compulsion.

Prediction markets sit uncomfortably between the two. The critics who see gambling aren’t wrong; those incentives are real and powerful. But there’s a virtue case too, and it’s structural: prediction markets financially reward being right and punish being wrong. Your P&L keeps score.

This is rare. Pundits can be confidently incorrect for decades and keep their platforms. Social media rewards engagement, not accuracy. PMs are different: they create hard incentives for calibration and updating when evidence changes.

But Munger would only credit those virtues if the system is engineered to be hard to defraud4. That’s the line between virtue instrument and vice engine. If prediction market prices are the product, then the product’s quality depends entirely on whether those prices can be trusted.

This is where the real work lies. Platforms must invest in market integrity, manipulation resistance, and robust settlement. The ecosystem around them—traders, analysts, journalists— must treat prices as epistemic tools worth explaining, not just numbers to flash on screen or potential payouts designed to hijack our dopamine receptors.

The price is the product. The question is whether we’ll build a product worth buying.

3

My friend Andrew Courtney took a quant trader’s approach to this in a recent substack post:

Whirligig Bear

Prediction Markets: Truth Engine or Casino? A Scoring Framework

Disclosure: I run Kalshinomics.com, which may earn Kalshi referral fees. I may trade event contracts on Kalshi and securities on other platforms. Readers should consider this relationship when evaluating my analysis. For educational purposes only, not…

Read more

6 months ago · 16 likes · 3 comments · Andrew Courtney

Kalshi recently announced a partnership with StockX that enabled people to trade on the future price of Labubus. Bloomberg’s inimitable Matt Levine had a minor and hilarious meltdown.

But beyond the absurdity of Labubu derivatives, Mr. Levine touched on something technical and in my opinion, under-discussed in the rise of prediction markets as an asset class:

Stop it. Why … why is this an event contract? Surely this should just be a regular futures contract? Like the natural way to bet on the price of a sneaker is to buy a futures contract on the sneaker. You buy a contract for future delivery of the sneaker today, and at maturity you get the sneaker or you settle up in cash. You pay $200 for the sneaker futures today, and if the sneaker ultimately trades at $175 or $250 you lose $25 or make $50.

He’s right to be exasperated. Prediction markets are forcing numeric outcomes—prices, percentages, temperatures—into binary YES/NO frameworks. It’s like trying to measure your speed with a series of “Are you going faster than X mph?” checkpoints instead of just looking at the speedometer. The tool fundamentally doesn’t match the task.

The same structural absurdity that makes Levine throw up his hands is crippling prediction markets’ ability to serve their most important functions: helping people hedge real economic risks and producing prescient information via the prices.

The problem is that this structure, while simple to understand, creates all sorts of weird incentives and terrible UX for market participants.

We’ll never see “hedging” behavior with this microstructure

Say you run a small business that gets hurt when inflation goes up.

The market offers you a contract: “Will inflation be above 3.0%?” It’s currently trading at 40¢, meaning the market thinks there’s a 40% chance inflation exceeds 3%. You buy $1,000 worth, so if inflation comes in above 3%, you get $2,500 back—a $1,500 profit to offset your compressing margins.

Great! Except... what if inflation comes in at 2.95%? Your costs still went up significantly—2.95% inflation still hurts!—but your hedge pays out exactly zero. The market said “above 3%” and 2.95% is not above 3%, so tough luck. You’re out your $1,000 hedge premium AND suffering from higher costs.

This is insane from a risk management perspective, and it’s not useful enough to hedgers for them to decide to accept the negative-EV that is the lifeblood of the derivatives industry.

The binary trap

The problems compound at every level. Want to express a view that inflation will be “highish”? You need to buy multiple contracts at different strikes, each with its own bid-ask spread and fees. A 3¢ spread on four contracts means 6¢ of edge lost just to put on a single view.

At low probability levels, this gets even worse. A contract trading at a 10¢ midpoint with a 4¢ spread means you’re paying 12¢ to get 10¢ of expected value—a 20% transaction cost.1

For market makers, it’s worse. In normal markets, they can hedge by trading the underlying asset. But what’s the “underlying” for a binary inflation contract? There isn’t one! Market makers are stuck warehousing risk with no way to hedge.

The binary structure unnecessarily amplifies small differences in beliefs. A maker can think the mean is at 4.0 and the taker can think 4.1, but binaries makes risk management extremely difficult. You can be wrong by a tiny amount but lose the maximum amount.2 This means market makers have to be super conservative with their spread and liquidity.

This totally freaking sucks if you’re the market maker in a new asset class that is trying to establish the liquidity and order flow flywheel and solve the cold-start problem that plagues any new tradable asset.

How did we end up here?

It’s bizarre - why are we doing these markets like this in the first place? To be honest, I’m not really sure. Kalshi was incepted as the “stock market for events” and thus born into this binary world. That might explain why they don’t have scalar event markets (yet).

That then begs the question: why haven’t CME, ICE, or CBOE done this then? Their bread and butter is scalar futures contracts. They could have been listing Beanie Baby futures back in the day, and Labubu futures today… right?

I think this is just the result of the strategies of the exchanges and regulatory path-dependence. Historically, the aforementioned derivatives exchanges have been largely focused on large, institutional market participants. Getting the right market microstructure, regulatory all-clear, and liquidity providers and demanders all together was paramount.

Kalshi came along and turned that on its head by pursuing a retail customer base that preferred straightforward ergonomics (it doesn’t get much simpler than YES/NO) and a large breadth of markets over depth of liquidity3.

Better microstructure. Better liquidity. Papa John’s.

What prediction markets really need for numeric events are... numeric payoffs. Revolutionary, I know. Instead of “Will inflation be above 3%?” you’d have a contract that pays out based on what inflation actually is. If inflation comes in at 3.1%, you get $31. If it’s 2.9%, you get $29. Simple, intuitive, and it actually works for hedging.4

The benefits cascade throughout the system:

• Hedgers can actually hedge (your payout scales with your risk)

• Expressing views requires one trade, not six

• Market makers can manage risk better and provide more liquidity

• Transaction costs drop dramatically

• The forecasting information created is higher fidelity

This is how futures markets have worked for the last century. But, what does the next century look like?

“Patrick Mahomes passing over 110 yards” is a binary market on both Kalshi and sportsbooks. What if we extended scalar logic to sports betting? How accretive and revolutionary would a more elegant microstructure be for the ages-old industry?

Or, culture? Imagine a scalar futures version of the Taylor Swift Spotify streams market, which is a binary ladder as of now. If it had the right microstructure, might it actually have a chance at becoming liquid enough that Ms. Swift herself could use the market to hedge the economic risk of her album flopping?

What if there was an index that produced a single scalar number by rolling up the market prices of a bunch of underlying events that actually are events, making something like the S&P 500 of Democrats? What if there was an ETF that accomplished the same thing, thus siphoning order flow and liquidity from the securities market into prediction markets?

The derivatives industry figured the microstructure for numbers a long, long time ago. What they didn’t have was the remit to tap into an entire world of measurable outcomes. That is perhaps the more important innovation of Kalshi’s strategy - not the YES/NO microstructure that defines prediction markets… for the moment.

I’ve seen people say that PMs are the hottest business space in the world outside of AI, and I have to agree. Exciting, but the noise can drown out the signal. If you can cut through the noise of bullpoasting, Twitter badge mania, and the incessant pace of headlines in this industry, there’s gold to be mined. Chris Fargis, who leads sports and prediction markets at Crypto.com, said it best in this post he wrote last week.

I graduated from college in the spring of 2023, and I knew deeply that I had to be in this space. I landed a job at Kalshi that summer and eventually became the Head of Operations. Even before prediction markets were a big deal, catching the attention of Kalshi was difficult. Nearly two and a half years ago, I wrote my first public piece for Kalshi on prediction markets, spotlighting my student loan forgiveness hedge. Even now, it’s a prime example of why prediction markets are essential.

