Formula 1
Analytics Platform
ML-driven race outcome prediction · Monte Carlo simulation engine
Select a race above to generate predictions
Grid slot — single strongest predictor of race outcome
Points-weighted average of last 5 race finishes
Driver & team historical performance at this circuit
Constructor gap vs field average over last 3 events
DNF / mechanical failure rate across last 2 seasons
Historical compound degradation vs field baseline
2026 uprated MGU-K usage pattern from prior events
Temperature, humidity, chance of safety car
Optimal stop window from Monte Carlo strategy sim
High historical accuracy — strong predictive signal
Solid prediction with moderate variance
Competitive field — prediction less certain
Highly contested — multiple realistic outcomes
How It Works
The core model is a Gradient Boosting Regressor (100 estimators, learning rate 0.1) trained on 2024–2025 F1 season data via FastF1. It predicts lap time per driver; drivers are ranked by predicted time to produce finishing order probabilities.
Qualifying position carries the highest feature weight (28%) as the strongest single predictor — track position in modern F1 is decisive. The model achieved a mean absolute lap-time error of 3.22 seconds and top-3 podium accuracy of 89.5% across the 2024–25 seasons.
For 2026 predictions, the model augments its historical baseline with live 2026 standings data — current championship points gaps are mapped to form adjustments on the Recent Form and Team Pace features, accounting for momentum changes from the season so far.
Monte Carlo simulation runs 10,000 race scenarios sampling from historical degradation distributions, reliability failure rates, and safety car probabilities to produce a statistically robust finishing position distribution per driver.
Data: Jolpica F1 API · FastF1 telemetry · Pirelli compound statistics · 2024–2026 race history