The market knew the difference. It just wasn't pricing it

For 43 of the 50 days both traded before the closure, Kalshi's Hormuz contract closed above the Polymarket contract that strictly contains it — a bound each venue keeps on its own book.

Published 2026-08-20 · analysis

Two contracts in the file ask about the same thing. Polymarket's is “Iran closes Hormuz by 2027.” Kalshi's is “Iran closes Hormuz 7+ days before 2027.” Whatever the second title means — a closure lasting at least seven days, or a deadline seven days earlier — every world that settles the Kalshi contract YES also settles the Polymarket one YES. The Kalshi contract is a subset of the Polymarket contract. Its price cannot be higher.

It was higher on 43 of the 50 days the two overlapped before the strait closed, 9 January to 27 February 2026. Polymarket was higher on five days and they tied on two. Kalshi averaged 32.04% against Polymarket's 26.58%, a mean violation of 5.46 points, median 6.25, peaking at 13.5 on the first shared day. Twenty-eight of the 50 days broke the bound by more than five points and six by more than ten. The longest unbroken run was 14 days.

Three explanations do not survive the file. Stale quotes: Kalshi's close is a bid–ask midpoint on no-trade days, but the Kalshi price changed on 59 of its 61 day-pairs, so it was not a parked number. A venue clock offset: the Kalshi series here run a day behind Polymarket's, and shifting them either way leaves the result intact — 46 of 51 days broken at minus one day, 41 of 49 at plus one. Sloppy pricing generally: each venue polices nesting on its own ladder almost perfectly.

That last one is the control. Across every pair of dated contracts within a single venue where one deadline is strictly earlier than the other — so the earlier one is a subset and must price at or below the later — there are 2,778 overlapping contract-days in this dataset and exactly one violation: two points on one day in April, between Kalshi's before-August and before-September contracts. Polymarket's own nested closure pair runs 58 days with the narrower contract sitting a mean 12.9 points below the wider one, which is what a correctly signed bound looks like. One violation in 2,778 within a venue, against 43 in 50 across them.

Then it reversed, and the reversal is dated. On 27 February, the day before the IRGC declared the strait closed, Polymarket closed at 42.50% while Kalshi ticked down to 36.00%. From that day to settlement, 13 trading days, the bound was never broken again.

What replaced it was a price on the clause. On 4 March, Polymarket closed at 85.45% and Kalshi at 34.00%, its lowest reading since the declaration. That 51.45-point same-day gap, on two contracts with identical time to expiry, is the seven-day duration clause and nothing else: implied conditional odds of 0.340 divided by 0.8545, or 39.8%. It was a day on which zero vessels transited Hormuz, with the 2–4 March mean of 2.67 a day running 96.1% below the 68.20 a day of the preceding thirty days. The ratio recovered to 0.910 by 7 March and 0.991 by the 11th as the week elapsed. Kalshi settled YES on 10 March, Polymarket on the 14th.

Read forward, that sequence is awkward. For fifty days the narrower contract was too dear by an average of 5.5 points. Then, the moment the event was real, the same market showed it understood perfectly well that the clause makes the narrower contract cheaper, and marked it to 40% of the wider one. Both cannot be beliefs about the world. The pre-event premium was basis — something about the two venues rather than about Hormuz — and it was larger than the thing the contracts actually differ on.

† Limits. One strictly-nested cross-venue pair, one crisis, 50 days — and a 14-day unbroken run means those are not 50 independent observations. Not an arbitrage claim: it needs collateral on two venues that do not net, across fees, and the Kalshi figure is a midpoint rather than a bid. The venues' resolution criteria are not in the data, so the nesting is read off the titles — though during the event week Kalshi's criterion proved harder to satisfy, not looser, which is the wrong direction to explain a pre-event premium. It says nothing about forecasting skill.

Open in the interactive site → · the feed · email me when something breaks