Working at Kalshi was the ride of a lifetime, and the role I played taught me so much about what problems need to be solved at every exchange and which kinds of people tend to succeed in these roles.

Working at a Prediction Market Exchange

The most obvious way to break into the industry is to get a job at a prediction market exchange. An exchange is a platform operating in the United States as a DCM (designated contract market) or operating internationally without a license. An incomplete list includes Kalshi, Polymarket, Robinhood (which recently acquired a DCM), FanDuel Predicts, DraftKings Predictions, Crypto.com, and a significant long tail of crypto projects.

This choice will give you direct exposure to the ins and outs that make a PM exchange tick. There is a considerable breadth of roles to play at a PM exchange.

1. Engineering (the most important)

2. Product/Design

3. Markets Operations (the most unique to the PM industry)

4. Liquidity/Trading

5. Partnerships/Growth

6. Support

7. Legal/Compliance

8. Security/Surveillance

Every exchange will need to execute on these functions at a high level to compete in this space. In general, all of these roles become more difficult if you are operating under the regulation of the CFTC. It’s my opinion that working under the jurisdiction of the CFTC is the key to an exchange succeeding (specifically in the United States), and those are the exchanges you should be most excited to work for.

Find your niche

So you’ve decided that you want to work for a major exchange like Kalshi or Polymarket. Unfortunately, so has everyone else.

You will need to differentiate yourself from your peers. The pattern that I noticed at Kalshi is that the most successful applicants were people that have traded a decent amount of volume on the product before, created a useful tool for the community, or have an insane passion for prediction markets (and it doesn’t really track or look great if you’re in the latter without either of the former). This is critical if you want to work in a non-engineering role at an exchange. Shannon, on the Kalshi operations team, started out as a trader on Kalshi and then built a community to better understand weather markets.

If you’re an engineer, things are simpler (but not easier). Exchanges will differentiate themselves by how fast they can ship quality infrastructure and products in the space. If you are a cracked engineer, resist the siren’s call of Big Tech, with their high guaranteed comp and nice cafeteria. Instead, just ship some really cool shit at a prediction market exchange. Hone your craft, and if you’re good enough, the exchanges will end up fighting over you to help them deliver and win.

Working at a brokerage

A less obvious route is to work for a brokerage that has integrated with a prediction market exchange. Integrations can be done a number of ways, the most common being via an FCM (futures commission merchant) or a TSP (technology service provider). These include Robinhood, Webull, PrizePicks, Underdog, Coinbase (rumored), Fanatics, MyPrize, and TruthSocial.

All of these companies are offering prediction markets to their users sourced from either Kalshi’s or Crypto.com’s exchanges. Brokerages don’t create, trade, monitor, and resolve markets, but rather pick from the existing markets on exchanges like you would at a buffet.

Brokerages cut the original list I wrote in half.

1. Engineering (the most important)

2. Product/Design

3. Markets Operations (the most unique)

4. Liquidity

5. Partnerships/Growth

6. Support

7. Legal/Compliance

8. Security/Surveillance

The list doesn’t look so cool anymore, right?

Exchanges are where you tackle the full complexity—legal/compliance, market operations, liquidity, and security/surveillance. Brokerages offer a limited view of these critical functions. It’s the difference between scratching the surface and diving in.

The markets are the ultimate teacher

If you aren’t already trading regularly, the odds of getting your foot in the door at Kalshi (and likely any other exchange) are significantly lower. I firmly believe that you will never fully understand prediction markets unless you have spent time trading on them.

Does this mean that you should full port on the first market you see on Polymarket that looks mispriced? Absolutely not. You’ll learn a lot more by splitting $100 over 10 markets than by putting $10,000 in one market.

It’s more about the breadth of exposure. Try different categories and different market structures. Read the rules for all of them, and understand why some markets work and some don’t.

Try to find some edge; you’d be surprised by how far a little bit of research can go! Right now, I can’t trade on Kalshi (there is a cool-down period post employment), but I still enjoy finding a little bit of alpha and sharing it with the community, like this.

I’ve never cared for the Victoria’s Secret Fashion Show myself, but I do know this: the more niche and random a market is, the easier it’ll be to find an edge. Markets for major sports and economic events are the most likely to already have very accurate pricing. The more popular a market is, the more money there is to be had, making it more accurate, so they tend to be so.

Another structural reason is that the information that goes into pricing these markets is more canonical and easily accessible. I would be hunting for edges in markets that are both not very canonical but also decently popular, like TIME’s Person of the Year, Rotten Tomatoes, or mention markets. Dancing with the Stars just did $5m in volume, an indicator that culture markets are going to be huge, and no one has ever priced them before. Your PnL opportunity is your edge multiplied by the volume, so look for places where both of those factors will be huge.

Collaboration is key. Adhi Rajaprabhakaran, the 2nd trader at Kalshi Trading, the first market making firm dedicated to PMs, has said that 75% of his PnL has come from other people’s ideas. There are simply too many markets and too much information to be an expert on all of them. There will be creative edges and sophisticated models that are the fruit of working with traders smarter than you.

The best way to collaborate is to be generous. If you’ve found some edge in a market, don’t be afraid to share it with other traders (after getting the shares you want, of course). Join Discords and other group chats, talk to other people that want to take trading seriously, and learn from others that consistently profit.

Study the top traders that are more public about their trading. You can see every position that PredMTrader has on Kalshi, who has made over $70,000 trading on Kalshi. Tune into a live stream that he hosts sometime to learn more about how he trades and how he finds alpha.

Level Up and Provide Liquidity

Market making on exchanges is one of the biggest opportunities to profit and deeply learn prediction markets. Despite large institutions like Susquehanna and Jump Trading entering the space, there is still ample room to provide liquidity on “less-solved” events. Your advantage is your taste for opportunity and execution speed.

When starting out as a liquidity provider, target the markets with lower volumes that don’t receive the love they deserve. Focus on cultural events, mention markets, and niche political events (avoid major elections)... the more miscellaneous, the better. In these spots, developing your own sophisticated model or process can quickly make you the sharpest trader in the order book.

Pull a stunt or create content

For better or worse, we live in an attention economy, and prediction markets are poised to become a new block of the information diet food pyramid.

Getting eyeballs is something that matters very deeply to the people who run Kalshi and Polymarket. Why else would they make partnership after partnership, or develop an army of twitter poasters?

They understand that the information created by these markets, when distributed massively, will change the way we interact with all media. You actually don’t need their permission to be a part of this revolution. Adjacent News is an early experiment in market-driven journalism. Three Markets is making TBPN but for the prediction market space. Prediction News has been great at covering the daily drumbeat of industry news.

Build Something New

Finally, there’s the hardest and scariest option: to build something on your own. My personal suggestion is to not try and build another prediction market exchange. This space is getting more crowded by the day by incredibly well-capitalized competitors. The regulatory moat and network effects are daunting.

If you don’t have access to $>100m of venture capital, think differently. Instead of trying to compete with Kalshi and Polymarket, think about the second and third order effects of what they have wrought. What do exchanges, brokers, and traders need? How can you solve those problems? Is there a better way for people to consume the information created by these markets?

These problems will become apparent to you if you trade regularly, get a job at an exchange, or build in public.

There is so much opportunity to create the infrastructure and define the culture around this new asset class, and the sky is the limit. If this post convinces even just one person to create something net-new, or to become a trader, or to get a job at an exchange, I’d be thrilled. Thank you for reading, and please never stop predicting.

My old thesis was simple: as sports betting financializes through disintermediation, sportsbooks start will to look like bond desks. They accumulate massive, concentrated risk—imagine their liability if the Chiefs win another Super Bowl 70-68 in 3OT—and need an efficient way to hedge. Prediction markets seemed like the perfect instrument.

