Do liquidity rewards and sponsorships actually tighten spreads and deepen books on prediction markets?
TLDR:
Rewards and sponsorships can move top-of-book liquidity, but only when daily spend is north of 1% of the existing book. Below that, the median program does basically nothing. And even above it, size barely predicts which markets actually respond.
Methodology and Background
Rewards are liquidity incentives awarded by Polymarket for placing passive orders within a certain price range. Sponsorships (which launched Feb 17th) is where any user can be the one who gives out rewards to makers.
We indexed all instances of liquidity sponsoring (including tests from Feb 10th to March 18th, 2026) as well as all rewards from 2026-01-06 to 2026-05-13 (non-exhaustive coverage).
For rewards, we indexed daily reward rates and start dates (rewards do not have granular minute-based start dates), while for sponsorships, we indexed total amounts and the start minute.
Then we measured the effect of different sizes of sponsorships and rewards by looking at the relationship between weighted tob uplift pct and daily rewards or sponsorships in dollar value:
We calculate weighted tob uplift pct as
weighted_top_of_book_uplift_pct = SUM(pre_total_top_of_book * top_of_book_uplift_pct) / SUM(pre_total_top_of_book)
As an example, say there are two markets. Market A has $10 usd of tob liquidity, and post-rewards it moves to $30 (+200%). Market B has $10,000 usd of tob liquidity and post rewards moves to $10,500 (+5%). If we were to JUST measure the average % increase, we would massively overweight the $20 usd increase and underweight the $500usd increase (the value will be +102.5%). The above weighted formula will calculate the aggregate uplift based on pre-dollars to post-dollars ($20usd uplift on A + $500usd uplift on B / $10 + $10,000 = +5.2%).
This is the percent change you’d get if you merged all events in the bucket into one giant combined book.
We have also removed the effect of resolved or near resolved markets by removing data points for markets that reached more than 0.98 and less than 0.02 and stayed there.
Liquidity weighted uplift segmented by incentive intensity.
Note, many of the smaller buckets are pure noise. A 10 usd sponsorship is not going to have any effect when tob liquidity is 100,000usd. We therefore only index buckets where daily rewards exceed 0.01% of pre-event tob liquidity, and give much more credence to 1%+ buckets.
One of the many confounding factors is the fact that sponsorships can occur right before resolution or around a critical piece of information. In this case isolating the increase in depth vs the increase of liquidity due to derisking of key information flow is very hard. We are however, reporting aggregate statistics, but one large thick event can move the aggregates.
The worst confounder is near resolution: when prices approach 0 or 1 and traders are confident in the outcome, makers happily provide liquidity as a bond for early exiters
Many sponsorships can occur within a tight band of each other and each episode contributes an entry to the weighted total. (ie. if 3 sponsorships occur within one minute of each other they are three entries into the weighted total). However, empirically this does not contribute to lowering the observed tob uplift. For example if we group sponsorship events by the hour, the uplift is still very large.
The sponsorship distribution is extremely right-skewed. A minority of episodes have enormous uplift, so if you look at the median it stays very modest while mean, weighted metrics stay very large. This can mis-attribute attention-based liquidity uplift (ie. in the Jesus market) to the effect of liquidity incentives, especially compounded with 1.
For larger markets the spread was already tight and was mechanically limited by tick size. The plateau at 50% shows that perhaps certain markets are ready to get more granular.
The weighted averages show very large top-of-book liquidity gains after rewards and sponsorships, especially in higher-intensity buckets, but the distribution is extremely wide. The median/Interquartile Range (IQR) evidence supports a positive relationship between incentive intensity and liquidity response for sponsorships, and a weaker, less consistent relationship for rewards. Importantly, the lower quartile remains negative even in high-intensity buckets, so incentives do not guarantee a liquidity uplift. If the goal is a meaningfully positive median response, the data suggests low-intensity incentives are often insufficient; sponsorships begin showing double-digit median uplift around the 0.1%–1% of pre-book/day bucket, while rewards only show a clear double-digit median uplift in the 3%+ bucket.
Below we plot tob + spreads for some of the markets that contributed the most to weighted average uplift, and present top 5 and bottom 5 uplifts + spread compression, tob liquidity decrease post sponsorship event filtered only to pre-event 1% of tob events.
Some of the biggest size-weighted top of book increases (Will Jesus Christ return in 2027?) are impossible to plausibly attribute to sponsorship vs. marketing-driven attention. For sports, we often also do see liquidity thickening as we approach game start.
Meanwhile some of worst weighted liquidity decrease, we believe, tend to be where exogenous information flows are drivers of price and liquidity, albeit, the data points are sparse enough, and the liquidity is controlled by such few makers that it is impossible to discern by looking at just the winners or the losers what market features will dictate winners or losers post rewards / sponsorships.
There is also no observed difference in the conclusion when we expand liquidity tracking to everything covered under reward minimal spread definitions (ie. for the Oil market rewards are given to anyone who is within +-4 ticks of spread). In the below chart we track pre +-4 ticks and post and observe that though it changes some of the chart values, it doesn’t change the conclusion.
The natural next question: what actually predicts whether incentives translate into uplift?
There is no observed median uplift for rewards. The forward r^2 (take an univariate linear regression model and a multivariate ridge of just incentive intensity to target which is ln (post 1d avg tob / pre tob), tends to be very low (~0.011 combined in incentive intensity and 0.119 in all features). When we removed a feature and observed r^2 drop, the pre-tob trend, and whether or not pre-tob liquidity was already thin were much larger drivers of variance in uplift. Therefore size of rewards (measured as % of existing liq) don't explain variance in uplift.
There is, however, in the 3%+ bucket clear demonstration of uplift in the median case, as well as the combined case. The low predictability just shows that differences in intensity fail to explain variance and intensity/size is not deterministic in uplift, rather other factors contribute.
Liquidity incentives on Polymarket appear to work, but not in a simple, linear, or universally predictable way. The clearest result is that rewards and sponsorships can materially improve top-of-book liquidity and compress spreads, especially when the incentive is large enough relative to the market’s existing liquidity. In practice, markets receiving daily incentives of roughly 1% or more of pre-event top-of-book liquidity are the ones most likely to see meaningful uplift.
The harder problem is not whether incentives can move liquidity, but how to deploy them efficiently. Much of the observed variance is still unexplained: some markets respond dramatically, others barely move, and many outcomes are confounded by attention, resolution risk, and exogenous information flow. The strongest predictors today are still basic pre-event conditions like spread width and existing liquidity, which suggests that incentive design remains under-optimized.
Incentives buy a better median chance of liquidity uplift, not a guarantee of closing liquidity gaps. Spending a lot can help, but it is not sufficient.
Our thesis is that prediction market liquidity will not be solved by rewards alone but rather fundamental market structure innovations that engender deeper liquidity and smaller spreads.
Reach out
We'd always love to talk to traders, builders, MMs. DMs are open. If you'd like to stay up to date with what we're working on sign up here: https://longshot.xyz/
Thank you to @_helmass, @Mikey0x_, @MovieTimeDev and @defiance_cr for review and feedback on these ideas.
Originally posted on X.