Polymarket Final-Seconds Data

In the published example, a resolving contract's final minute shows the bid pinned at 0.990 while the ask side appears in only 3 of 18 sampled frames — at 09:19:41 with 1,685 shares, 09:19:46 with 674, and 09:19:49 with just 80. For the remaining fifteen seconds nothing could be bought at any price.

The price series for that minute reads a flat 0.99. The book reads a market with one side gone. Charts display the first; backtests need the second.

• second-hammered UTC timestamps (09:19:41 through 09:19:59 in the example)
• bid and ask prices on the 0.01 grid
• top-of-book depth on each side, in shares
• a side-state label: two-sided, bids only, asks only
• cumulative volume so later fills can be matched to their frames

One-second frames derive from the 250ms archive by bucketing: take the last row of each second. The methodology carries to any series — cricket overs, election nights, or five-minute crypto — by filtering to final-window frames.

The pattern generalizes beyond the example: at settlement-scale, measured BTC books end one-sided in at least 90% of resolved contracts. The single-minute record is the microcosm.

• Take any resolved contract and isolate the final 60 second frames (240 rows at 250ms); count ask presence per second — the recorded 09:19:41–09:19:59 window shows the extreme: asks present in only 3 of 18 frames, at sizes 1,685, 674, and 80.
• Mark the descending size path (1,685 → 80) as the archetype of deliberate drain: the ask is being consumed or withdrawn as the contract resolution approaches.
• Compute the binding question at each final frame: was the executable floor, not the last mid, the price you could have exited at; the floor is the number that matters when the ladder is one-sided.
• Include the final-seconds view in the per-market settlement table so the resolution behavior and the exit liquidity are one rowset, not two.

The final seconds are where the venue rehearses its settlement: ask presence drops toward single-digit seconds, sizes shrink a full order of magnitude, and the market that had 35–36 sports levels or a wide Up/Down ladder ends as a one-leg book with a floor that quietly decides the outcome.

This is why final-second studies, not mid charts, are the right substrate for settlement forensics: an ask-less book is not a bug, it is the market's last statement, and the archive frames it exactly.

The final minute is a separate microstructure: ask presence in 3 of 18 frames at 1,685, 674, 80 is a measured, replayable pattern, not an anecdote.

Every exit-order strategy should benchmark against the final-second executable floor, because that is the price the book actually offers at the moment of truth.

Record the final-minute flags with every contract you study; the one-sided finale is information about the market's end that no interval chart preserves.

Final-seconds data is both an ending study and a liquidity study; read it with settlement-data for the outcome and with bid-ask-data for the mechanical shape of the one-sidedness.

Each measured case in this page is stored at 250ms with the same field names as every other dataset, so the switch from one page's material to another is a query-parameter change, not a migration.

That shared format is the reason a family can be studied across settlement, repricing, replay, final-seconds, and cross-market frames with one pull and one schema.

Final-second studies should always be anchored with the window boundaries in the report (start instant to end instant, UTC); a settlement number without its seconds window cannot be replayed by anyone.

The 3-of-18 ask-presence pattern and the 1,685–80 size drain are measured across the recorded window; treat them as the reference distribution for any final-minute claim you publish.

• Research blog
• Order book data
• Backtesting data
• Crypto data