I also felt super validated when Underdog announced that they’d start trading on Kalshi. All of my predictionpilled marketcel dreams were really coming true!

But after learning more about how sportsbooks actually manage risk, the bank analogy breaks down. They aren’t financial institutions, and their risk management playbook is entirely different. Their tools are designed for maximizing margin within their own ecosystem, not for competing for flow and offloading risk to external markets.

Banks and sportsbooks wear risk differently

Banks take risk in giant, concentrated bites; sportsbooks eat it as a buffet of small plates.

On a bond desk, one $500 million trade instantly puts you on a mark-to-market rollercoaster. A tiny price move swings real dollars, and because there’s deep two‑sided liquidity on both that instrument and correlated instruments, the rational move is to hedge fast and get back to neutral.

Books live in a different world. They’re holding thousands of $5–$500 bets spread across 10–15 NFL games, dozens of player props, and staggered kickoffs. Yes, fans love favorites and overs, but cross‑game outcomes don’t resolve in lockstep. Mahomes over passing yards and Saquon over rushing yards aren’t the same or even correlated risk. The calendar itself diversifies the exposure.

“Everyone bets the over” isn’t “everyone bet the same over.” Public money diffuses across teams, totals, and props. A bank can be handed a risk cliff while the sportsbook gets a gentle rolling hill.

Competition for flow vs. captive flow

Banks live in a multi‑dealer world, and sportsbooks live in a single‑dealer app.

An institutional bond client pings many dealers at once, and then it’s a horse race to quote tightest, win the trade, and hedge immediately. Spreads are razor‑thin because competition is savage. If you don’t hedge, you’re gambling on microscopic margins. A trading desk is actually quite happy with a couple bps of edge on a huge trade.

Books don’t face that pressure because the typical customer is locked into the app they deposited in. There’s no RFQ carousel or parade of dealers shaving basis points. There’s no need to beat a rival quote so they maximize margin, especially on props and SGPs where edge is fat and UX sticky.

External hedging taxes fat margins. If props/SGPs hold ~6,000 bps, giving up 10,000 bps to lay off risk is a real haircut on the core business. Banks hedge to stay price‑competitive in thin‑spread arenas, but sportsbooks don’t live there. In a captive, single‑dealer experience, most weeks the optimal move is to keep the basis points and embrace the volatility.

Books keep them coming back

Banks and sportsbooks diverge most in what happens after a “win.”

A bank’s trading desk don’t hold their clients hostage. That capital leaves and may not come back for months, if ever. There’s no “we’ll earn it back next week” dynamic, so drawdowns are final in a very real sense.

Books have a different customer journey… A lot of winners recycle their winnings into next week’s slate. They come back for the Sunday dopamine hit, stack another same‑game parlay, and—on average—leak it back over time. A book’s losses, perversely, can be retention fuel: a close miss or a narrow customer win keeps the habit alive.

That loop lowers the effective cost of a bad week and weakens the case for hedging routine volatility.

I might not be totally, “generationally” wrong though

There are some instances where sportsbooks may hedge out risk on a PM like Kalshi.

At season endpoints and tentpoles—the Super Bowl, Finals, World Cup—you see more “I’m done” bettors who withdraw and churn. So then, hedging might start to make sense: correlated exposure (everyone bets the over/star players) plus higher cash‑out risk.

But on a typical week, the earn‑it‑back dynamic turns drawdowns into temporary dents, not permanent impairments, which nudges the optimal strategy toward keeping the basis points of edge and embracing the volatility.

There’s also in-play trading. When the game is live, the job looks less like “guess the right price” and more like “manage risk while collecting the spread” because the fair price is so volatile.

Also, a sportsbook might heavily promote a specific long-shot parlay (say, the “+10,000 Dave Portnoy Parlay Presented by Barstool”) for user-acquisition purposes, and then if the risk gets to be too much, they can RFQ it away on Kalshi’s parlay engine.

Bottom line

Sportsbooks aren’t banks. Week to week, they wear diversified, retail‑driven risk and monetize margin; they won’t need to optimize capital turns or dump exposure into prediction markets.

The corrected model is simple:

• Diversified calendar risk != thin‑spread bank logic

• Hedging is a tool, not a religion… save it for tentpoles and true in‑play risk control

• PMs win in sports on liquidity, price discovery, and distribution, but they aren’t really for hedging

I was wrong because I imported the wrong mental model. Updating it clarifies where prediction markets fit, when sportsbooks actually hedge, and why most of the time the optimal move is to keep the basis points of edge and just let it ride. The point is to gamble, after all.

For a long time, my favorite podcast has been Odd Lots by Tracy Alloway and Joe Weisenthal. What makes it so good is simple: they bring on people who really know their stuff, and they’re not afraid to get into the weeds. It’s curious, technical, and deeply human all at once.

That’s the spirit I want to bring to the 50¢ Dollars Podcast, a podcast about the markets that put a price on tomorrow. Prediction markets sit at the intersection of finance, politics, culture, risk-taking, and forecasting — a strange and revealing mix that shows how people actually think about the future.

My goal is to dive into the history, microstructure, and regulation of this space with the people who actually built it. The traders, founders, and regulators who’ve been in the arena.

The name comes from an old meme in prediction market twitter: selling dollars for fifty cents. It’s a reminder that even the most “rational” markets are shaped by incentives, design choices, and human behavior.

The podcast is recorded at and produced by the fine folk at Novig, a sports prediction market. This wouldn’t have been possible without them!

Check out the podcast here, available on all of the typical platforms as well as Youtube:

History doesn’t repeat, but it rhymes

For the first episode, I wanted to start at the beginning — with Ron Bernstein, co-founder of InTrade, the original real-money prediction market. InTrade was the prototype for everything that followed: an online exchange where people traded on elections, economic data, and world events long before “prediction markets” were a mainstream idea.

Ron and I trace the rise, fall, and legacy of InTrade — how it pioneered the field, the regulatory fights with the CFTC that ultimately shut it down, and how its DNA still runs through modern platforms like Kalshi and Polymarket. It’s a rare firsthand account of the early days of prediction markets.

After the fall of InTrade, Ron pivoted to crypto for a fruitful second act. He shares his thoughts on how crypto may define the future of prediction markets in a way that was truly surprising and illuminating for me.

I hope you find it educational, entertaining, and worthwhile. I’m genuinely grateful for anyone who takes 60 mins out of their busy day to listen.

Thank you!

Here’s a question that stumps most prediction market enthusiasts: Why didn’t Betfair Exchange—with its superior pricing, lower fees, and fairer model—dominate the UK sports betting market?

Recent data tells the story: Betfair holds just 2% of the UK online sports betting market, while traditional bookmaker William Hill commands 36% and Sky Bet maintains 18%. This stark disparity, a pioneering exchange with superior pricing capturing less than one-fifteenth the market share of its old-school competitor, shows how cultural entrenchment and familiar distribution channels trump technological innovation.

Twenty years later, it’s still a niche player in the UK while antiquated bookmakers dominate. Why? Because Betfair Exchange brought market structure innovation to a distribution fight.

The UK has a uniquely entrenched betting culture

In Britain, betting shops anchor every high street. Multiple generations learned to punt at the local Ladbrokes. These physical touchpoints, cultural habits, and familiar rituals created an unbreachable distribution moat. And as online sports betting rose, it piggybacked on the brand equity and familiarity of that system. Betfair’s market structure innovation bounced off this infrastructure like rain off concrete.

They asked punters to abandon these touchpoints for marginally better prices. That’s not how the fish swim. They are simply not that price sensitive1.

