Polymarket review: makers pay zero, takers pay 0.00 to 0.07 by category

Polymarket charges the taker and never the maker. A resting order that somebody else hits costs you nothing, and that is zero rather than a discount. Take liquidity and you pay a per-category coefficient: 0.07 on crypto, 0.05 on sports, economics, culture, weather and other, 0.04 on finance, politics, mentions and tech, and 0.00 on geopolitical. Eleven categories, four distinct rates. The charge is rate x contracts x P x (1 - P), so the same $100 taker fill costs $4.90 at a price of $0.30 in a crypto market and $0.70 at $0.90 in that same market. The part most reviews get wrong: the United States sits on Polymarket's close-only list on the frontend and the API, not the frontend alone.

Status as of 18 August 2026. Fee coefficients read from each market's feeSchedule object on the Gamma API on 4 August 2026, schedule version 2026-08-04.a. Country list read from the geoblock endpoint on 5 August 2026. PredictionEdge currently takes no referral revenue from either platform; whether it ever will is undecided.

The fee table, all eleven categories

Nobody publishes this as a table. The coefficients live in the API, one per market, in a feeSchedule field, which is why third-party fee pages tend to quote one flat number for a schedule that carries four.

The two dollar columns below are what a $100 taker fill costs at two entry prices. They come straight from the formula, because contracts cancel: fee as a share of stake is rate x (1 - P).

Geopolitical is the row worth staring at. It is not a promotion and not a rebate. The coefficient is zero, so a geopolitics market is free on both sides of the book.

The coefficient is a ceiling, not a percentage

Read 0.07 as 7% of your money and you will be wrong almost everywhere. Two different things peak in two different places.

• Fee per contract peaks at a price of $0.50, where P x (1 - P) is largest. On a crypto market that is 0.07 x 0.25 = 1.75 cents per contract, the most a single contract can ever cost you in that category.
• Fee as a share of stake is rate x (1 - P), which rises the cheaper the contract gets. It never reaches the coefficient, but in the longshot corner of the board it gets close: at $0.05 in a crypto market it is 0.07 x 0.95 = 6.65% of stake.

So the expensive place to be a taker is not the middle of the book, it is the cheap end. A $100 fill at $0.05 in crypto costs $6.65; the same $100 at $0.95 costs 35 cents. That is a 19x spread in cost on identical size, in the same market, from price alone. Traders who size into 3-cent longshots because "it's only three cents" are paying the highest fee rate available on the venue.

Two of our own paper trades, same edge, opposite economics

On 28 July 2026 our scanner opened five paper positions on $500 of notional. It is not an arbitrage detector and we should say so before quoting it: intra-market arbitrage is dead here for the reason set out in the next section, and what the scanner actually emits is spread-edge and price-momentum signals. Those are directional bets with no locked outcome. Two of the five were entered on an identical claimed edge of 1.20% and the fee arithmetic sends them in opposite directions.

The arithmetic, in full, because this is the whole review in two lines:

• 478026: 0.07 x (1 - 0.8693) = 0.0091, so the trade needs 0.91% to break even and it was entered on 1.20%. It clears, narrowly.
• 4ad1b0: 0.04 x (1 - 0.2965) = 0.0281, so the fee alone takes 2.81% of stake and the position was entered on a 1.20% edge. It was a guaranteed loss at the moment of entry, before the market moved at all. The lower coefficient did not save it; the low price killed it.

Across all five positions the book was +$4.73 at entry on one leg and -$3.48 on two legs, meaning the cross-venue version was underwater before any price moved. Those are five worked examples, not a rate: n = 5 supports no percentage and we are not going to publish one. The full paper log stands at 10 trades, 6 closed, net -$28.30 on $500 deployed as of 4 August 2026. Our break-even figures were corrected on 4 August 2026 after a cross-agent review found the original table had hardcoded one category rate across all five rows.

You cannot lock YES and NO under a dollar here

The trade every beginner's guide describes, buying YES and NO for a combined price under $1.00, does not occur on modern Polymarket. The Gamma API exposes bestBid and bestAsk only, and the NO side is the arithmetic complement of the YES side, so yes_ask + no_ask = 1 + spread by construction. The sum is at or above $1.00 always. Two independent systems reached this from different directions, a session review on 28 July 2026 and our own collector on 31 July 2026.

Cross-venue arbitrage against Kalshi is a different question and it remains untested here. It needs a market-equivalence matcher that neither of our systems has built, and our own five-position book was already negative on entry as a two-leg strategy, at -$3.48 before any price moved.

The United States is close-only on frontend and API

The geoblock endpoint publishes three tiers, and reviews routinely collapse them into one.

The middle and bottom tiers are not the same thing, and the difference is the one that matters to anyone writing code. Ireland and Japan are restricted at the interface. The United States is restricted at the interface and at the API, so routing around the website does not change the status. Close-only is the endpoint's own label. Its ordinary meaning is that you may reduce or exit a position but not open one, and the endpoint does not expand on it.

Competitor restricted-country lists report 39 countries (Polymarket's own help centre), 33 (datawallet) and about 14 (polymarket101). All three disagree with each other and with the endpoint. Bulgaria appears on none of them and is not on the endpoint's list either.

Polymarket also runs a separate US property. We read polymarket.us on 5 August 2026 and it discloses nothing: no operator named, no CFTC or DCM registration status, no state availability list, no statement on whether a US user may open positions. A sign-up page and two links.

