Two exchanges. The same question. Two different prices. When Kalshi says a thing is 62% likely and Polymarket says 69%, one of those two crowds is wrong, and the gap between them is the most interesting number in prediction markets. Nobody publishes it.
The reason nobody publishes it is not the prices. It is the matching: proving that two differently-worded markets are actually asking the same thing. Get that wrong and you print an edge that does not exist. So we are paranoid about it, and we show you our confidence on every row.
A gap is not an arbitrage
Read this before you read the board. Kalshi gives us a firm ask, a price you can actually hit. Polymarket’s public feed gives us a last price, which is not an offer: the book behind it may be thin, wide, or gone. So a gap here means the two crowds disagree, not that money is lying on the floor. A real arbitrage needs a crossable offer on both sides at the same instant, in size, after fees, after the cost of parking capital on two venues, and only when the two markets settle on identical rules. Rules are where these trades die: two markets can ask the same question in English and resolve differently on the edge case that actually happens.
Comparing both exchanges…
Gaps close while you are asleep
Here is the honest problem with a board like this: a gap is only worth anything while it is open, and the wide ones rarely open at a moment when you happen to be looking at this page. By the time you refresh, the two crowds have usually converged and the moment has passed. Pro emails you the instant Kalshi and Polymarket diverge by 8 cents or more on a market we are confident is genuinely the same question on both venues. One email per market, ever, so it never nags. The board is free and always will be. Being told about it in time is $19 a month.
A pair has to clear four tests, and failing any one of them means no pair at all rather than a low-confidence guess.
1. Subject overlap. The meaningful words of the two questions (ignoring filler like “will” and “the”) have to genuinely overlap, not just share a topic. 2. Strikes must agree exactly. This is the one that matters most. Any threshold in the question ($100k, 3.1%) must be identical on both sides. “Bitcoin above $100k on July 31” and “Bitcoin above $120k on July 31” share almost every word, the month, and even the day, and they are completely different bets. A matcher that pairs them prints a thirty-cent gap that is not there. We separate thresholds from incidental numbers like dates so this cannot happen. 3. No contradicting month. “By March” and “by June” are different bets. 4. Settlement windows within a week. Two markets on the same question that settle three weeks apart are not the same market.
Prices refresh every couple of minutes. Kalshi figures come from the live book we already maintain for the rest of the site; Polymarket figures come from its public Gamma feed. When a board comes back empty, that is a real answer and not a failure: the two exchanges list mostly different questions, and an empty board beats a dishonest one.
Looking for a genuinely risk-free trade on one venue instead? Impossible Prices finds ladders where Kalshi’s own book has broken its logic, and Locks tracks one-winner fields priced under 100 cents. For how the two exchanges differ on fees, custody, and taxes, see Kalshi vs the alternatives. Educational, not financial advice.