Kalshi vs Polymarket arbitrage: we measured it, and here is what is left
The idea sounds free: the same event priced differently on two venues, buy both sides, lock the difference. We built a system that watched both venues around the clock for weeks, matched contracts by their published rules, priced both order books, and applied each venue's published fee. This page is what it found, updated nightly.
What the measurement found
- Against the venue a US account can actually trade (Polymarket US), 27 matched pairs produced zero executable gaps. The two US venues are linked by the same fast traders and agree on price.
- Every gap the system ever recorded was against Polymarket's offshore venue, which US accounts cannot use.
- The best case, taking the numbers at face value: about 3.4% annualized on capital locked for five to eight months, on order books holding tens of dollars.
- Fees are a curve that peaks at 50¢ on both venues (see the calculator). Most "gaps" people screenshot are smaller than the two fees combined.
- And the contracts are usually not the same bet: settlement sources, delayed-data rules, and rounding differ, so a "locked" position can settle two ways (real cases).
Today, across the pairs we track
24 US-tradeable pairs measured at the last rebuild; 0 show a positive gap after both venues' fees. A negative number means buying both sides costs more than the $1 you get back.
| Pair | Gap after fees | Contracts available | Rules |
|---|---|---|---|
| Will the Federal Reserve Cut rates by 25bps at their October 2026 meeting | -0.3¢ | 901.0000 | Not verified |
| Will real GDP increase by more than 1.0% in Q3 2026 | -0.8¢ | — | Not verified |
| Will the Federal Reserve Cut rates by 25bps at their September 2026 meeting | -1.1¢ | 1565.9000 | Not verified |
| Will the rate of CPI inflation be above 2.9% for the year ending in August 2026 | -1.1¢ | — | Not verified |
| Will the rate of CPI inflation be above 4.0% for the year ending in August 2026 | -1.1¢ | — | Not verified |
| Will the rate of CPI inflation be above 3.9% for the year ending in August 2026 | -1.1¢ | — | Not verified |
| Will the rate of CPI inflation be above 3.8% for the year ending in August 2026 | -1.1¢ | — | Not verified |
| Will the rate of CPI inflation be above 3.7% for the year ending in August 2026 | -1.1¢ | 25.00 | Not verified |
| Will real GDP increase by more than 0.5% in Q3 2026 | -1.5¢ | — | Not verified |
| Will the rate of CPI inflation be above 3.0% for the year ending in August 2026 | -2.1¢ | — | Not verified |
Gap = $1 minus the cost of buying YES on one venue and NO on the other at the best available prices, minus both venues' taker fees, per contract. "Contracts available" is the thinner side's depth at that price.
Where the numbers come from
The instrument is a paper-only research system with a public record: four ideas for making money in prediction markets, each tested against thresholds fixed before the data, each returning a negative. The write-ups are at https://github.com/omac049/atlas. Nothing here is a strategy or a recommendation; it is a measurement.