How Polymarket Markets Resolve

• Each market has a resolution source — reference prices and market definitions Polymarket publishes with the market.
• On resolution, winning Yes tokens settle at 1.00 and losing ones at 0.00; a winning No is mathematically the same as losing Yes.
• The No/Yes mirror: a No best bid of 0.99 is exactly a Yes best ask of 0.01 seen from the other side (and reversed).
• Markets stay tradeable until resolution; the archive keeps them queryable afterward, including final volume and settlement metadata.

In a sample of 298 resolved BTC five-minute contracts drawn from 8,349 that settled between 2026-08-26 and 2026-09-23, the books ended one-sided at least 90% of the time, and the winning side sat at 0.99 or better in 98.6% of observations. The losing side kept asking; the winning side kept bidding, both at the near-certain price.

The final-second books never had both sides empty across 17,880 snapshots. The market does not vanish at settlement; it tilts until one side effectively owns the outcome.

A strategy that buys the 0.99 winning side in the last minute is not adding signals; it is absorbing a trade the market will not give you back. Rules that assume fills at the bid in the final seconds are the most common silent backtest error on Polymarket.

Historical depth is the only way to see the resolution collapse as it happened — which is why settlement studies in the blog are built on 250ms records, not on 1-minute prices.

• Take a resolved contract and separate its final 60 seconds (240 frames at 250ms); count one-sided frames and the frames where the ask was absent entirely.
• Read the resolution outcome for the contract and compute the executable floor at the reference second; the floor-vs-outcome check is the discipline a resolution study should not skip.
• Check the round-grid exit: how many frames priced at 0.99/0.995 up to the final window; the measured 98.6% share says most of resolution trading happens within two cents of the end.
• Save a per-contract resolution table: slug, final-minute shares, last floor, outcome — the sentence-length record of how a market ended.

Resolution is the venue's payoff moment, and the archive turns it into the most auditable event in the catalog: the final-minute flags, the one-sidedness, and the executable floor at the reference second are all stored, so "how did this market end" is a query, not a memory.

The measured final-minute pattern — ask presence in only 3 of 18 sampled one-second frames at sizes 1,685, 674, 80 — is the resolution textbook case: the liquidity drains deliberately as the outcome locks.

Resolution is where prediction market data becomes settlement truth, and the final-minute frames are the primary source for studying it.

The flags are independent of the reported outcome; study the flags first and let the outcome confirm, not define, the analysis.

Every resolution claim should carry its window (start to end, UTC) and its floor computation; the archive provides both, so the discipline is free.

Resolution reads backwards from the settlement pages and forwards from the final-seconds page; it is the outcome layer of the Up/Down studies.

Pair it with the settlement-data page for the corpus (8,349 contracts) and with final-seconds for the drain pattern at the very end.

Every resolved-contract study in the catalog relies on the same final-minute definition this page establishes.

Resolution is the venue's moment of truth and the final-minute frames make it the most auditable event in the catalog.

Study the final-minute flags first, let the outcome confirm, and always record the window plus floor.

• Order book data
• Price history
• Crypto data
• Provider matrix