Polymarket vs Kalshi Midterm Odds: 65 Matched Contracts, and Why Few Gaps Pay

Polymarket and Kalshi mostly agree on the 2026 midterms. Across 65 contracts that exist on both venues, from control of each chamber to 31 individual Senate races, the median pair's daily prices sat 1.4 points apart over the last 30 days. On the questions that matter most on 12 September, they told the same story: Democrats around 85 to 87 cents to win the House, the Senate a coin flip, Texas and Ohio close, Maine and Michigan leaning Democratic.

Where they differ, the difference has a direction. In 15 of the 18 most-traded outcomes, Polymarket priced the Democrat higher or the Republican lower than Kalshi over the past month. A Democratic House was dearer on Polymarket on 319 of 379 days since August 2025. Vultax did not measure why; the venues have different users, fees and rules.

That makes the gap look like free money more often than it is. At 18:56 UTC on 12 September, 19 of the 65 pairs had quotes that crossed between venues, but only 4 of the 40 pairs with both order books cleared both venues' fees at the size actually on offer, and the best of them paid $62.57 on about $24,000 tied up until the result, under two rulebooks that do not say the same thing.

The table lists the most heavily traded contracts on Kalshi, those with open interest above 500,000 contracts; Kalshi files Ohio as a special election. Bid and ask are the best prices to sell and buy a Yes contract. The gap is Polymarket's daily price minus Kalshi's, as a median over the 30 days to 12 September; positive means Polymarket was higher.

Download table figures:CSVJSON with sources

On the chamber-control contracts the pattern is long-running. Since 9 August 2025, the first day both venues had a daily price, Polymarket priced a Democratic House higher than Kalshi on 319 of 379 days, by 1.85 points on average, and a Democratic Senate higher on 234 of 358 days, by 1.05 points. Over the last month the Senate gap widened: Kalshi's Democratic Senate price rose from 41 to 49 cents between 29 June and 12 September while Polymarket's rose from 42.5 to 52.5.

The race contracts mostly follow suit. Of the 18 outcomes with more than 500,000 contracts of Kalshi open interest, Polymarket was higher on the Democrat or lower on the Republican in 15. Iowa went the other way on both candidates, and South Carolina showed no gap. A consistent lean like this is what you would expect if the two venues draw different traders, but a price gap alone cannot say whose traders are right.

Download chart figures:CSVJSON with sources

Most of the time, not far. The chart shows the average absolute difference between the venues' daily prices over the last 30 days, for the more heavily traded side of each contract. Chamber control and Maine ran about 3 points apart; North Carolina, where Roy Cooper trades above 90 cents on both, ran under 1. For comparison, Polymarket's politics fee on a trade at 50 cents is 1 cent per contract and the reported Kalshi taker fee is 1.75 cents, so gaps of 1 to 3 points are close to what it costs to act on them.

Download chart figures:CSVJSON with sources

A cross-venue trade buys Yes where it is cheap and buys No on the same outcome where Yes is dear, so that one leg pays $1 whatever happens. It works only if the two asks add up to less than $1 after both venues' fees, at a size the order books actually hold. Vultax walked both venues' books at 18:56 UTC on 12 September for every pair where both books were available, charging Polymarket's 0.04 politics rate and the 0.07 rate reported for Kalshi's standard fee on each level.

Nineteen of the 65 pairs had crossed top-of-book quotes, eight still showed a profit after fees at the top of the book, and four of the 40 pairs with both books cleared fees at depth. The table lists them. None is a small edge on a large trade; the largest is a 0.26% return on capital that stays locked until the result is called or, on Kalshi's Senate-control contract, as late as 1 February 2027.

Download table figures:CSVJSON with sources

A hedge across venues is only a hedge if both contracts pay on the same facts. They mostly will, but the rulebooks differ in ways that matter at the edges.

Download table figures:CSVJSON with sources

• The 2029 trap - Kalshi's Georgia and North Carolina pages also carry contracts for the Senate term beginning in 2029, elected in 2028. Matching those against Polymarket's 2026 races produced apparent gaps of 25 to 31 points; they were different elections, not mispricing.
• Party or person - a Kalshi contract labelled with a candidate pays on the party; Polymarket's pays on the candidate who wins. If a nominee is replaced before November, the two can split.
• Timing - Polymarket settles when three networks call a race; Kalshi's fallback is the swearing-in or the chamber's organisation in 2027, so the same position can tie up money for different lengths of time.

Kalshi contracts came from its public events and candlestick endpoints for House control, Senate control and every Senate race series, keeping only events for terms beginning in 2027. Polymarket contracts came from its public events and price-history endpoints for markets tagged midterms. Pairs were matched on chamber or state, then on party label or candidate surname, and checked by reading both rule texts; Alaska, Kentucky, Louisiana and Nebraska had no clean match. Ohio and Florida pair Kalshi's special-election series with Polymarket's race markets for the same seats.

Daily gaps compare Polymarket's daily price point with Kalshi's daily closing price on the same UTC date, which are not taken at the same second. Kalshi's fee coefficient is the rate reported by independent fee trackers, because Kalshi's fee schedule could not be retrieved from our environment; its API lists these series under its standard quadratic fee. Polymarket prices are from its international exchange, which U.S. residents cannot use; see where U.S. law stands. An earlier Vultax study measured the same kind of gap on matched Fed contracts. Nothing here is a forecast or a recommendation to trade. Challenge a figure at [email protected]; material changes are logged in the changelog.