Race history
for reproducible research.
Free public archive access with canonical driver and constructor identifiers. Results extend back to 1950; lap timing, stints, pit stops and weather have narrower, era-dependent coverage. Check the historical reference and record your request, retrieval date and returned coverage.
What researchers and students build on RaceHooks
Reproducible datasets for papers
Publish the exact HTTP requests and save the responses used in your study, so a reviewer can inspect the collection process and reproduce your analysis. Driver and constructor identity is canonical across every season, so a merge across eras does not silently mismatch on a renamed team.
Teaching datasets & student projects
A course needs data that is interesting, well-shaped, and available on the first evening without a procurement conversation. Results across seven decades, plus lap times, stints and pit stops where covered are enough for a term of regression, time-series, survival analysis, or classification work.
ML training corpora
Lap-level records with the context that makes them modellable — compound and tyre age, weather on the same clock, and the classification each session ended in. Enough structure to train on without spending the first month reconciling identities.
Sports science & engineering studies
Reliability by constructor, career trajectories by driver, circuit-by-circuit records, and per-round weather — the joins a study usually spends its first month building, already made and already keyed.
One payload, straight onto the screen
What a research use of this actually looks like: a table you can join, a record you can cite, and the identity contract that makes both safe.
The difference between a dataset a reviewer can re-derive and one they have to take on trust is entirely in these three steps.
Stated up front rather than discovered at a rate limit. Everything a study needs from the archive is on the free tier; the paid step is about live delivery volume, not about access to history.
Example interfaces, rendered from the fields above: your product, your design system. Nothing here is a template you have to adopt.
The endpoints that build it
Every route is Bearer-authenticated and versioned under /v1. Base URL https://api.racehooks.io/v1.
| Method | Path | What it gives you |
|---|---|---|
| GET | /v1/historical/seasons | Every season in the archive — the index a collection script starts from. |
| GET | /v1/historical/:season/:round/results | Final classification for a round, with qualifying, sprint, and practice on sibling routes. |
| GET | /v1/historical/:season/:round/laps | Lap-by-lap times for a session — the row-level table most studies are actually built on. |
| GET | /v1/historical/:season/:round/stints | Compound and stint plan per driver, with pitstops and weather on sibling routes for the same round. |
| GET | /v1/historical/drivers/:driverId/career | A driver's full career record; the constructors routes carry the equivalent, including reliability by season. |
| GET | /v1/historical/circuits | Canonical circuit identity and the rounds each has held — the key a cross-era join needs. |
session.statusSession lifecycle — started, aborted, finished, finalised. Free, and enough to timestamp a collection run.session.track-statusGreen, yellow, safetyCar, red — one named status. Free, and the covariate most race models want.session.lap-countCurrent lap and laps remaining. Free.driver.listNumber, TLA, full name, team and team colour — the canonical identity every other payload keys to. Free.weather.dataAir and track temperature, humidity, pressure, wind, and rainfall. Free, and on the same clock as the laps.timing.dataPosition, gaps, lap and sector times, live. The Developer step up, when a study needs the session as it runs.Developer includes all 29 live data feeds, not just the ones above; the remaining 26 in the 56-feed catalog are the Custom analytics feeds. Every id, cadence, and payload schema is documented.
Three ways the data reaches you
Pick per integration. You can run all three against one API key. Webhook and SSE frames are both counted as deliveries (Free pauses at its monthly limit; Developer has no delivery limit); REST reads count against the request rate limit instead.
Webhook
Push · defaultRegister an HTTPS endpoint and subscribe it to feeds. RaceHooks POSTs events as source updates arrive. Every payload is HMAC-SHA256 signed, retried with exponential backoff on failure, and every attempt is logged in your console with headers, status, and latency.
POST /v1/webhooks
{ "webhookUrl": "https://your-app.com/hook",
"feedId": "events.race" }SSE stream
Push · no public URLOpen one long-lived connection and read frames as they arrive — no public webhook server to host or expose. Each frame is `event: <feedType>` + a JSON `data:` line, with a ping every 30s so proxies hold the connection open. Auth by Bearer header, or `?token=` for browser EventSource, which cannot set headers.
GET /v1/stream?feeds=raceevent,timingdata
Authorization: Bearer $TOKENREST
Pull · on demandQuery completed and reference data directly whenever you need it: results, standings, laps, stints, pit stops, and the segment endpoints above. Use it to backfill history, hydrate a new user, or reconcile after an outage rather than to follow a live session.
GET /v1/historical/2026/16/results
Authorization: Bearer $TOKENTest the whole thing on a quiet Tuesday
Simulate replays a real current-season session through your own webhooks at any speed from 0.1× to 120×. Your integration receives the exact payloads it would in a live race (real data, known result) so you can validate research logic end to end before a live session ever runs. Free includes 5 sessions, Developer 10; Custom replays the full archive.
How Simulate works →What the ML layer adds for research
The archive above is free and stays free. Custom exists for work that needs the modelled layer or the raw channels underneath it — normally a lab with a grant, an industrial partner, or a licensing question. Model outputs publish only on supported sessions, schedules and eligible race states. Review the output contracts and evaluation scope with us.
Mini-sector timing
timing.mini-sectors is the high-granularity companion to lap timing — mini-sector progress ticks and speed-trap detail, for work where a lap is too coarse a unit.
Model outputs as a comparison baseline
The analytics feeds carry our own fitted surfaces — position distributions, true pace, tyre state — which is the honest thing to benchmark a new method against rather than a naive baseline.
Full-archive replay
Self-serve replay is capped at distinct current-season sessions. Custom lifts it to the whole archive, which matters when the object of study is the sequence rather than the table.
Start on the free tier.
Build and test a complete integration on historical and replayed sessions, with no credit card. Go Developer when you need it live.