System Docs // v2.0

The Algorithm

First things first... This isn't that other Elo model. You might have seen "How I calculated an ELO rating for every F1 driver ever" on YouTube or Matthew Perron's work. Those are great. They are arguably more "statistically rigorous" using strict head-to-head teammate comparisons.

Our model does not do that. We adopt a different, slightly more abstract set of weighted performance factors. Different tools for different jobs!

Computation Engine

It's an iterative process. We update it after every Grand Prix. The core concept is the Race Performance Score.

1
Take an arbitrary race rating (e.g. 8.6)
2
Scale it by a multiplier (100)
3
Add a baseline (1000)
4
Compare against the pre-race Elo
5
Multiply the difference by a Weighting Factor (K)
6
Add that to the old Elo

Implied Ratings Translation

Abstract Elo numbers are hard to visualize. What does 1721 actually mean? Let's map it to something familiar... like the F1 25 game ratings (1-100). We assume a linear correlation between our Elo and the game's rating using a two-point linear interpolation ($y = mx + c$).

Upper Anchor (2025 Grid)

Max Verstappen

System Elo

1890

F1 25 Game

95

Lower Anchor (2025 Grid)

Lance Stroll

System Elo

1460

F1 25 Game

78

Derived Function: f(x) ≈ 0.03953x + 20.279
Telemetry Simulation

Case Study: Gasly (Bahrain)

Base Rating8.6
Perf. Score 1000+(8.6*100)1860
Pre-Race ELO1714
ELO Delta (1860-1714)146
Weighting (146 * 0.05) +7
Final ELO1721
Model Limitations

It is imperative to note that this is an estimation. It is not a precise reflection of the official F1 25 game logic. Assuming a perfect linear relationship based on two data points is a massive simplification. Treat the "Implied Rating" as an illustrative abstraction. It's a point of correlation, not a definitive prediction. But it works for us ;-)

Development Team

Free Formula 1 driver ratings and Elo statistics. Visualising F1 driver performance, teammate comparisons, and historical rankings. All inquiries regarding the ELO model, the specific dataset employed, or the calculation methodology are welcome.

Rahul RajaniShlok PatelHarrison LeeFred
Contact Admin