F1 fan and data scientist Mariana Antaya built her own AI model to predict race winners. So far, it has correctly forecasted three Grand Prix results in 2025 – and it keeps improving.
From fandom to data lab
Like many fans, Mariana Antaya used to guess who would win each Grand Prix. But unlike most, she took it one step further: she built a machine learning model using real race data to predict outcomes.
Trained on lap times from the 2024 season, and using open FastF1 API data, her first success came with the 2025 Australian GP, where her model correctly predicted Lando Norris as the winner.
Crowdsourced machine learning
Mariana opened up the project to the community, letting fans suggest what features to include — from weather to free practice data.
The model responded well. At Suzuka, she added rain probabilities and driver wet-weather skills. Again, the result: a correct prediction of Max Verstappen’s win.
Smarter than it looks
The tool isn’t perfect. Mariana admits that safety cars and crashes are hard to predict. But her model is growing more accurate with every race, especially as she feeds it with team performance trends, like McLaren’s progress or Red Bull’s inconsistencies.
It’s not just about guessing winners — it’s about understanding how the sport works through the lens of data.
What’s next?
Mariana wants to implement more advanced machine learning methods, reduce the mean absolute error, and possibly factor in crash likelihoods from past races.
While her predictions won’t replace the thrill of watching live F1, they highlight the growing intersection of AI, data science, and sport fandom — a future that’s already here.