Do Kalshi and Polymarket odds agree?

How the disagreement is distributed

Mean 1.23pp, widest 15.85pp. The mean sits well above the median because a small number of pairs are a long way apart, so the median is the number to quote. Each pair is counted once even where more than one page here points at it.

The widest gaps right now, and why

settled on one venue, still trading on the other — one venue has already resolved it in play while the other still shows a pre-game price.
both near the boundary, where a tick is a large percentage — both sit near 0¢ or 100¢, where a one-tick move looks like a large percentage gap.
sides do not match — the two legs are opposite sides of the event and the pair is not comparable.
no mechanical cause detected — none of the above applies, which is not the same as proving the books disagree — a stale quote on a thin book looks identical from here.

What a wide gap usually means

The useful reading of this table is the opposite of the one a scanner invites. Two independent books, with different users, different fee structures and different countries of origin, landing within half a point of each other on most of the board is strong evidence that both are priced about right. So when one pair is a long way apart, the first question is not "how do I trade this" but "what is different about this pair" — and the candidates are a clock, a boundary, a rule, or a book nobody is trading.

• A clock. Live markets move at different speeds on the two venues. A game that has effectively been decided reads 99¢ on one side and a stale pre-game number on the other. The gap is real and untradeable: by the time an order fills, the slow side has caught up.
• A boundary. Near 0¢ and 100¢ the tick size dominates. A contract at a fraction of a cent against one at 13¢ is a thirteen-point gap and, in ticks, almost nothing.
• A rule. The two contracts settle on different criteria, different deadlines, or different sources of truth. Of 131,140 candidate pairs checked, only 4,631 were confirmed to settle identically — see the methodology for how that is tested.

When none of those apply, the honest answer is that we cannot see a cause — which is not proof there isn't one. A stale quote on a book nobody is trading looks exactly like a considered difference of opinion from the outside. Where the gap does survive that doubt, fees decide whether it is worth anything: a spread smaller than about two points is inside the cost of crossing both books.

A note on what this page nearly said

An earlier version of this measurement reported a mean gap roughly twice as large. Three pairs had the Kalshi leg on the wrong side of the game — Kalshi names the side after a dash, so "GB Packers vs NY Jets — New York J" is a market on the Jets, while Polymarket's "Packers vs. Jets" resolves YES for the Packers. Subtracting those two prices produced a 23-point "gap" between two numbers that summed to 100.5¢, which is agreement.

Those pairs are repaired and the check that caught them now runs on every page, so a pair whose legs look like opposite sides has its comparison withheld rather than published. The figures above exclude any such pair; right now 0 are excluded, because there are none left to exclude.

Recomputed from the live board every time this page is served. Prices move; anything quoted from here is a snapshot.