Method

How it works

Every prediction on this site follows the same four steps, is locked before the session it forecasts, and is scored in public afterward. Here is exactly how — and where the honest limits are.

The pipeline

Gather

Before a session, we pull the current picture from public F1 data: recent form, qualifying and race history at the venue, championship state, reliability record, and the latest weather forecast for the circuit.

Frame

That raw data is compressed into a structured context pack — a driver-by-driver, team-by-team snapshot of who is fast where, who is fragile, and what the track and conditions reward.

Reason

A language model reasons over the context pack and produces a full forecast: a probabilistic finishing order, head-to-head calls, weather and incident estimates, and short written explanations for the headline picks.

Lock

The forecast is published with a timestamp and then frozen at session start. Locked predictions are never edited. If we ever regenerate, it becomes a new, separately labeled version — the original stays on the record.

The credibility mechanic

The point of this site is not to sound confident — it is to be accountable. Predictions are published before a session, frozen at its start, and graded after the official classification (including any penalties) is final. Nothing is quietly edited to look smarter in hindsight. The scoreboard shows the hits and the misses.

How we score

Winner: Did the driver we put P1 actually win? A simple hit-or-miss.
Podium: How close was our top three? Exact positions score full, right-driver-wrong-step scores partial; a set is a 'hit' above a threshold.
Full order: Rank correlation between our predicted grid-to-flag order and the official result, shown as a single percentage.
Props & head-to-heads: Each yes/no call and each teammate battle is checked against what happened, and we publish the strike rate.

What a model cannot know

We would rather be honest than impressive. Motor racing is full of genuine unpredictability, and no forecast should pretend otherwise:

Mechanical failures and first-lap incidents are close to random and can undo the best-reasoned call.
Safety-car and virtual-safety-car timing swings strategy in ways no pre-race model can pin down.
Weather that shifts during a session — a passing shower, a drying line — can invert an order.
Team orders and in-race politics are human decisions we can only estimate.

These forecasts are for information and entertainment only. They are not betting advice, and this site is not affiliated with Formula 1, the FIA, Formula One Management, or any team.