The Rubber Revolution: Tire Choice Modeling
This model analyzes historical weather and environmental data to predict the optimal tire compound choice, forming the foundation of any successful race strategy.
Formula 1 AI PitWall
CS 7641 - Summer 25 - Group 4
Introduction
Formula 1 isn't just the world's most popular motorsport; it's a high-speed chess match played out at 200 mph. Millions watch, and fortunes are won and lost on the razor's edge of performance. To find that edge, teams and fans alike turn to data. Our project dives deep into this world, leveraging a suite of powerful machine learning tools. We've integrated data from FastF1 for unparalleled access to timing and telemetry, the TUMFTM Racetrack Database for intricate track details, and Tomo Bacinger’s F1 Circuit Data for precise geographical insights. This fusion of data is the fuel for our predictive engine, aiming to decode the DNA of a race weekend.
Our Motivation
Formula 1 is a symphony of driver skill, engineering genius, and strategic cunning. With countless interdependent variables, it presents the perfect challenge for machine learning. Our motivation was to build a tool that serves three core missions:
Race Strategy Simulation
From the pit wall, every decision is critical. When to pit? Which tire compound to choose? How hard to push? Our framework serves as a powerful race simulator, running thousands of scenarios to help identify the optimal strategy before the lights go out.
Advanced Sporting Analytics
F1 operates in a unique small-data, high-complexity environment. With non-stationary patterns driven by rule changes and car upgrades, it's a formidable research challenge. We've developed models to tackle time-series forecasting and multimodal feature integration in this dynamic arena.
For the F1 Community
Public predictions are often simplistic. We aim to change that. Our goal is to deliver accurate, data-driven, and interpretable predictions to the global F1 audience, elevating the conversation and empowering fans with real analytical firepower.
The Pit-Stop Dashboard
Our project is built on 5 distinct analytical models, each designed to deconstruct a key component of a Formula 1 race weekend. Click on any module below to jump directly to its performance analysis and results.
Chasing Pole Position: Qualifying Pace Predictor
Focusing on the crucial battle for grid position, this model uses a range of performance metrics to forecast the ultimate single-lap pace of each driver.
Decoding the Circuits: Track Analysis
Using unsupervised clustering, this module categorizes every F1 circuit into distinct archetypes based on its geometric characteristics, such as cornering intensity and high-speed sections.
The Undercut: Tire Stint & Pitstop Predictor
This model predicts the viable length of a tire stint by synthesizing track characteristics, weather conditions, and tire compound data to inform pit stop strategy.
Driver DNA: Decoding Driver Performance
Analyzing driver telemetry data to understand driving styles and breaking down performance attributes. Furthers understanding of circuit layouts to discover the unsaid but ingrained rules of being an F1 driver.
