SRH Berlin University of Applied Sciences

In this project, I designed an ML model that predicts the championship probability for each driver after every round using historical F1 standings data.

What I did:

Built a full pipeline in Python (pandas + scikit-learn). Used races.csv, driver_standings.csv, and drivers.csv to create a round-by-round dataset, Engineered features like points/wins, gap to leader, and season progress (round_frac)

Trained an interpretable Logistic Regression model with scaling + imputation + class_weight="balanced" to handle class imbalance. Evaluated using forecasting-style metrics (Top1-by-round, Top1-final-round) and imbalance-aware metrics (Balanced Accuracy).

Next steps: add richer signals (race results, constructors, qualifying, reliability) to improve early-season forecasting.