The race, called in advance
Loading race…
Predicted finishing order
Every driver, ranked by the model, with their chance of winning and of finishing on the podium.
- Loading predictions…
Every driver, rated
Each card turns a driver's season into six ratings from 40 to 99, built from the same race results the model learns from. Open a card to see where the numbers come from.
The same cards for whole careers: every Grand Prix since each driver's debut, rated against today's grid. Open a card for career totals and teams.
How the ratings work
- QUA
- Qualifying: average qualifying position.
- PAC
- Pace: average finishing position in races the driver finished.
- RAC
- Racecraft: places gained or kept from the grid, as a share of what was possible.
- CON
- Consistency: how often the driver finishes, and how little the results jump around.
- FRM
- Form: points per race over the last five Grands Prix.
- H2H
- Vs teammate: share of qualifying and race head-to-heads won against the driver in the same car.
In the Career view, positions count as a share of the field (grids have had 18 to 26 cars), Form becomes SUC, success: win rate, podium rate and championships, and drivers with few starts are pulled towards the average, so a rookie's handful of races can't top the table. Our data starts in 2001.
Each stat is rescaled so the best driver on the grid gets about 97 and the slowest about 40. The overall rating weights pace and qualifying highest. Because two stats compare drivers with their teammates, a great driver in a slow car still rates well. Gold is 80 and up, silver 70 to 79, bronze below 70. Drivers with fewer than three starts this season have no card.
Teams & standings
Live championship tables, every Grand Prix finish this season, and each team's car in the garage.
2026 scorecard
Each race was predicted using only data from before it, and compared with the simplest rule: the driver on pole wins.
How accurate is it?
Back-tested race by race over three seasons. The baseline predicts that drivers finish in starting-grid order.
Can you trust the percentages?
Drivers grouped by predicted win chance. Dots near the dashed line mean the odds are honest.
What drives the prediction
Share of the ranking model's decisions explained by each input.
How it works
A four-stage pipeline that reruns by itself after every practice session, qualifying and race.
Every result since 2018
Race and qualifying results for every Grand Prix from the Jolpica F1 database, plus practice lap times from OpenF1.
20 signals per driver
Grid slot, gap to pole, recent form, team pace, retirement rate, track history and championship standing.
Ranking + odds
A LightGBM ranking model sets the order. Logistic models give win and podium odds. Recent seasons weigh more.
No peeking
Every past race is predicted using only earlier races, then scored against the grid-order baseline.