Parlays won’t scale until prediction markets solve this microstructure problem

Fully cash-backed trades prevent another FTX implosion, but makes market-making difficult

Parlays are the people's champ of betting: tiny stakes, blockbuster dreams. They're also a huge chunk of sportsbook revenue because people love stacking 4, 5, or even 10+ long-shot legs together and then overpaying for the longshot. That works great for FanDuel and DraftKings as it becomes an ever-larger share of their revenues.

It's trickier for CFTC-regulated prediction markets like Kalshi (for the moment). Every contract must be fully cash-backed the second it's traded - if you can win $100, someone's posting $100 in cold, hard collateral until settlement. No IOUs allowed.

Did you bring your calculator to class today?

Quick math: when a user stakes $S at +X odds, the counterparty must have $S × (X/100) of cash at risk. So a $5 bet at +100,000 ties up $~5,000 on the other side, because you get to buy ~5000 shares of YES (at 0.1¢ each). Do that a lot, and the numbers stop being charming. If you accept a 2 million parlay tickets in a day with a realistic mix of odds, the aggregate cash market makers would need to post runs to high hundreds of millions — and if the long tail gets fatter, billions.

In the plot, I slightly shift the distribution of parlay tickets to the long tail of extreme outcome bets.

The baseline model assumptions are hypothetical but “reasonable” for a given night of booking parlays for a large sportsbook, according to sports betting experts. Much thanks to @BigBuckHunterrr at NoVig for the modeling help.

Ok, posting over a billion bones is already a hard enough challenge, but here, the sensitivity is the point: small shifts toward the long shots explode collateral needs. Every extra 100,000 tickets in the ≥ +100,000 band at a $5 average stake adds roughly $500 million to the posted-cash requirement.

Sportsbooks don’t face this in the same way. They can comfortably live with latent, rarely realized risk in a way that market makers in a cash‑backed exchange legally and practically cannot. They’re not fully cash-collateralizing each parlay ticket. They run a house book with diversification across millions of wagers, can throttle stakes selectively, cancel “related contingency” exposures, and rely on the empirical reality that the catastrophic, all-legs-for-all-users-hit-at-once scenario basically never arrives.

That lets them comfortably price and promote +10,000 (1¢) and +100,000 (0.1¢) tickets without matching those worst-case payouts dollar-for-dollar in a vault. An exchange or its market makers can’t assume that luxury under full collateral rules.

This is why a cash‑backed exchange can’t simply mirror sportsbooks’ “pick‑your‑own” product without either (a) sharply limiting stake sizes in the tail, (b) funneling users into a small menu of pre‑made parlays, or (c) getting permission to margin portfolios instead of ponying up for every worst‑case payout up front.

The scary, dirty M-word

“Margin Call” is one of my favorite movies, even though I’ve never actually sat down and watched it. I’ve gleaned the gist of it by watching clips on YouTube… Zoomer-brain tendencies aside, the movie does a great job of scaring the crap out of you and is emblematic of how the concept of margin and leverage is politically charged in a post-2008, post-FTX world.

Me explaining to my girlfriend how the CFTC’s strict clearinghouse regulations are the reason why LedgerX was totally unaffected by the FTX implosion. Source.

Now combine that with the concept of sports betting and you have a nice little recipe for congressional hearings.

That said, it could work:

Give a few trusted market makers a line of credit, and the exchange doesn’t have to sit on every worst‑case dollar for every absurd parlay. Practically, the exchange would vet a handful of big counterparties - hedge funds, prop desks, established brokerages - sign legal credit agreements, set hard intraday limits and P&L triggers, and let those firms quote deeper sizes in the tail. The exchange stays the arbiter: it doesn’t take the bets onto its balance sheet, it merely treats some incremental exposure as contractually credit‑backed rather than cash‑backed in the account ledger.

This is appealing because it’s fast and surgical: liquidity arrives where it’s needed, you just create capacity in places market makers already live. Practically, it buys the exchange breathing room and lets real users get the same neon‑jackpot UX they expect from sportsbooks. But make no mistake, it concentrates risk. If those counterparties fail, the exchange (and customers) are exposed.

