# How Market Makers Set Prediction Market Prices and Spreads

> On order-book prediction markets the displayed price is the midpoint of quotes that venues pay makers to post. On November 11, 2026, Polymarket's rewarded markets had a 3.8-cent median spread against 48 cents for a sample of the rest, and depth fell 64% at the reward band's edge. Our view is that rewards buy tight prices without buying better information.

Published 2026-11-11 by PM Crypto Notes. Canonical: https://pmcrypto.blog/how-market-makers-set-prices/

## Key points

- Polymarket listed 11,093 markets in its liquidity rewards program on November 11, 2026, paying $115,218 a day; the top 100 markets took 28.6% of it.
- Kalshi had 6,561 live liquidity incentive programs worth about $140,600 a day, every one with a 1,000-contract target size and a 0.5 discount factor.
- Rewarded Polymarket markets had a median spread of 3.8 cents and 98.7% had two-sided books; in a random sample of 3,860 other markets the figures were 48 cents and 49.6%.
- Resting depth fell 63.7% from the last cent inside the reward band to the first cent outside it, against 35.7% at the same distance in unrewarded markets.
- In one $4,000-a-day market, a 1,000-share quote at the touch would earn an estimated $258 a day in rewards against $10 of spread per round trip.


Our thesis is that on order-book prediction markets the displayed price is mostly a market maker's quote, that the venues' reward formulas now decide where those quotes sit, and that paid quoting makes prices tight and continuous without making them better informed.

## Who sets the price on an order-book prediction market?

Polymarket's [documentation](https://docs.polymarket.com/concepts/prices-orderbook) states the rule plainly: "The displayed price is the midpoint of the bid-ask spread," and "If the spread is wider than $0.10, the last traded price is shown instead." Those best bids and asks come mostly from market makers, firms and bots that post on both sides and earn the spread when takers cross it. A taker moves the displayed price only by trading through resting levels.

Kalshi's [market maker program](https://help.kalshi.com/en/articles/13823819-market-maker-program), described on a help page dated April 28, 2026, asks designated firms for "consistent, two-sided liquidity" in exchange for reduced fees and adjusted position limits. Jump Trading, already a Kalshi market maker, was reported in February 2026 to be poised to take small equity stakes in both venues tied to its market making, [DeFi Rate reported](https://defirate.com/news/equity-for-liquidity-jump-trading-set-to-take-stakes-in-kalshi-and-polymarket/), citing Bloomberg.

## What do the venues pay makers to quote?

**Polymarket liquidity rewards.** Polymarket [samples the book](https://docs.polymarket.com/market-makers/liquidity-rewards) once a minute at a random moment. Each resting order within a market's maximum spread `v` of the midpoint scores its size times ((v − s) / v)², where `s` is its distance from the mid in cents. A maker's score is roughly the smaller of its two sides, with one-sided quoting allowed at a third of the credit when the midpoint is between 10 and 90 cents and disallowed outside that range.

At 22:42 UTC on November 11, 2026, Polymarket's public rewards endpoint listed 11,093 markets paying a combined $115,218 a day. The median market paid $3; 77% paid $5 or less; the top 100 markets took 28.6% of the pool. Most used a 4.5-cent (6,474) or 6.5-cent (3,342) maximum spread. By category, politics took $39,207 a day (34.0%), tech $18,536 and sports $14,318. On top of this, Polymarket's [maker rebate program](https://docs.polymarket.com/programs/maker-rebates) returns 15% to 25% of taker fees to makers, by category.

**Kalshi's Liquidity Incentive Program.** Kalshi [snapshots the book](https://help.kalshi.com/en/articles/13823851-liquidity-incentive-program) every second. A snapshot counts only if both the yes and no sides reach a target size. Orders at or better than a reference price, the first level where cumulative size reaches a fifth of the target, get full credit; worse orders get "the Discount Factor raised to the number of ticks away." The help page puts rewards at $1 to $1,000 a day per market and says the program ends January 1, 2027. An [amendment filed on October 14, 2026](https://www.cftc.gov/filings/orgrules/rules10142633813.pdf) caps rewards at $10,000 per business day for the most volatile markets and states the aim is to "enhance pricing efficiency."

Kalshi's public API listed 6,561 live programs at 22:54 UTC. Using the API's centi-cent [reward unit](https://docs.kalshi.com/api-reference/incentive-programs/get-incentives), they were worth about $140,600 a day, with a median of $100 per period. Every one used a 1,000-contract target and a 0.5 discount factor. Elections had 2,084 programs worth about $45,200 a day and entertainment 1,676 worth about $37,900; sports had 402.

**Volume-share rebates.** Polymarket US pays its makers by share of maker volume in each series, with a 5% minimum share, under a program [filed on September 30, 2026](https://www.cftc.gov/filings/orgrules/rules09302630733.pdf); our essay on [volume as a headline number](/volume-is-the-wrong-headline/) covers what volume-based pay does to reported volume. CFTC staff warned in August, [per Willkie Farr & Gallagher](https://www.willkie.com/publications/2026/08/cftc-advisory-on-incentive-programs-for-prediction-markets), that market maker programs which "guarantee net profits or cover participant losses through stipends and rebates may incentivize artificial strategies."

