Polymarket Arbitrage 2026: 716 Fields Priced
Polymarket arbitrage measured across 716 live multi-outcome fields on 19 October 2026: five were genuinely risk-free, worth $0.90 of profit in total.
Affiliate disclosure: contains affiliate links; we may earn a commission at no extra cost to you. 18+ only — see BeGambleAware.org. Last updated 19 October 2026.
Key takeaways
- Polymarket arbitrage is real and it is worth about a dollar. We priced every buyable multi-outcome field on the exchange. Five were risk-free after fees and depth. Buying all five takes $344 and returns $0.90.
- Twelve fields looked free. Seven survived a completeness check. Five of the twelve were only cheap because part of the field had stopped trading — including one at 49.7¢ on the dollar with just 12 of its 18 listed outcomes live.
- The taker fee is the wall. A field costs a median 3.22¢ in fees to buy, and the fee rises with the number of outcomes. Two of the seven honest underrounds died on it.
- At the midpoint the exchange has almost no margin at all — the mid-price field sums to 100.6%. The cost is not a built-in overround, it is the 7.10¢ round trip you pay to cross the spread.
- Liquidity is the whole story. Fields with over $200,000 of daily volume cost 1.00% above par; under $10,000 they cost 9.30%. Three-outcome sports fields are tightest at 2.00%, crypto widest at 34.10%.
On this page
Every guide to Polymarket arbitrage uses the same worked example: five candidates priced at 92¢ in total, buy all five, collect a guaranteed dollar, pocket 8.7%. It is a clean illustration and I have never once seen anybody check whether those prices exist.
That matters because the trade is advertised as risk-free, and risk-free trades attract people who cannot afford to be wrong. If the edge is really 8.7%, it is the best thing in crypto.
So we priced the field on every one of them. On 19 October 2026, between 18:41 and 18:46 UTC, we pulled 716 open multi-outcome events — 10,781 live contracts — and read the full order book on both sides of all 21,562 tokens straight from Polymarket's public CLOB. No estimates and no cached quotes: every number below is a price somebody was standing behind at that moment, and where our book disagreed with Polymarket's own cached quote we used the book.
What "buying the field" means on Polymarket
Polymarket groups mutually exclusive outcomes into a single event using a mechanism it calls negative risk. Exactly one outcome in the group can resolve YES, and the contract that does pays $1.
That gives you an identity to trade against. Buy one YES share of every outcome and you own a guaranteed dollar, whatever happens. So if the asks sum to less than $1.00, the set is mispriced against itself and the profit is locked at purchase. Here is a real field, the December Fed decision, the most heavily traded shape on the exchange:
Polymarket event fed-decision-in-december, $159,755 of 24-hour volume, read 19 October 2026 at 18:42 UTC. Adding the 1.89¢ taker fee, the guaranteed dollar costs $1.030 to buy and fetches $0.984 to sell.
There is the shape of the thing. A guaranteed dollar costs $1.03 and sells for $0.98, and the 4.6¢ between those two numbers is why nobody is getting rich here. The question is how unusual that is.
How often Polymarket arbitrage actually exists
Of the 716 events, 496 had every outcome quoted on the ask — the minimum requirement for buying a field at all. The other 220 had at least one contract with no seller at any price, which makes the trade impossible before you start.
Across the whole 10,781 contracts, that pattern is worse than it sounds. Only 4,756 (44.1%) were quoted on both sides. 3,071 had an ask and no bid, and 2,954 had no ask at all. More than half of all live multi-outcome contracts cannot be sold to anybody at any price.
For the distribution I use the 188 events where every outcome is quoted two-sided and every listed outcome is still trading — the only population where the field is genuinely both buyable and sellable. That is 1,101 contracts and $4.05m of daily volume.
188 two-sided, fully-tradeable multi-outcome fields, 19 October 2026. Median field cost $1.0415; the dearest in the sample asked $8.88 for a guaranteed dollar.
The median field asks $1.0415 for a dollar — a 4.15% overround. Two fifths sit inside 2% of par and two fifths cost more than 10% above it, and the split is almost entirely a liquidity story rather than a mispricing one. Only 2.7% of fields were under $1.00 before costs.
Twelve candidates, five trades
Across all 496 buyable fields, 12 priced under a dollar. That is where every arbitrage guide stops, and it is where the real work starts — because five of those twelve were not arbitrage at all.
