we shipped a big upgrade to how
@gap133xyz matches markets.
until today our verifier answered one question: are these two markets the same event. that turned out not to be enough, because
@Kalshi and
@Polymarket often list the exact same question and still settle it differently.
their "next prime minister of israel" markets read word for word the same, but kalshi pays on the first new person to hold the office any time before 2045, interim holders included, while polymarket pays on whoever gets sworn in after the 2026 election, by the end of 2027.
so jev now answers two questions for every pair:
1. whether it's the same event, and
2. whether both venues settle it on the same terms, and
every row on the board carries a label: same contract, close terms, or terms differ.
we only mark a gap net positive when the terms line up, because on a terms differ pair part of that gap is just the price of the different rules.
we also added hard blocks for questions that share words but ask different things, like an endorsement vs a win or one house seat vs the whole chamber, and every telegram alert now tells you which kind of pair you're looking at.
open any row to see both deadlines and both scores.
the full receipts are on
gap133.xyz/audit
matching is the hard part.
"will X run for the nomination" and "will X win the nomination" are different questions but they seem to be wearing the same words.
one pays on an announcement, the other on an election.
every pair passes a deterministic matcher, then a second model (JEV) reads both resolution rulebooks and scores the probability they're the same event. below 80%, it doesn't touch the board.
[deterministic matcher proposes, Jev verifies, staleness gate polices]