Pit wall strategy lab
Change the call. Replay the race.
Choose a Grand Prix, edit a driver's pit strategy, and see whether the alternate call would gain time, lose time, or get trapped in traffic.
Analyze race
Load real FastF1 lap data.
Edit strategy
Pick a driver, pit laps, and tyres.
Compare result
See positions, gaps, traffic, and delta.
Load a Grand Prix
Race Control
No race selected
Waiting for the backend schedule
Pick a driver and strategy
Strategy Bay
Strategy tools locked
Analyze a race first, then choose a driver and edit the stint plan.
Positions, gaps, and explanation
Alternate Strategy Result
Awaiting pit wall call
Run a structured strategy or ask the optimizer to produce the first alternate-strategy overlay.
Model credibility
Confidence Gate
No model yet
Analyze a race to see fit quality, reconstruction error, and confidence reasons.
Can we trust this race?
Data Quality
No data report yet
Analyze a race to see coverage, warnings, and recommended simulator use.
Want to know how the backend got this result?
Model Lab
Selected engine
No engine selected
Model Lab shows both available simulator paths. Auto mode compares their validation scores and uses the engine that reconstructs the real race most reliably.
AI can parse strategy text and write narration, but the simulator owns every lap time, gap, position, confidence score, and optimizer choice.
Model version
--
Clean laps
--
Confidence
--
Baseline engine
OLS pace model plus ghost-field replay
Fits trusted clean laps with a regression, then replays one driver's alternate strategy against real-race ghost cars.
- Predicts lap time from driver pace, tyre compound, tyre age, fuel burn, and pit loss.
- Keeps the rest of the field anchored to the real race.
- Useful as the conservative fallback when advanced validation is not better.
Fit laps
--
Fit RMSE
--
Pit loss
--
Fuel burn
--
Passability
--
State
--
Advanced engine
Calibrated full-field simulation
Starts from baseline, then adds historical priors, nonlinear tyre phases, residual offsets, calibration, and full-field interaction.
- Models evolving gaps, traffic trains, pit events, and deterministic overtakes.
- Replays the actual race first to learn capped race-specific corrections.
- Only wins auto-selection when validation says it reconstructs the race better.
Historical blend
0%
Residual error
--
Field anchor
--
Overtakes
--
Traffic events
--
Calibration
--
Pit-loss fix
--
Validation comparison
Why auto trusts one engine
Final error checks finishing order, trace error checks lap-by-lap position shape, and the score combines those with bounded gap and event-plausibility penalties.
Quick terms
F1 Glossary
Pit stop
A tyre change. It costs time now, but fresher tyres may be faster later.
Compound
Tyre type: soft is quicker but wears faster; hard is slower but lasts longer.
Stint
The laps between pit stops on one set of tyres.
Dirty air
Time lost when following another car closely and unable to pass.