Liquidity Rewards: Polymarket Offers $150,000 a Day; Kalshi Paid $2.95M in 30 Days

Both large prediction markets pay traders to keep their order books full, but they pay in different shapes. Vultax read every reward listed by each venue's public API late on 13 September 2026. Polymarket posted liquidity rewards on 17,866 markets worth $149,859 a day, paid daily to makers whose resting orders sit near the midpoint. Kalshi attaches a reward pool to one market for a set period; programs that ended in the 30 days to 12 September and were marked paid out carried $2,952,042, about $98,000 a day.

Neither venue puts the money where trading is heaviest. On Polymarket, finance and weather markets took 27.3% of daily rewards on 5.9% of the day's volume, while esports took 0.6% of rewards on 5.5% of volume. On Kalshi the median pool was $20, and the largest series in the period was a 15-minute crypto race. Rewards sit on markets that trade little on their own, and the typical reward is small and shared among everyone quoting.

When listed Kalshi rewards ended, quoted spreads widened in 23 of the 36 markets with enough quotes to compare, by a median 2.0 cents; only 36 of 445 sampled markets were quoted often enough, and the comparison holds under four of five coverage rules.

Polymarket: Vultax read the order book's current-rewards endpoint at 23:57 UTC on 13 September 2026. It lists each rewarded market's daily rate, maximum qualifying spread and minimum order size. The markets were matched to every open event on Polymarket's Gamma API, read between 00:01 and 00:03 UTC on 14 September, for tags, 24-hour volume and quoted spread. Categories are Vultax's grouping of Polymarket's tags; a market with several tags takes the first of esports, crypto, sports, politics, geopolitics, economy, finance, tech, weather and science, and up-or-down crypto markets are counted on their own.

Kalshi: Vultax read the incentive-programs endpoint at 23:57 UTC on 13 September: 207,407 programs, the earliest beginning on 18 September 2025, each with its market, dates, pool and whether it had been paid out. Kalshi reports pools in centi-cents; the figures here are in dollars. Categories come from Kalshi's own series list. Monthly and 30-day totals count programs by the date they ended and include only those marked paid out.

Polymarket's rewards reach about one open market in ten: 17,860 of 185,959 markets open for trading carried one. Most are small. The median rewarded market offered $3 a day in total, 79% offered $5 or less, and 882 offered $50 or more. The largest on the day were a market on whether the Clarity Act becomes law in 2026, at $2,502 a day, the two leading September Fed outcomes at about $1,000 each, and two markets on Sweden's next prime minister, for Ulf Kristersson and Magdalena Andersson, at $1,000 each.

Politics and sports took half of the money, well below their three-quarters of the day's trading. Finance and weather took more than a quarter on less than 6% of volume, about $14 to $22 of rewards for every $1,000 traded, against $2 to $3 in politics, sports and crypto and 42 cents in esports.

Kalshi's programs have grown quickly. Pools paid out on programs that ended in October 2025 came to $198,545; in August 2026 they came to $2,473,663. Most programs are short and small. Over the 30 days to 12 September, 53,781 programs on 1,058 series paid out, half of them with a pool of $20 or less, and the four Sundays averaged $263,740 against $72,965 on other days. Since the earliest program in the data, 187,690 programs marked paid out have carried $12,065,409.

The mix moves from month to month. Economics led the 30 days with 26.8%, 60% of it on gas and diesel price markets, and the largest single series were a 15-minute crypto race at $180,340 and table-tennis matches at $152,100. Sports took 8.4%, 62% of it those table-tennis programs, although sports held 43.6% of Kalshi's open contracts in Vultax's open-interest reading. In August as a calendar month, weather led instead, at 19.2%.

The two venues publish different things, so their totals are not added together here. Polymarket posts a daily amount per market and shares it among qualifying makers by a published score. Kalshi posts a pool for each program and reports whether it has been paid out.

Four points follow from the listings themselves.

• A reward is shared, not paid per order - Polymarket splits each market's daily amount by each maker's share of the score, so a $3-a-day market pays $3 in total to everyone quoting it.
• Rewarded markets were not the tight ones - among sports markets with under $1,000 of 24-hour volume, rewarded markets had a median quoted spread of 28.4 cents against 2.0 cents for the rest; in politics, 3.0 cents against 1.0. Read it as where rewards sit, not as their effect.
• Rewards turn over daily - 59% of Polymarket's reward settings, 10,509 of 17,866, had started on 13 September, the day of the reading.
• Kalshi's money arrives in many small pieces - half of the 30 days' paid-out programs carried $20 or less, and the three largest series were a 15-minute crypto race, table-tennis matches and hourly Miami temperature.