Critics watching prediction markets see another exchange trying to convert sports bettors. But a Designated Contract Market’s (DCM) exchange structure unlocks something Betfair never had: direct access to America’s financial infrastructure, which is really the world’s financial infrastructure.

CFTC regulation means integration with every broker, every trading platform, every fintech app. The same exchange model that limited Betfair in the UK hands a DCM the keys to global derivatives markets.

For prediction markets, a DCM exchange is just a means to the end of distribution, and Kalshi just picked the lock on the biggest distribution network ever.

The exchange model does have its flaws

When Betfair struggled in the UK, conventional wisdom formed instantly: exchanges are inherently niche, too complex, unsuited for recreational betting, too sharp, and a graveyard of liquidity. This dismissal became industry gospel, repeated without examination. But it misses the real reason exchanges struggled to win.

Here’s the uncomfortable truth: bookmakers make more money than exchanges. Traditional books profit from every losing bet, building 10-20% margins into their odds. Betfair just collect 2-5% commission on winning bets2. The math is brutal. Why run a low-margin exchange when you can print money as the house? How could you possibly compete on user acquisition and product velocity?

This economic reality, not market structure complexity, explains Betfair Exchange’s ceiling. They built a structurally less profitable business in a market with finite size and entrenched culture. Even if they captured every UK punter, they’d still make less per customer than Ladbrokes. Distribution challenges3 compounded this disadvantage, but the core problem was economic.

Earlier this year, Betfair ended it’s affiliate marketing program, a telling indicator of strategic retreat. In this industry, abandoning user acquisition is not to dissimilar to entering hospice care.

CFTC distribution makes it viable

Critics cite this as proof that exchanges can’t work. They’re thinking too small. The UK sports betting market handle is roughly £15-18 billion4 annually. Substantial, but still just 7-8% of the $260-300 billion global sports betting market. But even that pales against global financial derivatives markets: hundreds of trillions in notional volume. A tiny commission on those flows dwarfs any bookmaker’s wildest hold rate dreams.

This is a DCM’s edge: use exchange structure not as a business model choice, but as a necessary regulatory key that unlocks financial market infrastructure. Every broker, every regulated trading platform, every international market5. The same exchange structure that was an economic ceiling for Betfair in the UK becomes a passport to global capital when blessed by the CFTC.

CFTC exchanges sacrifice margin for scale. They’ll never match a sportsbook’s 10-20% take rate. But they don’t need to when they can tap distribution networks that are orders of magnitude larger than any sportsbook on earth. Would you rather have 10% of $100 billion or 2% of $1 trillion?6

The mistake is judging the vessel by the harbor. Betfair wasn’t wrong—they were just sailing in too small a sea.

In any exchange, the fundamental dynamic is simple: takers pay fees to trade immediately, while makers earn positive-EV and rebates for posting standing orders. But Wall Street’s traditional metrics miss what’s actually driving value in prediction markets.

Take Spruce Point’s widely circulated short report on DraftKings. They tried to calculate Kalshi’s “hold” – the traditional sportsbook metric of revenue divided by handle. But applying sportsbook metrics to exchanges misses the point.

Unlike DraftKings, Kalshi doesn’t take the other side of bets – it facilitates trading between users1, creating fundamentally different economics.

The fish are the product, and on parlays, the fish are the takers.

When DraftKings reports an 8% hold, that’s their cut of the fish flow – the systematic transfer of wealth from recreational bettors to the house. When Kalshi reports fee revenue of ~1%, that’s just their toll for facilitating a much larger transfer. The other 7% flows directly from takers to makers, invisible in Kalshi’s revenue but very real in economic terms.

The real product is the negative-EV flow itself, or the aggregate losses of takers that are the lifeblood of the entire ecosystem. This is what Citadel is paying Robinhood for—the infamous PFOF, or Payment for Order Flow— or the privilege of feasting on their customer base, of being the Brown Bear in the Alaskan river. That flow of losses becomes maker profits and exchange fees2.

Until now, nobody was actually measuring the fish on prediction markets. Traditional sportsbooks capture it all as revenue, making their unit economics opaque. Kalshi isn’t calculating and publishing these figures themselves either.

But prediction market parlays offer something unique: a controlled environment where we can finally cleanly quantify the fish in sportsbook terms. We can measure, with precision, exactly how much takers lose and where that money flows – not just to the exchange, but to every participant in the ecosystem.

The results reveal the true economics of modern prediction markets – and why comparing Kalshi’s fees to DraftKings’ hold is like comparing the NYSE’s revenue to Jane Street’s.

The perfect laboratory

Unlike most prediction markets where the same wager can be placed and unwound multiple times, Kalshi’s parlay structure makes round-trip trading nearly impossible. You can’t easily buy a parlay ticket and then sell it back – the combinatorial complexity and thin liquidity make it economically unfeasible3.

This constraint is actually a feature for our analysis. Every dollar of parlay volume represents pure directional betting – handle is simple to calculate directly. No need to wrestle with the “volume” number. There’s no noise from arbitrageurs, no day-trading activity, no sophisticated players gaming the system. Just users making directional bets and holding them to expiration.

For this analysis, I examined all parlay trading on Kalshi from September 29, 2025 (when parlays launched) through last night. The dataset includes every fill, with maker/taker flags, execution prices, fees paid, and settlement outcomes.

This data and technique lets us calculate the total economic transfer in the system4.

Finally, a “hold rate” for prediction markets

For the first time, we can calculate a true “hold rate” for prediction markets – the percentage of money that flows from takers to the broader ecosystem.

The formula mirrors traditional gambling metric: (Taker P&L × -1) / Taker Cost Basis, where the Taker P&L is the inverse of the sportsbook revenue, and Taker Cost Basis represents total “handle” wagered.

Across seven days of Kalshi’s parlay trading, we saw $5.6M of volume, $806k of “handle” and an overall hold rate of 11%.

What’s striking is that even in this transparent, competitive environment – where anyone can make markets or take liquidity, spreads are visible, and there’s no house edge – takers systematically lose. The exchange structure doesn’t eliminate the fundamental dynamic: less-informed traders transfer wealth to more-informed ones.

The data spans all parlay trades during since the launch of the product on Monday, capturing both high and low volatility periods. On the first day, the takers actually won in the aggregate. But as more data flowed in, the story is clear: the exchange and the makers are having a field day.

The bigger picture

The parlay data quantifies what Spruce Point missed: the total fish flowing through this platform.

The vertically integrated sportsbook model – where they’re simultaneously the broker, exchange, clearinghouse and market maker — creates operational bloat that obscures the core value transfer.

Kalshi’s parlay data proves exchanges can facilitate the same economic transfer – recreational bettors losing to sharps – without the overhead. No dealing with a state-by-state patchwork regulations. No cat-and-mouse game with profiling and limiting. Same fish, same flow, 90% less infrastructure and 10x the potential distribution through brokers in every country that allows futures trading.

Now that we can actually measure the fish, we can start asking whether the current ecosystem is catching them efficiently. Spoiler alert: it’s not.

Our gambit to get election markets for the 2022 midterms had failed because the CFTC blocked us. Trading volumes were low, growth was miniscule. We had to do a layoff, reducing our headcount from about 30 people to about 25. It might not sound like a big deal to most, but when you’re a team that small trying to do something this hard, it is devastating.

So we tried to realign the company with a company off-site trip to Joshua Tree. In the arid California desert, around a woefully inadequate campfire, we had some pretty difficult heart-to-heart conversations.

What are we going to do now? What’s the way forward? Do we triple down on this?

We did. Kalshi embarked on an incredible uphill battle, the battle of a lifetime: vs. regulators, vs. incumbents, vs. the non-believers… and despite the long odds, we won.

Now, on Monday, September 29th, in the Year of Our Lord 2025, two prediction market exchange CEOs are welcomed to Washington D.C. to face the very regulators that they sued or were investigated by. Here they sit, shoulder-to-shoulder, with the titans of the derivatives industry.

