What Actually Wins You F1 Points?

Inspiration

We love Formula 1 — the speed, the strategy, the drama. But one question always came up in the stands and group chats:

Is the race actually decided on Sunday, or was it over on Saturday?

That question became our project.

How We Built It

We used Google Colab to write the code and GitHub to collaborate.

Collected 26,692 laps of real 2025 F1 season data
Engineered 5 features: grid position, qualifying pace, recent form, team strength, and tyre compound
Trained two models — a Random Forest and a Neural Network
Used walk-forward validation so each race was predicted using only data from previous races — no cheating
Built a fully interactive dashboard with charts, driver cards, and a race-by-race breakdown

What We Found

We let the models learn from the data without telling them what mattered. Both independently agreed:

73% of your finishing position is decided before lights out.

Grid position alone explained more than half the outcome. Qualifying — not strategy, not tyres, not overtakes — is the real race.

The Random Forest predicted finishing positions within ±2 places 40% of the time. The Neural Network hit 33%. For a sport as unpredictable as F1, that's a meaningful result.

$$\text{What wins points} \approx \underbrace{55.8\%\ \text{grid} + 17.4\%\ \text{quali}}{\text{Saturday (73\%)}} + \underbrace{13.7\%\ \text{form} + 9.6\%\ \text{team} + 3.4\%\ \text{tyre}}{\text{Sunday (27\%)}}$$

Challenges We Faced

Handling retirements, missing laps, and messy real-world race data
Making sure the model never "saw the future" during training
Stopping the Neural Network from memorising instead of learning
Debugging JavaScript buried inside a Python-generated HTML file (one misplaced semicolon cost us an hour)
Learning to collaborate on GitHub under hackathon time pressure