Polymarket Trade History

• Per fill: price, size, side (buy/sell), market, and timestamp on the 250ms grid — the rows the tape is built from.
• Cumulative volume at the frame level reconciles with the sum of fills for the same market window — the reconciliation that validates ingestion.
• Fills are timestamped at the same grid as books, so impact studies join the tape to the ladder exactly.
• The tape covers every market side the books do; a one-sided book near settlement has a tape that shows it being drained.

The tracked-fills study (4 fills vs market prices) shows the slippage dichotomy: fills can match the executable price across, and slip the wide end of the spread on a reprice spike.

The VWAP-relevated tape is the honest basis for execution-quality claims: fill price vs the frame executable floor at timestamp.

The median size of 42,254 shares and p90 of 65,081 (September sample) frames the tape: fills at that size pull multiple ladder levels when depth is thin.

• Replay your own executions: match your order ids against the tape rows and compute realized slippage vs the floor.
• Flow attribution: buy-side concentration, by marker type, tells who is taking which side of a repricing.
• Settlement forensics: the final-minute tape around ask-absent frames (3 of 18) shows exactly how exit liquidity behaved.

• Pull one day of fills for a liquid market and compute total traded shares; reconcile against the frame-level cumulative volume at end-of-day.
• Split fills by side and interval; the buy-side share per five-minute window is a flow fingerprint of the market's activity.
• For a reprice window (the recorded 65-second ETH move is a clean example), join every fill to the frame at that second and compute the fill-vs-floor distance.
• Build a daily tape table (Parquet) sorted by (market, ts); it becomes the input for every VWAP and execution query downstream.

The tape is the only fully independent truth on the venue: books and prices can be synthesized or sampled, but a fill is an event that happened at a price and size in a timestamp, and nothing can claim it otherwise.

The measured good news: fills can execute at the published executable price; the measured bad news: in a reprice spike they slip the wide end of the spread — both statements read directly from the tape.

Size conditions everything: median fill of 42,254 shares (p90 65,081) on tops that are often smaller means the tape is routinely pulling through multiple ladder levels.

The tape powers VWAP, execution analytics, flow attribution, and settlement forensics; keep it append-only and reconcile it against frame volume daily as a data-quality tripwire.

Your fills and the archive's fills should agree on the same market, same second; when they stop agreeing, one of your toolchains is the bug.

The tape is the only fully independent truth on the venue: a fill happened at a price, a size, and a second, and nothing can synthesize that — penetration testing your tooling against the tape is the strongest hygiene there is.

Reconcile fills against frame volume daily, split by side into flow fingerprints, and join each reprice-window fill to its frame to see the executable floor distance: the measured pattern (clean fills, reprice-spike slippage) is right there in the rows.

Size is the co-star: median 42,254 shares on shallow tops means big fills routinely walk the ladder — tape analysis that ignores size will mislead you systematically.

Keep the tape append-only as your highest-trust dataset; it is the numerator for VWAP, the evidence for impact studies, and the audit trail for every execution number you publish.

• trades data
• vwap data
• order flow analysis