It is an incredible and awe-inspiring story.

So when Terry Duffy, the Chairman of the $95 billion-market-cap CME, the entrenched incumbent with all the resources you could possibly imagine, asked for a “level playing field”, I took it as a personal affront.

The playing field was never level

Mr. Duffy said1:

…They would be exempt and core institutions would not be allowed to do the same exact thing, or if they are, it’s definitely not on the same exact timeline. And if you’re not first to market, you might as well not go to market because you’re never going to be successful.

So I appreciate the innovation up on this panel. I think it’s critical to the future…

[…]

Maybe you can’t make it fit on the blockchain, but that’s not the problem of the core institutions that have been doing business forever. Allow us to do it just as well on the same speed.

So I really think it’s important that we have a level playing field, whatever it is. I always have said I will play on any regulatory platform you want.

[…]

I have an application that Craig is well aware of from 2015 that sat at the SEC for securities for six years. And I pulled it. Six years. I mean, how does that speed to market when other people want to do things overnight or in a one-day self-certification?

Or, whatever the case may be, we need to have a level playing field. So, if you want to talk about harmonization, you cannot take away this innovation exemption and say, “that’s okay, but let’s have harmonization against everybody”.

So, I think it’s really important to level set that. Love innovation, love it. Go forward. I just did a deal with FanDuel, as you know. I’m all in on this stuff. At the same time, you can’t exclude incumbent players in a different way.

Mr. Duffy is making two points here:

1. He’s asking the regulators to not “harmonize” regulation by cracking down on everyone: let the crypto, DeFi and event contracts thrive with innovation exemptions.

2. But at the same time, do it slowly so the “core institutions” and “incumbent players” don’t get left out!

What I’m hearing is, “Hey, CFTC, you need to slow down Kalshi and Polymarket to make it fair for us. Wait for me and FanDuel to catch up!”

Basically: “Ref, do something!”

So yea, the playing field wasn’t level - there’s nothing level about twenty-five lost souls in the desert versus the $95 billion behemoth. But Kalshi fought all the fights that no one else had dared to take. This new playing field was completely empty, so why is anyone surprised that Kalshi is now 10 laps ahead?

I find it hard to have much sympathy for the Goliath who was beat to a new asset class by small group of Davids.2

Not the fight I was expecting

I was excited to attend this roundtable to hear more about how event contracts can start crossing into the world of securities—arguably the biggest untapped use case for prediction markets. I’ve written about why merger markets could unlock real economic value across derivatives, so I went in hoping to hear concrete paths there.

I also thought there might be some fireworks around the supposed Tarek–Shayne rivalry.

Instead, a different tone was set right off the bat.

ICE’s Jeffery Sprecher, another titan of the industry, opened with a jab at Polymarket’s Shayne Coplan, sitting directly to his right, referencing the FBI raid. It felt like the schoolyard bullies punching down at the upstarts.

Mr. Sprecher said:

Entrepreneurs like Shayne and Tarek who have started relatively new marketplaces have been able to use those technologies to overcome some of the regulatory hurdles that I’ve mentioned. And I would say both gentlemen have gone through significant personal sacrifice to get to where they are today and a sacrifice that the public exchanges up here couldn’t tolerate.

We cannot tolerate receiving a Wells Notice. We cannot tolerate the FBI showing up in senior executive’s homes because our shareholders will not allow that.

I hope that not only can we find ways of harmonizing the CFTC and SEC, but across those two jurisdictions we find ways of harmonizing all the different innovative technologies and exchange ideas that are going on within the space.

From then on, the conversation that was supposed to be about how the SEC & CFTC will hold hands and sing “kumbayah” instead became about how the exchanges will compete with each other3.

So what exactly is Mr. Sprecher’s point here? He is implying that if it weren’t for ICE’s obligation to be so responsible, that maybe they would be the dominant player in a world-changing asset class. He is saying that Kalshi and Polymarket outcompeted ICE and CME because they were willing to take more risk.

But, this entire business is about taking risk. We’re in the business of trading, after all, so we should know this. Risk vs. reward, expected value, position sizing - it’s in our bones. It’s what defines us.

So if Mr. Sprecher had gone to his shareholders and said, “we want to do this new thing, and there’s a 10% chance of creating $100B of enterprise value”, would they have said “No, no, we don’t want $10B of expected shareholder value because we’re worried that the FBI might get mud on your carpet when they raid you”?

I don’t know. Maybe yes. Maybe the blame does lie with the shareholders.

But that is the proposition Kalshi made to Sequoia, and the proposition Polymarket made to Founder’s Fund, and, perhaps most importantly, the proposition both made to all of their employees.

And they said, “yes, that sounds like a great bet to me. Where do I sign?”

We were twenty‑five in the desert once. Now the caravan is bigger. The old guard can ask for harmonization; the new blood will create it by leading. We’ll see you on the playing field.

I’ve played tennis for over 20 years now. I’m a decent club amateur, a “good 4.5”, as they say. I can hang with D1 players without fully embarrassing myself. My forehand is struggling since I toy with my grip too much, my backhand has gotten quite decent after a ton of work, but my real strength is my serve. When I’m dialed, I’ve clocked it at 125mph. And once in a blue moon, I’ll blast one that I’m fairly sure that no one in the world would be able return.

Even so, I have <1% chance to hold serve against Jannik Sinner. Most people wouldn’t call us “peers,” because we load that word with skill and status.

But as soon as we step on court, he’s my counterparty. In that moment, we’re in the same arena, striking the same ball, and breathing the same air. That’s the sense in which he’s my peer.

That’s how I feel about trading.

Citadel is better than me at stock options like Sinner is better than me at tennis. But, I’m better than both in predicting how many senators will vote for a cabinet nominee. Both of them are my peers as soon as I buy a call option or toss a ball for my serve, and if either of them want to bet on how Susan Collins (R-ME) feels about Kash Patel, I am their peer as well.

You may disagree with the analogy. Fair. My aim isn’t to win a semantics fight; it’s to compare and contrast exchange market structure with a sportsbook.

I truly believe that prediction markets are more than just “a sportsbook with extra steps”, even if that is what looks like right now. This is how I think about “peer‑to‑peer” and why it matters.

Culture clash, and an apology

Before the nitty-gritty, let’s clear the air.

There’s friction as PMs collide head-on with sports betting. Some of it is vocabulary: “vig” vs “fees,” “hold” vs “PnL,” “the house” vs “market makers.”

But there’s something deeper here… a “vibe gap.” Sportsbooks are culturally everyman; PMs can read as intellectual or tech-elitist.

We’re not gambling, we’re predicting. Our foundational texts don’t come from card counters like Ed Thorp, but from academics like Phil Tetlock and Robin Hanson. We’re not “degens”, we’re “Effective Altruists” and “futarchists”. Our conference is not on the Vegas Strip, but actually in a quirky Berkeley backyard.

It can be really annoying.

So when someone from the PM world tells you that you’re wrong, implies that you’re too dumb to read a long tweet, that everything anti-Kalshi is really just FUD spread by the entrenched gambling industry, and that you don’t really understand this because this is actually High Fenance and you’re just a degenerate, it’s insulting. I get it.

On behalf of PM people: sorry. We’ve crashed your party and often failed to introduce ourselves with humility. We can and have to do better with tone and medium1.

That said, this humble Substack post is both an olive branch and a full-throated defense: PMs are “peer‑to‑peer” in a way that matters. I’m just going to try not to be a dick about it.

Intermediation is why exchanges aren’t just fancy sportsbooks

Prediction markets are now being discussed through a sports‑betting lens. Many smart bettors are used to a “house” model, so when they see pro market makers and sports on exchanges, it looks the same, smells the same, and feels the same. At that point, “peer-to-peer” can sound like marketing spin. I get why that rubs people the wrong way.

