Every Formula 1 race weekend, our AI generates a full slate of predictions: the race winner, the complete podium, qualifying pole position, sprint results, and driver head-to-head matchups. The model works from the same evidence a paddock analyst would — championship standings, circuit history, each driver's recent form, prior session results from the same weekend, and live news gathered through real-time web search in the days before the race.

What makes this different from a pundit's hot take is discipline: every prediction is locked before the session starts and never edited afterwards. When the chequered flag falls, the prediction is scored against the official classification — winner called or missed, podium order compared position by position, head-to-heads settled. The hits and the misses stay on the page permanently, on every race, all season.

That track record is the whole point. Formula 1 is genuinely hard to predict — a random pick from the 22-driver grid wins about 5% of the time, and even naive favourite-picking breaks down at chaotic weekends. Judge the model on the public scoreboard: every race page shows exactly what was predicted and what actually happened, and the accuracy tracker aggregates the full season.

Confidence percentages next to each pick show how strongly the model leans. A 70% winner call is a strong conviction; three drivers clustered at 25-30% means the model genuinely sees a toss-up. Treat them the way you'd treat a good analyst's gut feel, backed by data — not a guarantee.

Predictions are published for every Grand Prix in the season calendar below. Race pages update through the weekend as qualifying and sprint results come in, and switch to a full prediction review once the race is done.

See the full accuracy track record Read the methodology