This is an interesting theory. One potential area for future exploration: once you're past a given minimum liquidity threshold in a market, is the informational value *inversely* correlated with liquidity?
Here's what I mean: one potential reason Kalshi markets may be more informationally accurate than Fed funds futures is that, at sufficiently deep liquidity, people are using Fed funds futures in order to hedge, not in order to get a prediction "right."
If that's true, it's not really the case that "you only need a small amount of liquidity to be very accurate." It's actually that, at sufficiently *large* liquidity, your markets (eg Fed funds futures) become less accurate because they are big enough to be used in other ways that wouldn't make sense (for most large financial institutions) on low-liquidity markets like Kalshi.
if liquidity begets liquidity is flipped over its head, then is the sourcing of niche experts the fastest and framing questions most clearly the moat? would this not turn pms from financial business into some sort of a media or even logistics if you will typa business
Great piece. Your "Minimum Viable Liquidity" framework changed how we think about market selection. We weren't thinking of it prior to finding this post.
We built autonomous AI agent prediction market skills on SerenAI that implement MVL as a pre-trade filter across three Polymarket strategies:
For example, we built a maker rebate bot, a paired basis maker, and a directional scanner. Markets where spread < 50bps AND book depth exceeds 50% of daily volume get dropped before the agent spends compute or capital. The insight that over-liquid markets offer no edge for makers. We're aligning with your insight that more liquidity is not necessarily more in prediction markets. Our filter enforces this check and removes non-qualifying markets.
The skills run 90-to-270-day backtests with order-book-aware fills, pessimistic spread decay, and inventory guards before any live execution. MVL scoring happens on existing CLOB data.
Would love to connect! We are building the infrastructure for agents to autonomously discover and trade prediction markets, and your framework is now a core filter in the pipeline.
Interesting concept. For long I've thought that PMs should earn from allowing institution to sponsor liquidity awards on specific questions by taking a cut out of it vs trading fees. Example: CDC sponsoring rewards on measles markets. And based on your thesis the MVL is not that high! Makes the whole approach entirely viable, especially with breadth vs depth approach.
This is an interesting theory. One potential area for future exploration: once you're past a given minimum liquidity threshold in a market, is the informational value *inversely* correlated with liquidity?
Here's what I mean: one potential reason Kalshi markets may be more informationally accurate than Fed funds futures is that, at sufficiently deep liquidity, people are using Fed funds futures in order to hedge, not in order to get a prediction "right."
If that's true, it's not really the case that "you only need a small amount of liquidity to be very accurate." It's actually that, at sufficiently *large* liquidity, your markets (eg Fed funds futures) become less accurate because they are big enough to be used in other ways that wouldn't make sense (for most large financial institutions) on low-liquidity markets like Kalshi.
We see a slightly different version of this on Pinnacle (long the preferred bookie of sports betting sharps). Their closing lines are sometimes NOT the most efficient/accurate, because people are using Pinnacle as one leg of an arbitrage play: https://www.football-data.co.uk/blog/wisdom_of_crowd_betting_system_closing_odds.php
I discuss pieces of this in my own post on Pinnacle from a few months ago: https://networked.substack.com/p/a-view-from-the-pinnacle
if liquidity begets liquidity is flipped over its head, then is the sourcing of niche experts the fastest and framing questions most clearly the moat? would this not turn pms from financial business into some sort of a media or even logistics if you will typa business
yea it’s kinda like consulting
I am loving your Substack.
Great piece. Your "Minimum Viable Liquidity" framework changed how we think about market selection. We weren't thinking of it prior to finding this post.
We built autonomous AI agent prediction market skills on SerenAI that implement MVL as a pre-trade filter across three Polymarket strategies:
For example, we built a maker rebate bot, a paired basis maker, and a directional scanner. Markets where spread < 50bps AND book depth exceeds 50% of daily volume get dropped before the agent spends compute or capital. The insight that over-liquid markets offer no edge for makers. We're aligning with your insight that more liquidity is not necessarily more in prediction markets. Our filter enforces this check and removes non-qualifying markets.
The skills run 90-to-270-day backtests with order-book-aware fills, pessimistic spread decay, and inventory guards before any live execution. MVL scoring happens on existing CLOB data.
Would love to connect! We are building the infrastructure for agents to autonomously discover and trade prediction markets, and your framework is now a core filter in the pipeline.
Our Skills for Polymarket are: https://github.com/serenorg/seren-skills/tree/main/polymarket
Interesting concept. For long I've thought that PMs should earn from allowing institution to sponsor liquidity awards on specific questions by taking a cut out of it vs trading fees. Example: CDC sponsoring rewards on measles markets. And based on your thesis the MVL is not that high! Makes the whole approach entirely viable, especially with breadth vs depth approach.