In online sports betting, DraftKings is your broker, exchange, clearing and counterparty. It’s fully vertical, very opaque, and ripe for abuse2. In PMs, these are separated: brokers like Robinhood take your deposits and route your orders, exchanges like Kalshi match them to others, clearinghouses safeguard funds in the process, and counterparties are anyone on the orderbook.

The concept is called “intermediation”, and it’s the backbone of modern financial markets. That separation of duties creates anti-fragility, and it’s necessary to facilitate the transfer of trillions of dollars of risk and value.

It is also what enables “peer‑to‑peer” trading.

“Peer-to-peer” is a spectrum of market structure

If “peer‑to‑peer” (“P2P”) means “you and I can face one another directly, without a house taking the other side,” then it lives on a spectrum.

At one end is the purest form: a cordial handshake bet with a friend at the bar—no intermediary, just price and settlement between you and I.

Next is a central limit order book (CLOB) exchange.

1. Anyone can make or take at any price they like

2. The matching engine pairs the best bid and offer deterministically, according to price and time

3. Your counterparty could be anyone—an amateur, a pro, an “institutional market maker”

4. Crucially, the trading rules are the same for all of us3: price-time priority, transparent depth, and no one can “limit” you for being too sharp.

Slide further away from “P2P,” and you hit the online sportsbook. Yes, their prices are informed by a global ecosystem of sharps, quants, feeds, and risk teams, but when you press “place bet,” your sole counterparty is the house, every time.

Keep sliding and you reach table games like blackjack, where you face the house through a non‑discretionary dealer, and further still you get slots, where your “counterparty” is literally a random number generator.

Ordered that way - handshake → exchange → sportsbook → blackjack → slots - the dispute isn’t whether exchanges are P2P, but how P2P they are relative to the “house” model. By structure, exchanges sit firmly on the P2P side4.

“Peer-to-peer” isn’t about skill

This is where the confusion creeps in. Anyone would reasonably conflate “peer” with “equal skill,” if only because that’s how the word is used colloquially. On an exchange, your counterparty might be better capitalized, faster, or more informed - just as Sinner is better than me at tennis.

But skill distribution doesn’t change market structure. The book is still open; anyone can quote or trade with you; and identity of the parties doesn’t alter the matching rule. In practice, professional participation often makes a better product for the user: tighter spreads, deeper size, faster fills, and visible prices that you can undercut or improve with your own orders, which is my point here!

If what you care about is outcome—did you get the best available price? could you rest at your price and get filled?—then the relevant axis is structure, not skill. If what you care about is the identity of your counterparty, then exchanges are not for you.

If we’re judging by orderbook openness, matching rules, and who can be your counterparty, prediction market exchanges are peer‑to‑peer in the ways that matter: you’re not captive to a house; you can meet any willing counterparty at a common price; and the path to better execution is the same for pros and weekend traders—post, improve, get filled.

Put differently: if you value sovereignty over your price and the ability to compete for liquidity, you will have to step on the same court as Jannik Sinner.

I want to expand this - if you have a strong POV on a topic related to prediction markets, say thoughts on market structure, an amazing trade that you made, or a sophisticated market-driven interpretation of recent news, please reach out!

With that, please enjoy Flup’s perspective!

When a tennis match never happened and Kalshi’s market didn’t yet void, I tried to take advantage and lost over $30k — in a very public manner at that — because I didn’t understand Kalshi’s rules. I come from a sportsbook background where “player didn’t play” usually means “void.” But Kalshi resolved the market at the “last traded fair price” and that cost me, dearly.

What does “last traded fair price” mean here? In Kalshi’s book, it’s the exchange‑determined estimate of the contract’s fair value immediately prior to the disqualifying event, typically computed from recent prints and resting depth over a defined window. That’s very different from sportsbooks’ “Did Not Play = void” norm.

Annoying? Yes. Useful? Also yes. It forced me to learn why exchanges can’t “just void,” and why player props turn a thorny edge case into a daily problem.

Why exchanges can’t “just void”

It seems simple, but the exchange plumbing makes it hard.

Quick example:

1. I buy 100 YES Falcons at $0.40 for $40 from John (John is now short YES, with $60 max liability).

2. I then sell 100 YES at $0.45 to Mary for $45. I’m flat and withdraw my $5 realized P&L.

3. Net state: Mary is long 100 YES (paid $45). John is short 100 YES (posted $60). The exchange is holding $45 + $60 = $105 to cover all outcomes.

If the Falcons win: Mary receives $100 (profit $55), John loses $60, and my $5 is already withdrawn. Books balance.

If the Bucs win: John receives $40, Mary loses $45, my $5 is already withdrawn. Books balance.

Now suppose the game is canceled and the exchange “voids” all positions. A void implies Mary gets back $45 and John $60 = $105 back. But $5 already left the system to me. Unless the exchange claws it back, there’s a hole. Scale that “hole” across thousands of round‑trips and you’ve created balance‑sheet risk amidst perverse incentives.

That’s why regulated, fully cash‑collateralized exchanges are allergic to post‑hoc voids.

Player props make this a lot worse. There are going to be hundreds of instances of players listed who do not end up playing in the game at all. Sportsbooks usually void those. Exchanges can’t do that safely. So “did not plays” (DNPs) can turn into tens of thousands of unhappy customers unless the grading rules are crystal clear and priced in.

Policy options, none of which are perfect:

1. Void with escrow of realized P&L. Void the market, but hold back realized profits from round‑tripped trades until the event fully resolves (played or canceled). Likely regulatory friction: this is de facto withholding customer funds, which may run afoul of CFTC rules and customer‑fund segregation norms.

2. Grade at “last traded” when DNP becomes known1. In theory it approximates EV “as if” the player played. In practice it’s fragile: thin books and fat‑finger prints, timestamp ambiguity (when exactly is “known” — first tweet, team report, or market halt?), and pre‑news drift that punishes early buyers.

3. $1‑or‑$0 regardless of participation (DNP = NO wins). Simple, operationally clean, and consistent with collateralization. But it clashes with sportsbook norms; I’d expect serious user backlash unless pricing and disclosures2 make the DNP risk obvious. You could argue that injury risk should always be “priced in” - the thing either happens or it doesn’t.

4. List props late (post official injury reports), then apply 2 or 3. This reduces frequency of DNPs but shrinks the trading window and some of the edge that attracts market makers. Still won’t catch warm‑up or back‑to‑backs scratches, but it’s a directional improvement.

The “last traded fair price” mirage

Even with a crisp formula, “last traded fair” can be unknowable in rumor regimes. Consider Jannik Sinner at Wimbledon: on 7/7, a widespread injury rumor briefly knocked his outright price toward ~10%. He never withdrew — and ultimately won the tournament. If he had scratched, what’s the “fair” settlement price? The rumor low tick (path‑dependent and manipulable)? A time‑weighted window (which one)? Or a model-based estimate conditional on playing, which is, by definition, unobservable once we condition on non‑participation?

This is the core problem: when participation is uncertain and news trickles in, prices reflect a changing mix of “plays” vs “doesn’t play” probabilities. Any ex‑post attempt to divine the counterfactual “as if he played” fair is arbitrary, invites dispute, and is easy to game around halts. If you need a rule that is objective, auditable, and robust to rumors, grade it 0 or 1. That pushes the DNP probability into the price transparently and eliminates post‑hoc counterfactuals.

Here’s what I’d do

If regulators bless it, (1) Void + escrow is the most intuitive for users and preserves balance‑sheet integrity.

If not, (3) $1‑or‑$0 is the right default: it’s simple, hard to game, and compatible with cash collateralization. While we’re at it, pair (3) with (4): list the props late to reduce DNP incidence.

