Polymarket Raised Its Fees — How Much More Does a Trade Cost?
We compare what one trade cost on Polymarket before and after its fee change, for the same kind of games, and also check whether the gap between buy and sell prices changed.
- Result status
- Preliminary results
- Last updated
- Start date
- Duration
- Day 3 of 14
- Research team
- Mysports.AI Research Team
- Plan registered
Polymarket Fees: Key Findings
From MLB games in summer 2026: about 1.2¢ per share on average across 120 games before the change, and about 1.7¢ across 167 games after; most of the increase is the fee itself. Each share is worth at most $1. This is a direct before-and-after comparison on Polymarket itself.
- Most of the extra cost is the fee itself: fee = rate × price × (1 − price), and the rate went from 0.03 to 0.05. For one share at $0.50 (odds 2.0), the fee went from about 0.75¢ to about 1.25¢.
- The gap between buy and sell prices was almost always at the minimum of 1¢; different methods give different answers, so we cannot tell whether it widened.
- The planned comparisons with another prediction platform (Kalshi) — the difference in cost increases, odds accuracy and trading volume — are not published until we get written permission, because its data terms restrict public use.
In plain words
Why study this?
People who accept an existing price usually pay a fee. How trading costs, spreads and the accuracy of the odds change around a fee change affects whether trading on Polymarket is still worth it. For the model, before using a prediction platform’s odds as a reference, we need to know whether rule changes affect the quality of those odds.
What do we want to know?
On 2026/7/10, Polymarket (a prediction platform where people buy and sell shares in game results) raised its fees for sports games. We mainly want to know: does one trade now cost more? Has the accuracy of the odds at predicting game results changed? For trading volume, we only have a reference analysis for the NFL, so we can’t yet say whether trading dropped overall.
How do we study it?
We compare buy prices, sell prices and the fees actually collected on the blockchain for MLB games on Polymarket, before and after the fee change, and work out how much more one share cost to trade on average. The plan also compared with another prediction platform, but that platform’s data terms restrict public use, so for now the page only shows Polymarket’s own before-and-after comparison.
What have we found?
After the fee change, the cost of trading one share on Polymarket rose from about 1.2¢ to about 1.7¢ on average, and most of the increase is the fee itself. The gap between buy and sell prices stayed at the minimum of 1¢ almost the whole time; different methods disagree, so we cannot say whether it widened. Whether the odds became more or less accurate was meant to be checked against another platform, and that part is not published for now.
How much did fees go up?
Polymarket’s fee is worked out as “rate × price × (1 − price)” (the price is between $0 and $1), and the rate went from 0.03 to 0.05. For a share priced at $0.50 (an implied chance of about 50% = odds of about 2.0), the fee per share went from about $0.0075 to about $0.0125 — roughly 1.5% to 2.5% of the price. The closer the price is to $0 or $1, the smaller the fee.
What can’t we say yet?
On July 10 the fee went up and the reward for people posting orders went down, so we see the combined effect. The comparison window is short because the free archive stopped (15 days before, 26 after), the World Cup was on at the same time, and the page has no comparison group, so market-wide changes are not taken out. The spread was almost always at its minimum, so changes are hard to see; not seeing it widen does not mean there was no effect. The NFL data are incomplete, so treat them as reference only.
Terms used on this page
- Polymarket
- A prediction platform where people buy and sell shares in game results. Prices run from $0 to $1 and roughly show the chance the traders think an outcome has — not necessarily the true chance. A price converts directly to odds: odds = 1 ÷ price.
- Order book
- The list of buy and sell orders on a platform that are still waiting to be filled.
- Spread
- The gap between the highest buy price and the lowest sell price. For example, if someone will buy at $0.48 and someone will sell at $0.52, the spread is $0.04; to buy right away, you have to pay $0.52. The bigger the gap, the more it costs to trade right away.
- Taker / maker
- A taker accepts someone else’s existing price and trades right away; a maker posts their own price and waits for someone to trade with it.
- Maker reward
- The platform gives part of the fees back to makers, to encourage people to post prices.
- Midpoint
- The value halfway between the highest buy price and the lowest sell price, often used as the current reference price.
Detailed results
After the fee change, trading on Polymarket cost more
Scaling the average cost to 100 shares makes the amounts easier to read. All amounts below are US dollars.
Average cost per 100 shares
1.22 USD
Average cost per 100 shares
1.74 USD
An increase of
1.74 − 1.22 ＝ 0.52 USD / 100 shares
Most of the increase is the fee itself: the gap between buy and sell prices stayed at the minimum of 1¢ almost the whole time, before and after.
Cost means the extra paid relative to the reference price, plus fees; it excludes the share price itself. Scaling to 100 shares is a unit conversion, not a simulation of buying 100 shares.
Original units and calculation details
Original units are cents per share. Since $1 = 100 cents, 1.22 cents per share scales to $1.22 per 100 shares.
Total cost = half the spread + fee. This is a direct before-and-after comparison on Polymarket itself, with no comparison group, so market-wide changes over the same period are not taken out.
The same 287 MLB games on Polymarket (120 before the change, 167 after, all with fees confirmed on the blockchain); each share is worth at most $1.
Main analysis (v1.1 plan, corrected after external code review) | fees determined only from on-chain charges before each quote time | the switch actually happened around 03:00 UTC on July 10.
