Method
Data pipeline
Every number on this site comes from one deterministic pipeline. Same inputs, same outputs, every time.
Inputs
Data sources
Session timing
Lap, sector and speed-trap times for every session since 2018.
Car telemetry
Throttle, brake, gear and speed traces sampled per lap.
Weather
Track and air temperature, humidity, wind and rainfall per session.
Tyre and stint logs
Compound, stint length and pit-lane deltas per driver.
Pipeline
Four stages
- 01Ingest — pull session, telemetry and weather data, then freeze it as a versioned snapshot.
- 02Normalise — clean out-laps, traffic laps and red-flag runs so pace comparisons are like-for-like.
- 03Model — fit per-circuit pace, degradation and pit-loss models, then produce projections with spreads.
- 04Publish — write the read-only artefacts this site renders; each carries its run timestamp.
Last run 24 Aug, 18:50 UTC · complete through R12
Honesty
Known limits
- Projections are pace-based. Strategy calls, safety cars and incidents are not predicted.
- Model spread comes from historical outcomes for similar circuits, not a probability of the exact position.
- New circuits and regulation changes reduce the similarity weighting and widen the output range.
- Everything is computed offline and versioned. Nothing here updates live during a session.