Here’s how I’d message $1-or-$0 so users don’t revolt:

• “DNP is priced in” as a feature, not a bug. YES will often be cheaper here than at sportsbooks because if the player sits, NO wins outright. This is transparent, tradable risk — not a hidden trap.

• Disclose in‑app, in‑line, and in rules: “If the player does not play, NO settles at $1.00.” Make it flashing neon rainbow colors if it helps people notice.

• Show implied DNP component (e.g., an info tooltip: “This price reflects X% chance the player does not play”).

• Educate market makers: you must model DNP odds; that’s the point.

Currently, Kalshi goes with (2) “last traded fair”, but it’s my least favorite solution and it’s the reason I’m out 30 large.

Bottom line

Void rules are a solution to a balance‑sheet problem. If you can’t claw back realized P&L, you can’t “just void” without blowing a hole in the books. That’s why the only two defensible paths are:

• Best: Void with escrow of realized P&L (if regulators permit). It preserves user intuition without under‑collateralizing the system.

• Otherwise: Make DNPs an explicit part of the bet. DNP = NO, list late, halt on official status changes, and beat traders over the head with disclosures and a perhaps visible “DNP effect” in the price to really drive it home.

This isn’t anti‑user. It’s the opposite: predictable rules, priced risks, fewer rugpulls. I learned that the expensive way. Exchanges can avoid a thousand similar blowups by being painfully clear up front and operationally rigid when scratches hit. Do that, and player props stop being a daily customer‑service crisis and start behaving like the rest of an exchange: transparent, tradable, and impossible to game3.

You can learn more about what’s going on here in my article about the doctrine at question here:

TL;DR of today’s coverage:

• New Jersey leaned on the “sports ≠ swaps” line and the presumption against preemption.

• The judges kept dragging everyone back to the Commodity Exchange Act’s text—especially the CEA’s “exclusive jurisdiction” over trades executed on a federally regulated exchange.

• Kalshi’s best moment was a clean conflict-preemption example: a 50-state licensing overlay would make a national DCM unworkable.

• Kalshi’s scariest moment was when a judge pressed them on the legitimacy of some contracts. (But it’s not a big deal IMHO, as I write below.)

Net-net: serious grilling on both sides, no obvious loser, but as one lawyer in attendance said: “I’d rather be Kalshi, that’s for sure”.

What happened this morning

Both sides got 15 minutes to argue their case in front of a three Judge panel. The panel opened with some discussion of the pronunciation of “Kalshi” and then got straight to New Jersey’s three-part frame: (1) sports contracts aren’t “swaps,” (2) modern preemption cases (think Wyeth) caution deference to state police powers like gambling, and (3) Kalshi can’t win under the ordinary preemption tests.

You could hear skepticism from the bench that “sports ≠ swaps” settles anything; Kalshi’s attorney emphasized that the elected branches, not the court, ultimately draw the line. then pivoted back to the statute’s actual words.

Kalshi led with the statute: the CEA gives the CFTC “exclusive jurisdiction” over accounts and transactions “traded or executed on” a Designated Contract Market—which cordons off the on-exchange order book as a single federal lane. That’s the heart of a field-preemption story limited to DCM activity. Kalshi coupled that with the “event-contracts” provision (the public-interest check that tells the CFTC to consider things like “gaming” and activity unlawful under state law). In Kalshi’s telling, that language makes state illegality an input to a federal decision, not a roving state veto over on-exchange trading.

The opposing counsel said vigorously that the CFTC’s regulation is “not comprehensive”, to which I would point to this exchange on twitter:

Where conflict preemption bites

The sharpest practical moment: NJ’s attorney said Kalshi could do this if they simply registered with them as a gambling operator. But, New Jersey licensing rules require in-state–only wagering. That’s fine for sportsbooks, but it’s non-sense on a federal exchange where a single national order book clears and settles uniformly. Multiply NJ-only constraints by 50 and you get operational impossibility—classic conflict preemption.

IMHO, this moment betrayed either a fundamental misunderstanding or intentional ignorance of basic market structure realities. It was really not a good look for NJ and a slam-dunk for Kalshi.

Bench temperature checks

Signals cut both ways. A judge1 remarked that Kalshi’s argument was “significantly better than negligible,” which is not faint praise. You also heard a playful “ping-pong match” hypothetical about economic value, which is more policy administrability than doctrine. The persistent judicial move, though, was to force both advocates back to text and structure: What exactly is a “swap” here? Which contracts, precisely, fall outside the definition? And how should the court reconcile “exclusive jurisdiction” with Congress telling the CFTC to account for state illegality in its public-interest review?

Kalshi’s scariest moment, perhaps, was when a judge pressed their lawyer for examples of contracts that should not be permissible under the CFTC’s “public interest” test.

In what I think was a minor unforced error, the lawyer meekly suggested that there may be some kinds of “player props” that would be a bridge too far. What he should have said is that it’s not up to him or the court to decide that - the whole point of this is that this is the CFTC’s jurisdiction.

As Kalshi expands their market offering into player props, that utterance may come back to bite them. But not in any real, material way. While the members of the casino-shill-media-complex will parade it as an example of hypocrisy, hypocrisy is not the matter before the court.

How this fits the bigger doctrinal fight

The CEA speaks out of both sides of its mouth. On one side, “exclusive jurisdiction” over on-DCM trades sounds like a strong federal fence. On the other, the event-contracts rule (Rule 40.11) pulls state law into the CFTC’s analysis—fuel for courts that start with a presumption against preemption in traditional state domains like gambling (see Wyeth and company). Expect the written opinion to wrestle explicitly with field vs. conflict preemption and how far either extends.

We might even end up wading into the waters of whether Rule 40.11 is even constitutional or not, but that’s outside the scope of this humble substack post.

I’m bid 75@85 in Kalshi’s favor

(That’s trader-speak for: “I think there’s an 80% chance that Kalshi wins this, but have decently wide error bars.”)

If today is any guide, the panel is alert to both the CEA’s text and the real-world plumbing of exchanges. Kalshi’s clearest lane is: federal control on the order book, with states free to regulate sportsbooks, taxes, advertising, and everything in their lane—but not to bolt a 50-state licensing regime onto a DCM. New Jersey’s best lane is the modern caution against sweeping implied preemption. With a related appeal running in the Fourth Circuit, a split—and a Supreme Court date—remain very much on the table.

Here’s what the markets think right now:

What’s next?

Expect an opinion within a few months. The Fourth Circuit appeal on a similar record is running in parallel—divergent outcomes would raise the odds of Supreme Court review.

In the meantime, Kalshi will get to print millions of dollars every weekend as the NFL season develops.

Disclosure: I am not a lawyer, and I have a significant equity interest in Kalshi. Forget grains of salt - if you know what’s good for you, ignore everything I’m saying about this.

Federally CFTC‑regulated prediction markets have waded into sports, and state gambling regulators are livid. Since the Supreme Court overturned PASPA seven years ago, states have built a lucrative patchwork of rules and taxes for online sports betting (hefty entry fees here, per‑wager taxes there). This year, CFTC‑regulated futures exchanges began listing sports‑lookalike event contracts, trading on federally supervised order books rather than state‑licensed sportsbooks. States say that is illegal “sports betting” squarely within their jurisdiction.

The fight is another chapter in the storied history of a legal doctrine that has defined the power struggle between the federal government and the states: preemption. When and where does federal law “preempt” state law?

This chapter’s dispute turns on a single term in the Commodity Exchange Act (CEA) - “exclusive jurisdiction” - and whether it preempts state gambling authority over contracts traded on a CFTC‑regulated Designated Contract Market (DCM)1.