The fourth row is the one that should worry anybody running a scanner. Polymarket lists outcomes that later stop accepting orders, and the event keeps its remaining prices. Sum those and you get a number under a dollar that is not a claim on a dollar, because the outcome that actually wins may be one you could not buy.
The VALORANT line is the one to remember. A scanner reading live quotes sees twelve teams totalling 49.7¢ and reports a 101% return. Six listed teams are not trading, so what you would actually be buying is a bet that the player signs for one of the twelve that are — which is not a guaranteed dollar, it is a punt with extra steps. Polymarket's own documentation makes the condition explicit: the outcome universe has to be complete before a conversion is worth anything.
The taker fee is the wall
Most arbitrage write-ups model Polymarket's cost as "about 2% on winning positions". That is not the rule. Per Polymarket's own fee documentation, the taker fee is charged in USDC at match time, on the trade, and it is:
fee = shares × feeRate × price × (1 − price)
Two consequences matter here, and both work against the field buyer. First, you pay it on every leg — a ten-outcome field is ten taker trades. Second, because of the p × (1 − p) term, the per-dollar cost of a contract is feeRate × (1 − p), which is largest on the cheap contracts that make up most of a wide field. Spreading a dollar across more outcomes therefore costs more in fees, not less:
Fee column assumes every outcome quoted at 1/n and a 5% rate. Overround column is the measured median across the two-sided fields in each band.
The median real field costs 3.22¢ in fees to buy. Rates come from each market's own live schedule, not a published table: politics, tech and finance run 4%, sports 5% on the newer schedule and 3% on the older one, crypto 7%, and 239 of the 10,781 contracts we read carry no fee at all. The direction of travel is simple — any underround narrower than about 3¢ is already gone, and that killed two of the seven honest candidates.
Depth: the trade that exists for $24
A best-ask price is a price for a quantity, and the quantity is usually small. Walking every ask ladder and buying the same number of shares of every outcome gives the real cost curve:
Cost includes the taker fee at each fill price. A third of fields cannot supply 10,000 shares of every outcome at any price.
Not one field on the exchange was profitable at 1,000 shares. At 100 shares exactly one was. The arbitrage is a quote-sized phenomenon, and a field's capacity is set by its thinnest leg — the German chancellor field was stopped at a single share by one outcome with 0.8 shares offered.
The best trade available anywhere on Polymarket that evening was the OpenAI IPO market-cap field: seven buckets including a "No IPO by December 31, 2027" catch-all, so the universe really is complete. The asks summed to 92.80¢, a 4.51% edge after fees — the kind of number the guides promise. Here is the trade in full:
Buy 25 shares of all 7 outcomes → capital $23.57 + fee $0.68 = $24.25
Guaranteed payout at resolution: $25.00 → profit $0.75
Twenty-five shares is where the ladders run out. Past that the field prices above a dollar, so the entire 4.51% edge on the best arbitrage on the exchange is worth seventy-five cents, locked up until OpenAI either floats or does not by the end of 2027.
What the whole thing pays
Five fields survived every filter. This is all of them — the complete set of risk-free trades available on Polymarket's multi-outcome markets at 18:46 UTC on 19 October 2026:
Ninety cents. That is the total value of every risk-free multi-outcome position on Polymarket at one moment, for $344 of capital locked up for months, before a cent of gas or a minute of your time. Four of the five pay less than a dime.
Set that against the one serious study of this. Researchers at IMDEA Networks and the Oxford Internet Institute traced $39,587,585 of realised arbitrage profit across 86 million bids on markets resolving between April 2024 and April 2025. The money was there. Their window is also the 2024 US election, on an exchange with no taker fee at the time — and the fee arrived in 2026. That number describes a different market, and it is the one every 2026 guide still quotes.
The overround, and how it compares to a sportsbook
Sum the implied probabilities of a mutually exclusive field, subtract one, and you have computed an overround — the same arithmetic that tells you a bookmaker's margin. So the 4.15% median is directly comparable to a betting market, and it splits cleanly by liquidity and category.
Category assigned from each field's live fee schedule. Bands with fewer than five fields omitted.
Now the comparison worth making. A published 2026 margin panel of over four million odds observations from 52 bookmakers puts the average three-way football margin at 7.67%, two-way head-to-head at 4.77% and player props at 7.93%. Polymarket's deep fields sit at 1.00–2.00%, and even its 4.15% median beats the three-way bookmaker figure. On the liquid markets it is genuinely the cheaper place to price a field.