A reward listing shows where a venue is paying for liquidity, which is often where a book is thin. The depth study shows what thin means in dollars for esports and NFL books, and the arbitrage study shows how fees decide whether a quoted gap is worth taking. Vultax's prediction-market desk follows books, spreads and depth on both venues.

What happens to a book when the reward stops? Vultax took 445 Kalshi markets whose listed liquidity rewards ended between 1 and 16 September 2026 and compared each market's quoted spread in the 24 hours before the ending with the 24 hours after. In the 36 markets quoted often enough to compare, spreads widened in 23, tightened in five and did not move in eight. The median market went from a 1-cent spread to 4 cents, and the median change within a market was +2.0 cents.

A bootstrap over market series puts the median change between 0.0 and 3.5 cents. The comparison has no matched control group: six markets with continuing rewards were candidates and none passed the quote-coverage rule, so this measures what happened around reward endings, not what the endings caused.

The 18-hour rule used above is one of several reasonable choices. Holding the 445 markets, the reward-ending times and the definition of a valid quote fixed, we re-ran the comparison requiring 1, 6, 12, 18 and 24 valid hourly closes in each 24-hour window. Four of the five rules give a median spread change of +2 to +3 cents, with markets that widened outnumbering those that tightened by more than four to one.

The strictest rule, a valid quote in every hour on both sides of the ending, keeps 11 markets. Their median change is zero: four widened, one tightened and six did not move. These are the markets that were still quoted around the clock after rewards ended.

The rule decides which markets are in the result as much as what the result is. One valid hour keeps 394 of the 445 markets; all 24 keeps 11. The 18-hour row reproduces the comparison above exactly: 36 markets and +2.00 cents.

Across all 445 markets, a valid two-sided quote appeared in 8,076 of 10,680 expected hourly closes before the reward endings and in 2,790 after. Kalshi's public API returned no candle at all for 7,495 of the later hours.

A missing candle is a missing observation. It can mean nobody updated a quote or that the venue did not report the hour; it is not a measured empty book, and no missing hour was filled in or counted as a zero spread. One sparse market fetched again through the single-market endpoint returned the same 27 observations.

Keeping only markets priced between 10 and 90 cents leaves 32 of the 36 and the same +2.0 cents. One market per event family leaves all 36. Stretching the after-window to 48 hours keeps 26 markets at +1.0 cent; that is a different set of markets, so it does not show spreads recovering on the second day.

Vultax read Kalshi's full liquidity-programme listing on 19 September 2026 and kept the first ending per market between 1 and 16 September that followed at least 24 hours of rewards and had no new programme for 48 hours. Markets had to be open through the day before and the two days after, which separates a reward ending from a contract expiring.

One election series supplied 7,832 of the 12,609 eligible contracts, so the sample takes at most two event families per series, fixed by a hash before any price was fetched: 445 markets across 408 series.

One reading of each venue's reward data on 13–14 September 2026. Polymarket's daily amounts are offers: what makers receive depends on who else quotes and on the $1 minimum. Kalshi's pools are counted when a program is marked paid out; 3,480 programs that ended in August, with $184,815 in pools, were not marked paid out at the reading. Quoted spreads come from Polymarket's Gamma API and can lag the order book. Polymarket figures cover its international exchange. Nothing here is a recommendation to trade. Challenge a figure at [email protected]; material changes are logged in the changelog.

The reward-ending comparison reads Kalshi's hourly closing bid and ask, keeps a quote when both are inside $0 to $1 and the bid does not exceed the ask, and gives each market one observation: its median spread after the ending minus its median before. A market enters with at least 18 valid hours on each side. Closing quotes are not executable depth, and the schedule is a 19 September snapshot the venue can revise.

The downloadable Kalshi-related input and derived data files are temporarily unavailable while source redistribution permissions are reviewed. Methods and limitations remain described here; contact [email protected] for questions. This is not a claim that Vultax owns the underlying venue data.