## How does a reward formula shape a single quote?

Take the largest reward in our pull: $4,000 a day on Polymarket's "Jensen Huang announced resignation or removal by December 31, 2026?" contract, with a 3.5-cent maximum spread and a 200-share minimum. At 22:43 UTC the book showed a best bid of 65 cents for 11,429 shares and a best ask of 66 cents for 54 shares, a midpoint of 65.5 cents.

Suppose a maker joins the touch with 1,000 shares a side, at 65 and 66 cents, half a cent from the mid. Each side scores 1,000 × (3.0 / 3.5)², about 735. The rest of the visible book, counting levels of 200 shares or more, scores about 12,240 on the bid side and 10,660 on the ask side. If that book held all day, our hypothetical maker's share would be about 735 / (735 + 10,660), or 6.5%, worth about $258 a day on roughly $990 of capital. Buying at 65 and selling at 66 earns $10 per round trip.

Moving the same quotes to 64 and 67 cents cuts the estimate to about $119 a day; 63 and 68 cents, about $30; at 3.5 cents or more, nothing. So the reward pays a maker roughly 26 times the spread to stand at the touch, and about an eighth of that two ticks back. Kalshi's discount factor of 0.5 is gentler: an order one tick behind the reference price earns half credit, two ticks a quarter, and it never falls to zero.

The risk sits on the other side. Polymarket's API reported a one-day price change of 54 cents on this contract, and its best quotes moved from 69 and 70 cents in our 22:39 gamma read to 65 and 66 cents about four minutes later. A 1,000-share quote run over by a 54-cent move loses up to $540, about two days of reward. Two sister contracts, paying $2,000 and $1,000 a day for the November 13 and November 30 deadlines, had spreads of 20 and 45 cents with nothing resting inside their reward bands, so makers had left those rewards unclaimed at the time of our pull.

## What does the order book data show?

At 22:42 UTC we fetched the order book for all 11,088 open markets in Polymarket's program that accepted orders, and for 3,860 markets drawn at random from the 198,269 open markets outside it.

| Polymarket, November 11, 2026 | In rewards program | Not in program (sample) |
|---|---|---|
| Markets | 11,088 | 3,860 |
| Two-sided book | 98.7% | 49.6% |
| Median spread, two-sided books | 3.8¢ | 48.0¢ |
| Spread of 2¢ or less, share of all markets | 34.6% | 7.9% |
| Median spread, midpoint 20¢ to 80¢ | 8.0¢ | 90.0¢ |
| Depth drop, last cent inside band to first cent outside | 63.7% | 35.7% |

The gap is mostly sports. Rewarded sports markets had a median spread of 3.2 cents against 84 cents for unrewarded ones. In politics (3.0 against 2.3 cents) and crypto (3.0 against 2.0), the unrewarded markets in our sample were as tight, though the samples were small (158 and 77). Spreads also narrowed with the size of the reward: a median of 4.0 cents in markets paying under $2 a day, 3.0 cents at $10 to $50, and 1.0 cent in the 64 markets paying $200 or more.

The depth profile is the clearest sign of the formula at work. For the unrewarded control we used a nominal 4.5-cent band. Across rewarded markets, resting dollars fell 63.7% between the last cent inside each market's band and the first cent outside it. In unrewarded markets, depth fell 35.7% across the same distance. Depth thins with distance everywhere; in rewarded markets it falls off at the line where payment stops.

Kalshi shows the same direction. On open contracts priced between 10 and 90 cents, markets with a live incentive program had a median spread of 5 cents, against 9 cents for the rest. The gap was 1 against 7 cents in sports, 4 against 10 in elections and 1 against 10 in economics. Entertainment ran the other way: 18 cents with an incentive, 14 without.

## What does paid quoting do to price accuracy?

Our read is that rewards buy three things and leave one out.

They buy tightness and continuity. In Polymarket sports, the reward line separates a quoted market from a placeholder; our [liquidity essay](/liquidity-is-the-product/) found thousands of unrewarded props at 3 and 97 cents. A tight market lowers the cost of trading on information, which should help accuracy at the margin.

They buy a particular shape. Depth bunches inside the band and thins past it, so a book resists a small order and gives way to a medium one. In 89.1% of rewarded Polymarket markets priced between 10 and 90 cents, less than $1,000 of offers sat between the midpoint and 5 cents above it; the median was $40.

They buy coverage where the venue chooses to pay: a third of Polymarket's pool went to politics, and one contract drew 3.5% of it.

What they leave out is information. Both formulas score distance from the current midpoint, wherever it is. A maker is paid the same whether the mid is right or wrong, so paid makers follow the price while takers move it. When the news risk is large, as on the Huang contracts, makers step away from the reward, and the price is left to whoever trades.

## What is the strongest case against our view?

The strongest objection is that paid makers are the marginal traders the theory relies on (our [wallet-concentration essay](/wallet-concentration/) covers that evidence), and that our comparison is biased. Venues reward markets they expect to trade, so rewarded markets would be tighter anyway, and depth always thins with distance from the mid.