But the structure is different in a way the single number hides. At the midpoint, Polymarket's fields sum to 100.6% — essentially no margin at all. A bookmaker's overround is built in and you cannot avoid it; Polymarket's is a 7.10¢ round-trip spread you pay for crossing, and a patient order avoids most of it. That is the real difference between the two, and it is also why the arbitrage is so thin: there is no systematic margin sitting there to be collected.
Check the field before you trust the sum
Every figure here is one five-minute snapshot of the order book. Prices, depth and fee schedules all move — and a part-closed field will lie to you. Read the live market before you act on any of it.
Visit Polymarket.com →Affiliate link. 18+ only. Polymarket restricts access from the United States, Canada and other listed jurisdictions. Prediction markets carry a substantial risk of loss. Nothing here is financial advice.
Who actually captures Polymarket arbitrage
The fee schedule answers this on its own: makers are never charged. Every number in this article is a taker's number, and the taker is the one paying 3.22¢ a field to find out there is no edge.
Post limit orders instead and the arithmetic inverts. You pay no fee, you earn the spread rather than crossing it, and on eligible markets you can collect maker rewards on top. The catch is that you are no longer arbitraging: a resting order fills when somebody chooses to hit it, so you might end up with four legs of a five-leg field and an open position you did not want. That is market making with inventory risk, not a locked dollar.
The conversion mechanism is the other asymmetry. Polymarket's adapter lets you turn a NO share in one outcome into a YES share in each of the others, atomically — but only in that direction. The reverse trade has no equivalent shortcut, so the sell side of the field has to be assembled leg by leg. That shows up in our data: five fields had bids summing above a dollar, and not one of them survived the fee.
What this cannot tell you
It is one snapshot, five minutes wide, on a Sunday evening in October. Arbitrage is an event, not a state: a field that prices at par now may gap at the next piece of news, and a five-minute read cannot see how often that happens or how long it lasts. Measuring frequency properly needs a websocket feed running for weeks.
We sampled the 716 highest-volume open multi-outcome events, not all of them, and the tail below $1,000 of daily volume is thinner still — so if anything these figures flatter the exchange. We also priced the YES side only; a mixed basket of YES and NO legs is a wider search than this.
Finally, "risk-free" is doing some work. A complete, fully-trading field still carries resolution risk — the UMA oracle decides what happened, and disputes do occur. Capital is locked until settlement, which on the OpenAI field could be the end of 2027. A 4.51% edge held for two years is not 4.51% a year.
My verdict
The mechanism is real, the mispricings are real, and the opportunity is not there. Of 496 buyable fields, five were risk-free after fees and depth, and together they were worth 90 cents. If you are reading a guide that promises an 8.7% edge on a basket of candidates, it is describing an exchange with no taker fee, which stopped existing this year.
What the exercise is genuinely good for is reading the market. A field's overround is the best single liquidity gauge on Polymarket: under 2% and you are in a market that can take size, over 10% and you are the liquidity. That is worth more than the ninety cents.
And the trap is worth internalising even if you never run a scanner. Five of the twelve cheap fields we found were cheap because part of the field had stopped trading — including one apparently offering a 101% return. Before you believe any sum under a dollar, count the outcomes and check every one of them is still open. That check is the difference between an arbitrage and a bet you did not know you placed.
Related guides
- Polymarket Fees 2026: What You Actually Pay — the full taker schedule by category, and the live rates that decide every number above.
- Polymarket Slippage 2026: 897 Books Measured — the single-market version of the depth problem that caps every field at quote size.
- Polymarket Odds Accuracy 2026: 9,836 Scored — whether the prices in these fields are any good in the first place.
- Polymarket Maker Rewards 2026: $852 a Market — what the exchange pays the makers who never pay a fee.
- How Does Polymarket Work? Explained for 2026 — the YES/NO share mechanics the $1 identity rests on.
- Polymarket vs Kalshi 2026 — the regulated venue, and why cross-platform arbitrage is a different animal.
Polymarket arbitrage FAQ
Is Polymarket arbitrage still profitable in 2026?
Barely. Across 496 buyable multi-outcome fields on 19 October 2026, five were risk-free after the taker fee and real order-book depth, and buying all five would have turned $344.10 of capital into $0.90 of profit. The mechanism works; the opportunity has been priced out.
How does multi-outcome arbitrage on Polymarket work?
