Prediction market liquidity: why thin markets jump 10 points

You open a market you have never seen before. It says 41¢, you think the true figure is nearer 60¢, and you put in a modest bet. The price jumps to 55¢. Nothing happened in the world. You simply walked into a market where very little money was standing by, and your own trade was most of the market.

That is the short answer to why a thin market moves ten points on one trade: the price is set by whatever orders or pool of shares exist at that moment, and when there are few of them, a small trade uses them up. The rest of this post explains the two main ways prices get set, what slippage costs you, why headline volume can mislead, and how to read a market that three people have touched.

What does liquidity actually mean in a prediction market?

Liquidity is how much you can trade before the price moves against you. A liquid market lets you buy or sell a meaningful amount close to the quoted price. An illiquid one quotes a price that only holds for a very small trade.

Polymarket's documentation describes the order book as "a list of all open buy and sell orders for a market" and says "tighter spreads mean more liquid markets." The same page warns that "large orders may move the price significantly." The spread tells you the cost of a small trade, and the depth tells you the cost of a large one.

How does an order book set the price?

In an order book, buyers post the prices they will pay (bids) and sellers post the prices they will accept (asks). A trade happens when someone agrees to meet the other side. The gap between the best bid and the best ask is the spread.

If the best bid is 34¢ and the best ask is 40¢, a market buy fills at 40¢ and a market sell fills at 34¢. Polymarket's documentation says the displayed price is the midpoint of that spread, so this book would show 37¢. It also says that if the spread is wider than $0.10, the last traded price is shown instead. Neither is a price you can necessarily trade at.

How does an automated market maker set the price?

An automated market maker (AMM) is a pool that is always willing to trade, with the price computed by a formula rather than negotiated.

Robin Hanson's 2007 paper, Logarithmic Market Scoring Rules for Modular Combinatorial Information Aggregation, in the Journal of Prediction Markets, is the standard reference for one design. Hanson describes a market scoring rule as "a continuous inventory-based automated market maker" with a zero bid-ask spread for infinitesimal trades. For the logarithmic rule, the paper gives an exponential form with a parameter b, the dial for how far a trade moves the price.

Here is that idea in numbers for a yes-or-no market, using the standard logarithmic formulas (my arithmetic, not a figure from the paper). The price of YES is 1 divided by (1 + e raised to the power of minus the net YES shares sold, divided by b). With b set to 100, the market starts at 50¢. To lift the price to 60¢, the maker has to sell about 40.5 net YES shares, and the buyer pays about 22.3 in total, an average of roughly 55¢ a share. With b at 1,000, the same shares barely move it. A bigger b is a deeper market, and someone pays for that depth.

The other common design is the constant-product pool, borrowed from decentralised exchanges. The pool holds YES and NO shares, and a trade must leave the product of the two counts unchanged. Take 100 YES and 100 NO (price 50¢). You spend 50. The pool becomes 150 and 150, then YES is removed until the product is back at 10,000, leaving 66.7 YES. You receive 83.3 shares at an average of 60¢, and the price is now about 69%. One trade moved it 19 points.

Now make the pool 10 times deeper: 1,000 YES and 1,000 NO. The same 50 buys 97.6 shares at an average of about 51.2¢, and the price ends near 52.4%. A 2.4 point move instead of 19. Manifold, a play-money market that describes liquidity as a super weird and hard-to-grok concept in a discussion of creation costs, says market creators choose a subsidy at creation time "to inject liquidity into the system." Whoever funds the pool decides how deep the market is.

What is slippage, and how do you measure it?

Slippage is the gap between the price you saw and the average price you got. In the constant-product example above, the quoted price was 50¢ and the average fill was 60¢, so slippage was 10¢ a share, or 20 percent of the quoted price. In the deep pool it was about 1.2¢.

Why do volume figures mislead?

Volume can also be manufactured. A paper posted to SSRN in November 2025 and reported by Fortune estimated that "artificial trading" accounted for an average of 25% of all buying and selling on Polymarket over the previous three years, and 45% of sports-related trading. The authors called their identification not definitive, and the paper had not been peer reviewed. Treat it as an estimate, but a large volume figure is a claim about activity, not a promise that counterparties are standing by.

How do you read a market with three traders?

A market with three traders is not worthless. It is a small sample, and you should say so.

First, look at who moved the price and by how much. If one trade took it from 50¢ to 70¢ and nothing happened afterward, the 70¢ reflects one person's opinion plus the depth of the pool, not a crowd.

Second, check the cost of a small trade in both directions. If a 10-unit buy and a 10-unit sell land far apart, the price is a range, not a number.

Third, ask whether anyone had a reason to be right. Three traders who each looked up the same news item are not three independent estimates. Compare against an outside reference such as a poll.

None of this means the price is wrong. It means the uncertainty about the price is wider than the single number suggests. For the broader accuracy evidence, which is mostly about liquid markets, see Are prediction markets accurate?, and for how to interpret a number, how market odds and resolution work.

What can liquidity not fix?

Depth makes a price harder to push around. It does not make it correct. A deep market can still be wrong if the people trading in it share a bias, and a shallow one can be right if the one person trading knew something. Liquidity is about the cost of expressing a view, not about the quality of the view.

For a free-to-play market the practical points are the same, with a gentler cost. Your Sense is play credit, so slippage costs you ranking and prize chances rather than money, but a trade that moves a price 15 points on its own is still not telling anyone much. Size your trades to the pool, and read thin prices as drafts.

Ravioli is a free-to-play prediction market. You trade with Sense, which is free, and the prizes are real.

Create a free account and browse the board.