HypothesesFor deeper reading
- H1After the fee change, the gap between Polymarket’s buy and sell prices gets bigger → not seen with the method in our plan; a recalculation done after seeing the results shows a small increase; the two disagree, so we can’t tellInsufficient evidence
- H2(Reference only) After the fee change, the share of trades made by accepting an existing price falls → it rose slightly instead; this share is of limited use as a referenceExploratory
- H3After the fee change, the cost of buying or selling one share rises more than on another platform → analysed; this uses the other platform’s data, whose terms restrict publication, so the result stays unpublished until we get permissionNot published yet
- H4After the fee change, how accurately the odds predict game results (compared with another platform) changes → same as above; unpublished until we get permissionNot published yet
- H5(Reference only) During the NFL’s no-fee period, the gap between buy and sell prices is smaller and there is more trading → the gap was always the minimum of 1¢; the volume comparison uses another platform’s data and is not published for now; the data are incomplete, so this is reference onlyExploratory
Further analysis
Further analysis done after the main results. Items marked “post-hoc” were added after seeing the data and are for reference only; items marked “pre-registered” or “Extension 1” had their rules written and timestamped before anything was calculated.
- Underdog bettors pay the most: the fee per share is rate × price × (1 − price). Per $100 staked, buying an underdog at decimal odds of about 5.0 (price $0.20) cost $2.40 in fees before and $4.00 after; buying a favourite at odds of about 1.25 (price $0.80) went only from $0.60 to $1.00. Conversion: a price of $0.60 ≈ a 60% chance ≈ odds of about 1.67.
- Next (pre-registered and timestamped): Polymarket has waived NBA game fees since Oct 20, expected to end Nov 1, while NHL games kept their fee. We are recording both leagues' order books 24, 6 and 1 hours before each game and will estimate in mid-November how the cost of an NBA bet changed once fees returned.
Limits of this analysis
- Fees are list prices; Polymarket refunds part of the fee to high-volume traders, so some pay less. The per-$100 cost uses the best quotes and ignores slippage on large bets.
Polymarket Trading Costs: Research Methods
Research questions
- After the 2026/7/10 fee change, did the gap between the highest buy price and the lowest sell price (the spread) get bigger?
- Did trading volume and the number of posted orders drop? How much did the real cost of one trade change for traders?
- Do the odds still predict game results as accurately? In the NFL, do the effects of “fees → no fees → fees again” mirror each other?
Method summary
- The fee change: from 2026/7/10, the fee rate for sports games went from 0.03 to 0.05, and the reward for people who post orders dropped from 25% to 15%. Polymarket said it took effect at midnight, but the actual charges on the blockchain switched about 3 hours later, which doesn’t change which games fall before or after. Fee = rate × price × (1 − price), worked out per share.
- Planned comparison group: buy and sell prices for the same games on another prediction platform (Kalshi), to take out changes that hit both together. Its data terms restrict storing and publicly using the data, so these results are not published for now and will be added once we have written permission.
- Days compared: only days where the free archive has Polymarket data (15 days before the fee change, 26 days after).
- What we measure: the gap between buy and sell prices, the cost of one trade (estimated from the buy and sell prices; what people actually pay depends on the order), and how accurate the odds are 1 hour before the game. We also check that the prices are recent enough.
- NFL: games are split into three periods — fees charged, no fees, and reduced fees — based on each game’s own fee setting. The data are not complete enough, so this is reference only.
Data and limitations
Data we use (all public and free)
- Polymarket hourly order-book archive (PMXT; MLB) and Pendulum Flow archive (NFL)
- Polymarket’s official fee explanation, market data, and the actual fees charged on the blockchain
Limitations
- The free archive stopped partway (Polymarket after 8/10). After switching to other sources, fewer days can be compared.
- On 7/10, Polymarket both raised the fee and cut the reward for posting orders, so we can only see the combined effect.
- The fee listed on the website is not always what is actually charged, so we checked against the fees actually collected on the blockchain. The page only shows Polymarket’s own before-and-after comparison with no comparison group, so market-wide changes over the same period, such as the World Cup, are not taken out.
Latest progress
Latest：Removed numbers compared with another platform (data terms restrict it), keeping only Polymarket’s own before-and-after comparison; a forward comparison around the NBA fee return is collecting data
Next step：Write up the results and confirm data licensing, then publish
Study timeline
- Research direction set
- Writing the plan and timestamping it
- First data checkpoint failed; plan revised to v1.1 and re-timestamped
- Collecting data and analyzing
- Publishing results and tracking
- External code review and publicationIn progress
Update log
Version history
- v1.2-fix
The planned comparison platform (Kalshi) restricts storing and publicly using its data, so we removed every number and conclusion that compares with it until we get written permission. The comparison platform’s cost figures shown earlier were also wrong because a fee multiplier was miscalculated, and have been removed too. Polymarket’s own before-and-after numbers are not affected
- v1.1-fix
Fixed after a second AI tool checked the code: corrected when the fee change took effect, stopped using later data to work out the fee, and used the same games for NFL volume
- v1.1-results
Main analysis finished (after seeing the results, we changed to using only regular trades to work out the fee; both methods are kept)
- v1.1
The free archive stopped, so we switched to other data sources and only use days with Polymarket data; guess 2 was changed to a reference-only result. Re-timestamped
- v1.0
Study plan finalized and timestamped; data collection started
Study record
459f124a326bf093e16f6f332f70db9c765e8d4c49aa005a5033711541f55049effb078814f65d16c6b15949a9cf45edb93dddf4a8f203d76b76e5092b59a33e91cc6a8cdf3157be482fd2ca5b037a6b529f6fd8336a6c3262d0a83d109da7495360c662804dd03c3f81c35c0da25248e16de053e8b330ae8ce500be04c9571eThis page is a research log for reference only and is not betting advice or a guarantee of profit.