In a negative-risk event only one outcome can resolve YES, and that contract pays $1. Buy one YES share of every outcome and you hold a guaranteed dollar. If the asks sum to less than $1.00 you have locked a profit at purchase. The catch is that the fee applies to every leg and the quote only covers a few shares.
What fee does Polymarket charge on arbitrage trades?
The taker fee is shares × feeRate × price × (1 − price), charged in USDC at match time, with rates from 3% to 7% depending on the market's own live schedule. You pay it on every leg, so a median field costs 3.22¢ in fees for each dollar of guaranteed payout. Makers pay nothing at all.
Why do some Polymarket fields sum to less than $1 without being arbitrage?
Because part of the field has stopped accepting orders. Five of the twelve underpriced fields we found were artifacts of this: one VALORANT market summed to 49.7¢ with only 12 of its 18 listed outcomes still trading. You cannot buy the outcome that might win, so the sum is not a claim on a dollar. Always count the live outcomes.
How much money can you put into a Polymarket arbitrage?
Very little. No field in our sample was profitable at 1,000 shares per outcome, and only one was at 100. The best trade on the exchange — the OpenAI IPO field at a 4.51% edge — ran out of depth at 25 shares, or $24.25 of capital for 75 cents. A field's capacity is set by its thinnest leg.
What is Polymarket's overround compared to a bookmaker?
The median multi-outcome field asked $1.0415 for a guaranteed dollar, a 4.15% overround, falling to 1.00% on fields above $200,000 of daily volume. A 2026 panel of four million odds observations from 52 bookmakers puts the average three-way football margin at 7.67%. Polymarket's liquid markets are cheaper, and at the midpoint its fields sum to just 100.6%.
Can you still make $40 million from Polymarket arbitrage?
That figure comes from an IMDEA Networks study of markets that resolved between April 2024 and April 2025 — the US election period, on an exchange that charged no taker fee then. It found $39.6m of realised profit across 86 million bids. It is a real number about a market that no longer exists, and it is the figure most 2026 guides are still quoting.
Sources and method
Population: the 716 highest-volume open negative-risk events on Polymarket carrying at least two live order-booked contracts, read between 18:41 and 18:46 UTC on 19 October 2026 — 10,781 contracts, $21.96m of 24-hour volume, from 1,400 open events scanned. Event and fee metadata came from the Gamma API paged at 20 events per request; full bid and ask ladders for all 21,562 tokens came from the CLOB /books endpoint in batches of 25, returning 16,233 books (the rest had no resting orders). Distribution figures use the 188 events where every contract was quoted on both sides and every listed outcome was still accepting orders; the funnel uses all 496 events whose every contract had an ask. Field cost is the sum of best asks plus the per-leg taker fee from each market's own live feeSchedule; capacity is found by walking each ask ladder and buying equal share counts across all legs. As a join check, our book-derived best ask matched Polymarket's cached quote within 1.1¢ on 93.4% of 7,827 comparable contracts, with a median difference of zero. One five-minute snapshot is a measurement, not a law, and nothing here is financial advice.
- Polymarket docs: fees — the fee = shares × feeRate × price × (1 − price) formula, charged in USDC at match time, and "makers are never charged fees".
- Polymarket docs: negative risk — the one-winner guarantee, the NO-to-YES conversion, and the requirement that the outcome universe be complete.
- Saguillo, Ghafouri, Kiffer & Suarez-Tangil, "Unravelling the Probabilistic Forest: Arbitrage in Prediction Markets" (AFT 2025) — DOI 10.4230/LIPIcs.AFT.2025.27. $39,587,585 of realised arbitrage profit across 86 million bids, 8,659 single-condition and 1,578 negative-risk markets resolving 1 April 2024 to 1 April 2025.
- Glush, "Bookmaker Margin Panel" (Mendeley Data, 3 August 2026) — DOI 10.17632/yt3xsvk7zn.1, CC BY 4.0. Over 4 million odds observations from 52 bookmakers across 22 leagues, 29 June to 2 August 2026: three-way soccer h2h 7.67%, h2h 4.77%, spreads 4.90%, totals 5.45%, player props 7.93%. A commercial dataset, not peer-reviewed.
- Gamma API: /events — the event grouping, negative-risk flag, per-contract fee schedule and order-acceptance state used throughout.
- CLOB API: /book — the single-token version of the endpoint every price and depth figure above is read from.