Every stored NHL player-prop observation for one game or one player, newest snapshot first. 16,807,052 observations across 4,231 events, 2023-05-03 to 2026-06-14, 18 vendor markets over 9 canonical stats. Requires an Edge plan or higher (Free / Solo / Trio receive 403).

This is the ARCHIVE. It is a different product from /v1/nhl/players/{player_id}/props, which serves the CURRENT line: one row per book and market type, picked by a stated precedence. Here nothing is collapsed: the response grain is one observation, and season, snapshot minute, book, market, market type, line and outcome are each carried on their own row and each addressable as a filter. That is what lets you reconstruct line movement yourself.

One of match_id or player_id is required. An unanchored window query cannot be answered from the archive's indexes, so it is refused rather than accepted and timed out.

Unattributed rows are served, not dropped. Roughly 4% of observations name a player the bridge could not pin to exactly one of our players. Those rows are returned with player_id: null, the vendor's player_name intact, and attribution set to ambiguous or unmatched, never guessed onto a player id. meta.attribution_counts gives the split for YOUR selection. Note that a player_id-addressed query cannot return them at all, by construction; meta. unattributed_rows_reachable says so on every response.

There is no opening or closing label, deliberately. The board opens roughly 48h before puck drop, so no stored row is provably an opening price. Each row carries snapshot_at and the game's commence_time, plus before_kickoff, which is exactly the comparison snapshot_at < commence_time, made at read time. Derive "closing" as the last observation strictly before commence_time.

Response: { data: [{ match_id, home_team, away_team, commence_time, player_id, player_name, attribution, market, market_type, outcome, line, price_american, book: { id, name, logo_url }, source, snapshot_at, before_kickoff }], meta: { total, limit, offset, addressed_by, attribution_counts, unattributed_rows_reachable, markets_selected, snapshot_labels, generated_at } }.