Conclusion of an exhaustive hunt (Kaggle, Betfair, Oddsportal/BetExplorer/ OddsChecker, GitHub/HF/odds-APIs, academic replication data, archive.org): NO ready-made multi-season per-driver dataset exists anywhere public.
Viable routes, in order:
- Wayback-archived Oddschecker winner pages (chosen). ~370 captures, 2007-2025, gaps 2011-12 and 2017-18 (URL migration); ~85-95 races with clean pre-race (<=3 day) snapshots carrying the full driver x ~22 bookmaker grid, server-rendered, verified parseable across three page eras. Personal/academic use of archived pages; publish only derived consensus implied probabilities, never the bookmaker matrix (compilation/database rights). Pipeline: research/f1_odds_fetch.py.
- Betfair Historic "Other Sports", Basic tier (free): systematic full-field exchange prices ~2015-onward; US-geo-blocked; bz2 JSON stream files needing replay to last-pre-off prices; strict no-redistribution license. The free Kaggle zygmunt/betfair-sports week (2014 Italian GP) previews the schema. Layer in later from a UK/AU IP for closing-price quality.
- Fragments: Henderson & Kirrane 2013 championship ante-post series (github.com/d-a-henderson/F1); Fahy/Butler/Butler 2016-2020 top-5 odds (unpublished, r.butler@ucc.ie); Polymarket championship archiver (OnlineAdventuress/gridodds) as a forward-collection template.
Dead ends, verified: Oddsportal has NEVER carried F1 (all motorsport paths 404; the circulating URL pattern is apocryphal); BetExplorer likewise; Kaggle has no multi-season F1 odds; The Odds API has no motorsport; no academic deposits of race-winner odds.
Power note: ~90 races cleanly measures the market ceiling (large effect) but gives only wide error bars on a beat-the-market pool exponent; the derivative-market thesis (H2H/top-N priced off Harville vs our slab pricing) requires odds these pages lack and stays prospective.