The story so far

Early in the year, exchanges started listing sports‑event contracts. Crypto.com rolled out a Super Bowl market, and Kalshi followed with vigor: it self‑certified a slate of sports markets under the CEA’s “list first, review later” process. The CFTC didn’t block them at the gate - not a gold star of approval, just an acknowledgment that, by default, self‑certified products can trade unless the Commission says otherwise.

States that have grown to enjoy the tax revenues generated by their bespoke regulatory systems moved quickly. Nevada and New Jersey sent cease‑and‑desist letters, saying that if it looks like sports betting and quacks like sports betting, you need a state license (and you need to follow state rules on who can bet, how you market, and how you pay taxes). Maryland went further and threatened enforcement, framing Kalshi’s markets as illegal sports wagering unless the exchange got licensed or geofenced the state.

Three federal district judges then read the same statute and split 2–1 in Kalshi’s favor. In Nevada (Apr. 9, 2025) and New Jersey (Apr. 28, 2025), courts granted preliminary injunctions, treating the CEA’s “exclusive jurisdiction” over on‑exchange trades as a strong signal that state gambling overlays can’t control trading on a federally regulated order book. In Maryland (Aug. 1, 2025), the court denied an injunction, starting from the presumption that gambling is a traditional state police power and concluding that state licensing can coexist with the CFTC regime.

Those rulings set up the appeals now in the Third Circuit (New Jersey) and Fourth Circuit (Maryland).

Preemption 101

Preemption is the Supremacy Clause in street clothes: when valid federal law conflicts with state law, the state law gives way. Courts use three buckets; and Kalshi only needs one:

1. Express preemption: Congress says “states, keep out.”

   1. Securities law does this: the NSMIA bars state registration/merit review of “covered securities” (with a carveout for antifraud). That’s why buying Palantir call options works the same way in Michigan and in Florida.

   2. There’s no similar magic sentence for event contracts, which is why everyone is reading the CEA with a jeweler’s loupe.

2. Field preemption: Congress meant to regulate the activity so comprehensively that there’s no room left for states.

   1. Examples: Rice v. Santa Fe Elevator2 (federally licensed grain warehouses) and City of Burbank v. Lockheed3 (aircraft operations). Both of these cases held that when Congress builds a uniform federal scheme, conflicting local rules often yield.

3. Conflict preemption: Compliance with both regimes is untenable.

   1. Two flavors of this one: (i) Impossibility: literally can’t do both, and (ii) Obstacle: state rules thwart specific federal objectives.

The CEA’s mixed signals

How does this apply to the CFTC? The crux—CEA § 2(a)(1)(A)—gives ammo to both sides.

If the litigation is a seesaw, the fulcrum is a handful of words. The first sentence of § 2(a)(1)(A) says the CFTC has “exclusive jurisdiction” over accounts and transactions “traded or executed on” a DCM.

Those verbs are the obvious hook: “exclusive” sounds like a ban on competing regulators, and “traded or executed” narrows the clause to on‑exchange activity. Read together, the provision cordons off the exchange order book as a single, federally governed space.

The law has a second wrinkle that cuts the other way: the event‑contracts special rule, aka 7 U.S.C. § 7a‑2(c)(5)(C) and Rule 40.11. Congress told the CFTC to run a public-interest check when products touch categories like terrorism, assassination, war, “gaming,” or “activity unlawful under State law.”

That phrasing drags state law back in: the special rule instructs the CFTC to consider state illegality when deciding whether an event contract is contrary to the public interest.

So: “exclusive jurisdiction” for on‑DCM trades, but a federal review that takes account of state illegality. Which governs when they collide?

Kalshi’s case

Kalshi’s core point is simple: § 2(a)(1)(A) gives the CFTC “exclusive jurisdiction” over trades “executed” on a DCM. On‑exchange trading sits in one federal box—listing, clearing, surveillance, access.

The special rule is a federal decision‑making prompt. Under this reading, Congress put the CFTC in the driver’s seat. State laws are relevant inputs to the CFTC’s public-interest analysis, but they don’t get to veto an on‑exchange product. If the CFTC permits a contract on a DCM, states can’t independently shut down on‑exchange trading of that contract.

In short: if Congress told the CFTC to police state-illegality concerns for on‑DCM contracts, that implies the CFTC, not 50 state agencies, makes the call for on‑exchange trading.

That’s a field preemption story limited to on‑DCM activity.

Judge Kiel, of the New Jersey case, writes:

Even if express preemption did not apply, the exclusive-jurisdiction language reflects an intent to occupy the field and the defendants cited no authority to the contrary.

Then there’s conflict preemption: specific state rules break specific exchange functions. For example, a state rule that voids wagers for in‑state users would wreck uniform clearing and settlement. Divergent per‑wager fees or marketing rules would fragment access and pricing.

It’d be unworkable in practice. These are the kinds of concrete harms courts look for in conflict preemption.

It’s worth noting that Judge Gordon (Nevada case) didn’t even bother considering conflict preemption because he felt the field preemption bar was met:

…because Kalshi is a CFTC-designated DCM, it is subject to the CFTC’s exclusive jurisdiction and state law is field preempted. I therefore need not address conflict preemption.

The states’ case

The states start from a simple, powerful default: gambling is a traditional state power, so courts presume Congress didn’t mean to wipe that power out unless the statute is crystal clear. That presumption matters here.

Textually, the CEA doesn’t scream “states begone.” § 2(a)(1)(A) contains a savings clause - a provision that preserves other laws unless the statute clearly says otherwise4 - and the event‑contract rule explicitly tells the CFTC to consider contracts that involve “activity unlawful under State law.”

States read that as Congress wanting state illegality to remain relevant, not to be nullified.

Practically, the states make two linked points: (1) when Congress wants to preempt state regulators it knows how to do it explicitly (see NSMIA in securities law), and (2) the special rule’s nod to state illegality shows Congress envisioned a federal decisionmaker taking state law into account rather than stripping states of enforcement power.

Courts that resist broad preemption note modern precedents (Wyeth5, Hillsborough, Oneok6) cautioning against inferring a field‑wide ouster from regulatory complexity alone.

The Maryland opinion takes that path, starting with the presumption against preemption of the traditional state turf of gambling, reading the savings language and event‑contract text together, and concluding that coexistence is plausible unless the exchange can show a concrete operational impossibility.

Judge Abelson, of the Maryland case, writes:

That plain text clearly reflects an affirmative intent to preserve state laws governing whether particular conduct is lawful or unlawful. The fact that Congress expressly authorized the Commission to prohibit particular categories of transactions as contrary to the public interest based on the fact that the conduct at issue would violate state law severely undercuts Kalshi’s suggestion that Congress intended to displace all state laws that would otherwise apply to transactions that fall within the scope of the CEA.

What’s gonna happen?

Kalshi’s text‑first case is strong for on‑exchange trades, but modern courts are wary of broad implied preemption in traditional state domains.

The “presumption against preemption” is real. If appellate judges want clear textual signals, the CEA’s mixed messages make this close.

The Third and Fourth Circuits will now decide. A split is plausible, which would tee up Supreme Court review. Tomorrow, the Third Circuit will hear the case, and we will gain a lot of signal from the oral arguments.

To that end, I’ve created several play-money Manifold markets to let my readers speculate on the case or learn from the price signals. Expect the prices to move drastically during and after the oral arguments tomorrow!

The stakes (pun intended) here are massive, both judicially and for the industry, and I think this case is going to go all the way to the highest court in the land.

Disclosure: I am not a lawyer. I’m just a prediction market trader who is used to diving into legal/political rabbit holes to understand an issue before betting on it. I am implicitly “betting” on this case because I have equity interest in Kalshi. So not only am I not a lawyer, I am probably biased as well. I tried my best to be impartial here, and I hope that shows, but for these reasons, please take everything written here with a grain of salt.