We built an F1 engine, not just an opinion
Before writing a single hot take, we built something to argue with: a full F1 analysis engine. Not a dashboard that re-skins someone else's API — a system that ingests timing data, models race outcomes, and tells you why it thinks what it thinks.
Here's what's running under the hood:
- Driver and season comparison — head-to-head across eras, normalised so the numbers actually mean something.
- Track dominance maps — GPS traces that show, corner by corner, where one driver is genuinely faster and where they're just braver.
- A Monte Carlo race predictor — thousands of simulated races factoring in tyre strategy, finishing order, weather, and the current FIA regulation constraints. It doesn't give you a winner; it gives you a distribution.
- Championship odds and an accuracy tracker — every prediction is logged and graded against what actually happened, so the model can be wrong in public and held to it.
Wrong in public, on purpose
The most honest thing a prediction model can do is keep score. Ours does. Every race weekend, the predictor commits to its numbers before the lights go out, and afterwards we grade them against reality — no quiet edits, no retroactive genius.
This past weekend was its first real test in the wild. From here, every Grand Prix is another data point in a very public track record. Some weekends the model will look sharp. Some weekends it'll end up in the marbles with everyone else. Both are worth writing about.
You will not be surprised to learn a lot of this is done with AI. So if you see a problem, say something.
What's coming
This is the introduction, so here's the promise. Expect:
- Race-weekend predictions — what the model says, and how it did.
- Behind-the-build posts — how the tools work, what broke, and what the data taught us.
- Off-the-line analysis — the overlooked corners of a race weekend, backed by numbers instead of vibes.
Pull up a chair. We're going off the racing line.