The referral programme pays on a fee one side of every trade never pays

Worth understanding before you read any review that carries an invite code, because the incentive shape explains the genre.

"Net" means what Polymarket keeps after the referred user's own rebate. Since makers pay nothing and geopolitical takers pay nothing, a referred user who posts resting orders or trades geopolitics generates exactly zero for the referrer, forever. A separate partner programme run through Dub pays $0.01 per click and $10 per first deposit with no volume gate mentioned on the page.

We take neither. PredictionEdge currently takes no referral revenue from Polymarket or Kalshi, and whether it ever will is an open decision.

Against Kalshi, in one paragraph

Kalshi charges takers a flat 0.07 on the same shape of formula, with one difference that matters at small size: their published version rounds the result up, roundup(0.07 x C x P x (1 - P)), documented on docs.kalshi.com for engineers rather than traders. Their HTML fee page is a stub that punts to a rate-limited PDF. Between that and Polymarket keeping its rates in an API field, no per-price fee grid for either venue exists on the indexable web. On coefficient alone Polymarket is cheaper in ten of eleven categories and equal in the eleventh, but coefficient alone is not the comparison: entry price moves the cost more than the venue does.

What we could not verify

• Whether Polymarket rounds fractional fees. Kalshi's published formula rounds up explicitly. We have the Polymarket coefficient from the API and no human-readable Polymarket fee schedule page. docs.polymarket.com/programs/builders/fees has never been fetched by us. At small size, rounding is the difference between the table above and what you actually pay.
• Whether a US resident can open a Polymarket account at all, whether they are routed to Polymarket US, and what a Polymarket US signup does. The geoblock endpoint gives a status, not a signup flow, and polymarket.us answers none of it.
• Whether the country list is current. The geoblock page carries no date and no version number, so it can change underneath a reader silently. Ours was read 5 August 2026 and that is all we can honestly say about it.
• How many of our own paper positions are open right now. Our correction register records 10 trades with 6 closed, which implies 4 open; the scanner note lists 5 open from 28 July. Our own records disagree and we have not reconciled them.
• The BBB and Trustpilot complaint figures circulating in this niche. Our research corpus holds a real, verifiable count of 174 of 214 complaints unanswered, but the record does not name which platform's profile it belongs to, so we are not attaching it to Polymarket here.
• The claim that a quarter of Polymarket trades are wash trades, an SSRN finding that winning accounts are limit-order makers, and an NPR report finding no sign of Polymarket at its stated Panama headquarters. All three are secondhand in our corpus with no document we can link. Treat them as unsourced until we can.
• Whether Kalshi charges makers, which the comparison above needs and cannot have. Their newsroom and their help centre contradict each other in our reading, one saying resting orders are fee-exempt and the other saying maker fees are charged, and we hold no URL for the newsroom statement. Nobody has reconciled it publicly.
• Why competing fee pages quote a single flat coefficient. Our own notes say one competitor page has asserted 0.0625 for five months, but the same notes record that domain returning HTTP 403 to fetchers, so we cannot re-check it and will not assert it.

FAQ

Does Polymarket charge a fee on every trade? No. Only the taker is charged, and only in ten of the eleven categories. Makers pay zero, which is a rate of 0.00 rather than a discount off something. Geopolitical markets carry a taker coefficient of 0.00 as well, so both sides of a geopolitics trade are free.

How much does a $100 trade cost on Polymarket? It depends entirely on price, not size. Taking $100 at $0.90 in a politics market (coefficient 0.04) costs 40 cents. Taking $100 at $0.30 in a crypto market (0.07) costs $4.90. The share of stake is rate x (1 - P), so cheap contracts cost the most as a percentage.

Are limit orders on Polymarket free? Yes, if they rest and get hit. The maker rate is 0.00 across every category. The cost of being a maker is not a fee, it is adverse selection: your resting order gets filled fastest by whoever knows something you do not. Free is not the same as costless.

Can I use Polymarket from the United States? The United States sits in the close-only tier, one of three tiers the geoblock endpoint publishes, and it is restricted on the frontend and the API both, as read on 5 August 2026. A separate polymarket.us property exists but discloses no operator, no registration status and no state list. We cannot tell you whether a US signup works, and nobody else has published a sourced answer either.

Is Polymarket cheaper than Kalshi? On coefficient, yes: 0.00 to 0.07 by category against Kalshi's flat 0.07, with geopolitical at 0.00 and every maker at 0.00. But Kalshi's published formula rounds the result up and Polymarket's rounding behaviour is undocumented, so at 10 or 20 contracts the coefficients are not the whole answer.

Can you arbitrage YES and NO on Polymarket? Not within a single market. The Gamma API exposes only the best bid and best ask, and the NO price is the arithmetic complement of YES, so the two asks sum to $1.00 plus the spread by construction. The classic under-a-dollar lock cannot appear. Cross-venue against Kalshi is a separate and untested question.

Why did two markets at the same price charge me different fees? Category. The coefficient is stored per market in a feeSchedule field, and there are eleven categories carrying four distinct values: 0.07, 0.05, 0.04 and 0.00. Two markets at $0.60, one crypto and one politics, differ by 75% in fee on identical size.

Where to go next

• Polymarket .com versus Polymarket US, the split behind the close-only status
• Polymarket fee calculator, category-aware, all eleven rates
• Kalshi review, the same arithmetic on the other venue