How Polymarket Price Discovery Works

• Quotes: makers post bids/asks on a 0.01 grid; the mid is the display value.
• Crossing: a taker crossing the spread fills against the book, moving price up/down.
• Cancellation: makers withdraw in bulk under uncertainty — the 26-second event showed five markets repricing as depth departed.
• Arbitrage: cross-venue bots pin Polymarket to reference prices; measured with Binance at the same second.

Three measured studies anchor the velocity claims: the strike ladder repriced from 0.044 to 0.996 in one second; the ETH high/low matched Binance's move to the same clock second (0.815 to 0.065 during a $59/65-second selloff); and a five-asset lower-bound cluster closed within 26 seconds.

Information travels first as depth. Bid size drains, ask size multiplies, spread widens — then the price moves. The interval between depth-first and price-later is the edge measurable on 250ms records.

• Off-chain book: depth is invisible on-chain, so discovery metrics need recordings.
• Minute grids: aggregate the exact sequence discovery happens at, hiding the ordering.
• Terminal windows: settlement itself is discovery ending, not a tradeable move.
• Thin books: with few quotes the mid is closer to noise; the archive lets you measure depth-adjusted confidence.

• For a reprice window at 250ms, order the events: reference print second, floor move second, mid move second, depth change frame; the ordering is price discovery in miniature.
• Identify which side led (the executable floor or the mid) and by how many frames; on one-sided books the floor leads and the mid follows.
• Join the venue rows to the external reference on the UTC second (the 14:42:50 pattern) and confirm the same-second alignment; that alignment is the discovery mechanism's proof.
• Write the window as a four-column table (second, reference, floor, mid) — the discovery paper in rows.

Price discovery on Polymarket is the ladder re-anchoring to its reference: the floor moves into alignment, the mid follows the floor, and the depth redistributes around the new level — a sequence the archive sees at 250ms but a daily chart never will.

The measured cases — the same-second futures/floor alignment at 14:42:50, the 26-second four-coin family reset, the 37-minute BTC lag — are the documentation of discovery rates as data, and they are the venue's own response to "do these markets price correctly."

Discovery is a rate, not a state: the second the reference prints, the second the floor moves, the second the mid catches up, the frames of depth redistribution. Publish the four columns with the window stated, and every reprice claim becomes a reproducible rowset.

The four-column discipline also forces completeness: a study that names a reprice must fill all four columns for at least the seconds around it, including the pre-move and post-move frames. That boundary requirement is what separates measured discovery work from commentary, and it is exactly what the archive stores for every recorded second.

Discovery is the floor leading, the mid following, the reference aligned, and the depth redistributing — all timestamped.

The measured alignment (same-second, 26-second, 37-minute) is how fast discovery really happens per market.

Replay the window to see discovery: the four-column table is the mechanism, made readable.

Discovery is the synthesis page: it welds the ladder, the reference, and the depth into one mechanism.

Read it after the mid, price-bounds, and repricing pages, and before the arbitrage page — it sets the price path the arb page trades.

The four-column recording (second, reference, floor, mid) is this page's method and the catalog's standard for reprice studies.

Discovery is the floor leading, the mid following, the reference aligned, and the depth redistributing — all timestamped.

Record it as the four-column table and reprice studies across the catalog become readable from rows.

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