RaceIQ F2
Formula 2 · open-wheel
A purpose-built spec-series model: a leakage-safe driver-skill blend (rookie-aware ratings and finishing history) routed through two race-type heads — a merit feature race and a reverse-grid sprint — with simulated win/podium probabilities, finishing-range intervals, a sprint+feature-aware championship projection, forward-eval, and drift/promotion monitoring. Full website with per-round sprint/feature detail, head-to-head matrices, and an accuracy dashboard. Runs on the real 2026 FIA Formula 2 season — official results are scraped into a committed, reviewable snapshot with model forecasts for upcoming rounds, refreshed on a schedule. ~75% reuses MotorsportVerse core.
Datasets
- fiaformula2.com (official results, scraped to a committed snapshot)
- synthetic latent-pace (offline fallback)
Model components
- Driver-skill model
- Reverse-grid sprint head
- Probability calibration
- Championship simulation
- Standings projection
Shared core modules
- Probability calibration
- Championship simulation
- Standings projection
- Driver-skill ratings
- Uncertainty intervals
- Forward evaluation
- Drift monitoring
- Gated model promotion
- Leakage guards
- Predictor interfaces
Tags
- feeder series
- spec series
- reverse grid
- calibrated probabilities
- race simulation
Registered 2026-06-15
Related projects
View all →
F1 Predictions
Formula 1
Calibrated race-outcome forecasts for every Formula 1 Grand Prix.
IndyCar Predictions
IndyCar
Race and championship forecasts for the NTT IndyCar Series.
RaceIQ F3
Formula 3
Sprint, feature-race, and championship forecasts for FIA Formula 3.