We accept that our data cannot separate cause from selection across markets. We doubt selection alone explains an 81-cent gap in sports, but the politics and crypto comparisons show it matters. The band-edge test is harder to explain away: at the same distance from the mid, depth in rewarded markets fell nearly twice as fast as in unrewarded ones, at the point where the formula stops paying. On accuracy, we agree tight spreads help informed takers. Our concern is narrower: rewards make a price look settled to the fraction of a cent, and readers cannot see whether that comes from information or from a $4,000 daily subsidy.

## What would change our mind

- Calibration data showing rewarded markets resolve closer to their prices than unrewarded markets with similar volume and spread.
- A market whose maximum spread changes, with its depth profile staying put, which would mean the band edge is coincidence.
- Makers holding quotes through large news moves in high-reward markets.
- Venues publishing per-market reward spend beside maker profit and loss, so readers can see what each tight spread costs.


## FAQ

### How are prediction market prices set?

On order-book venues such as Polymarket and Kalshi, the displayed price is usually the midpoint between the best resting bid and ask. Polymarket shows the last traded price instead when the spread is wider than 10 cents. The resting orders mostly come from market makers.

### How do Polymarket liquidity rewards work?

Polymarket samples the order book once a minute and scores each qualifying resting order by size and by the square of its distance inside a maximum spread from the midpoint. Each maker's share of the day's scores sets its share of that market's daily reward, paid at midnight UTC.

### What is Kalshi's Liquidity Incentive Program?

It pays for resting orders that help both sides of a market reach a target size, scored by size and discounted by a factor for each tick below a reference price. Kalshi snapshots the book every second, pays $1 to $1,000 a day per market, and says the program ends January 1, 2027.

### Do market makers make prediction markets more accurate?

They make prices tighter and more continuous, which lowers costs for informed traders. In our November 11, 2026 data, rewards also bunched depth just inside each reward band, so a displayed price can look precise without much information behind it.

### How much does Polymarket pay market makers?

Polymarket's public rewards endpoint listed $115,218 a day across 11,093 markets at 22:42 UTC on November 11, 2026. The median market paid $3 a day; one market paid $4,000. Makers also receive 15% to 25% of taker fees as rebates, depending on category.

## Sources

- [Liquidity Rewards, accessed November 11, 2026](https://docs.polymarket.com/market-makers/liquidity-rewards), Polymarket Documentation, 2026-11-11
- [Maker Rebates Program, accessed November 11, 2026](https://docs.polymarket.com/programs/maker-rebates), Polymarket Documentation, 2026-11-11
- [Prices & Orderbook, accessed November 11, 2026](https://docs.polymarket.com/concepts/prices-orderbook), Polymarket Documentation, 2026-11-11
- [CLOB rewards endpoint: current reward markets (pulled November 11, 2026, 22:42 UTC)](https://clob.polymarket.com/rewards/markets/current), Polymarket, 2026-11-11
- [CLOB order books for 14,948 open markets (pulled November 11, 2026, 22:42 UTC)](https://clob.polymarket.com/books), Polymarket, 2026-11-11
- [Gamma API: all open events (keyset, pulled November 11, 2026, 22:39 UTC)](https://gamma-api.polymarket.com/events/keyset?active=true&closed=false&limit=100), Polymarket, 2026-11-11
- [Liquidity Incentive Program](https://help.kalshi.com/en/articles/13823851-liquidity-incentive-program), Kalshi Help Center, 2026-11-11
- [KalshiEX LLC: Amendment to August 2025 Liquidity Incentive Program](https://www.cftc.gov/filings/orgrules/rules10142633813.pdf), CFTC, 2026-10-14
- [Trade API: incentive programs, status active (pulled November 11, 2026, 22:54 UTC)](https://api.elections.kalshi.com/trade-api/v2/incentive_programs?status=active&limit=1000), Kalshi, 2026-11-11
- [Get Incentives (API reference), accessed November 11, 2026](https://docs.kalshi.com/api-reference/incentive-programs/get-incentives), Kalshi Documentation, 2026-11-11
- [Trade API: open events with nested markets (pulled November 11, 2026, 22:13 UTC)](https://api.elections.kalshi.com/trade-api/v2/events?status=open&with_nested_markets=true&limit=200), Kalshi, 2026-11-11
- [Our Market Maker Program is now Live on Kalshi](https://help.kalshi.com/en/articles/13823819-market-maker-program), Kalshi Help Center, 2026-04-28
- [QCX LLC (Polymarket US): 2026 Amended Series-Based Market Maker Program, Appendix A](https://www.cftc.gov/filings/orgrules/rules09302630733.pdf), CFTC, 2026-09-30
- [Equity for Liquidity: Jump Trading Set to Take Stakes in Kalshi and Polymarket](https://defirate.com/news/equity-for-liquidity-jump-trading-set-to-take-stakes-in-kalshi-and-polymarket/), DeFi Rate, 2026-02-10
- [CFTC Advisory on Incentive Programs for Prediction Markets](https://www.willkie.com/publications/2026/08/cftc-advisory-on-incentive-programs-for-prediction-markets), Willkie Farr & Gallagher, 2026-08-24