We saw how this plays out in the worst possible way with FTX: private credit lines and opaque interdependencies turned into systemic failure. That history matters. Any credit‑line program needs ironclad guardrails: audited capital tests, conservative haircuts, realtime reporting, instant kill switches, and legal arrangements that make recovery practicable rather than aspirational.

So credit lines are a pragmatic near‑term patch - they buy liquidity and time - but they are not the best possible solution IMO. They trade immediate UX gains for concentrated counterparty risk and heavy regulatory scrutiny.

The solution I have in mind is better, but involves a bit more math…

A V12 risk engine

The clever way to stop tying cash is to stop pretending every ticket’s worst‑case has to sit in a vault separately. A “risk engine” is bookkeeping with imagination: aggregate a market maker’s entire book, enumerate the realistic resolution states that actually matter, and hold collateral equal to the maker’s worst net loss across those - not the arithmetic sum of every standalone payout.

That simple change (portfolio worst‑case instead of per‑ticket worst‑case) is the difference between billions of idle collateral and the kind of capital efficiency that lets an exchange offer the same neon‑jackpot UX without forcing market makers to back up an army of Brinks trucks.

Let's make this concrete. Right now, if you're a market maker quoting parlays, your world looks like this:

Every parlay requires its own dedicated pile of cash, sized to the worst-case payout. A $12,000 max payout parlay needs $12,000 in the vault, a $8,000 max payout needs another $8,000, and so on. It's simple but wasteful- especially when you notice something interesting about these particular parlays...

Look at the possible states of the world. If Team wins, Parlay B can't possibly pay out. If Team loses, Parlay A can't pay. We're holding $20,000 against two events that can never both happen. A risk engine would recognize this and do something clever...

Instead of blindly adding up worst-cases, we can look at what's actually possible. The true worst case is one parlay hitting ($12,000), never both.

That means $8,000 of that capital is just sitting there, serving no real purpose except satisfying a blunt regulatory requirement. So the exchange can and should just give the money back to the market maker.

Scale this thinking up to millions of parlays and billions in posted cash, and you start to see why this matters: smart collateral management could unlock the same user experience without requiring market makers to back up that army of Brinks trucks.

This is the simplest possible example… the combinatorics can explode with every leg you add. It’ll be a pretty incredible feat of engineering to ship and scale this¹.

Then there’s the regulatory aspect: it will be prudent to launch it in lower leg parlays at first and scale it very carefully as a public stress‑test regimen to prove you didn’t miss a plausible margin risk catastrophe.

This is harder than “give SIG daddy’s credit card,” but it’s the principled fix. It reduces capital needs without concentrating credit risk in a handful of counterparties, and it’s the sort of thing a regulator can understand and sign off on if you publish methodology, backtests and governance.

Market structure informs strategy

Different sportsbooks adopt different strategies. Some focus on delivering the best liquidity for those medium-risky parlays, and some really push their users to longer and longer odds. In the case of a prediction market exchange, they are limited by the current state of margin rules and tick-width².

That segment of the base that loves +100000 (or 0.1¢ on a PM) bets won’t be won over. Traditional sportsbooks will have this advantage over prediction markets until either an exchange can deliver margin or a risk engine or, in the best case scenario, both.

1. That said, my former colleagues at Kalshi are absolutely cracked. The engineers there have repeatedly shipped impossible things inconceivably fast. I wouldn’t bet against them here.

2. For the sake of brevity, I didn’t get deep into tick-width here. The idea is that since the minimum price of a share is just 1¢, you can’t actually book a trade that is +100,000 or whatever. You need deci-cent pricing (e.g. 0.1¢ shares).

50CD understands that this microstructure issue will not be a great hurdle and will probably be fixed in the near future. So that point will become moot soon. Besides, if an exchange ships this feature, they will really need some margin/risk solution to let meaningful amounts trade in that sub-1¢ range.