What-If Lab
Explore how changes to driver attributes affect win probability predictions
How It Works
Select a 2024 race and driver, then adjust the sliders to see how changes to qualifying position, driver form, team form, and reliability affect win probability. All 8 models are available for comparison.
Why 2024 Races Only?
This project uses the official F1 API (Ergast/FastF1) which provides data through 2024 only. Even though 2025 races have occurred, the API has not released that data yet. What-If analysis requires actual race features which are only available for 2020-2024.
1. Select Race, Driver & Model
AUC: 98.3%Brier: 0.024XGBoost
2. Adjust Scenario
10 places better10 places worse
-50% (worse)+50% (better)
-50% (worse)+50% (better)
More reliableLess reliable
All 8 Models Available
LR
Logistic Regression
AUC: 98.7%
RF
Random Forest
AUC: 98.5%
CAT
CatBoost
AUC: 98.5%
XGB
XGBoost
AUC: 98.3%
QUALI_FREQ
Qualifying Frequency
AUC: 98.1%
LGBM
LightGBM
AUC: 97.5%
NBT_TLF
Neural Bradley-Terry
AUC: 95.0%
ELO
Elo Rating
AUC: 44.0%