How PITWALL records race data, analyses strategy and runs its simulations
This page explains where PITWALL's data comes from, how the race analysis on each race page is calculated, and how the simulator produces its numbers. Throughout the site, recorded race data and PITWALL's own model output are kept apart: anything produced by the simulator is labelled as PITWALL model output and is never presented as an official result.
For every completed Grand Prix PITWALL stores the finishing classification, the winner, the fastest lap, the weather, and the key moments of the race with the lap they happened on — including Safety Car, Virtual Safety Car and red-flag periods. The qualifying result and the starting grid after penalties are stored as well.
Each driver's strategy is recorded as the starting tyre plus every tyre change with the lap it was made on. A change made while the race was red-flagged is marked separately: it is a free change, not a pit stop, and it is never counted as one.
When PITWALL fills in tyre strategies, pit laps come from the official Pit Stop Summary published on formula1.com and compounds from Pirelli's per-driver tyre stint lists. When the two sources disagree for a driver, that driver's strategy is left unrecorded rather than guessed, and the race page states how many drivers its analysis is based on.
Each circuit has its lap length, race distance, pit lane length and speed limit, and the Pirelli compound allocation for its 2026 event. Base lap times are derived from the circuit's real fastest lap, scaled by 1.02 to approximate race pace. Values that could not be sourced — for example at circuits without a recent Formula 1 race — are marked as estimates in PITWALL's circuit data.
Everything in the analysis sections of a race page is calculated only from the recorded data above; nothing in them is simulated.
Coverage rule. Tyre strategy, pit stop and strategy-mix sections appear only when at least 50% of classified finishers have a recorded strategy. Below that, a summary would mostly describe gaps in the data, so the section is left out.
Stops after a Safety Car or VSC. The race data records when a neutralisation was called but not when it ended, so a stop is listed as "after" a Safety Car or VSC when it was made within 2 laps of the call. The page states this rule next to the list.
Why they lost. Factors are reported only when they clear fixed thresholds: an extra pit stop compared with the winner, a first stop 5 or more laps earlier or later than the winner's, finishing 5 or more places behind the starting grid slot, a retirement, or setting the fastest lap without converting it into a podium.
PITWALL's simulator runs a race lap by lap, with each lap split into circuit segments such as straights, hairpins and chicanes. The main inputs are:
For the Daily Grand Prix and the stored strategy comparisons, random elements such as small lap-time variation and overtaking outcomes use a seeded random generator, so the same inputs always reproduce the same race.
The tyre, fuel and pit-stop parameters are fitted to lap-by-lap timing of the 2026 dry races, taken from OpenF1 — an unofficial service built on Formula 1's live timing feed. Laps behind a Safety Car or VSC, pit in- and out-laps and traffic-affected laps are removed first. The data shows that 2026 tyres lose time slowly and roughly linearly, with no sudden "cliff" within the stint lengths teams actually ran, so the model uses linear wear with softer compounds always wearing at least as fast as harder ones. Wear rates are measured for 14 circuits and pit-stop time loss for 10; other circuits use pre-2026 estimates. Beyond the stint lengths observed, the model extrapolates.
Each circuit guide includes a strategy profile computed with the same model for a single car in clean air: how a one-stop compares with a two-stop, and the one-stop pit window. Because this setting has no traffic, track position or Safety Car, the profile gives stop count and timing but no compound order — in 2026 the model prefers finishing on softs, while teams mostly ran medium to hard. Across the 2026 dry races, the profile's one-stop/two-stop verdict matches what most of the field actually did at 12 of 14 circuits.
Some race pages compare the winner's actual strategy with alternatives. These results are computed offline, stored, and shown with a "PITWALL model output" label.
The race is rebuilt from recorded data: every classified finisher starts from their real grid slot on the strategy they actually used, and Safety Car and VSC periods start on the laps they were called. Retired cars are left out, because the simulator cannot make a car retire on a given lap. Each strategy is run 40 times.
Only the winner's strategy is changed; every rival keeps their real one. PITWALL tests every one-stop and two-stop plan that uses at least two dry compounds. One-stop plans are scanned every 2 laps, with each stint at least 20% of the race distance; two-stop plans are scanned within 3 laps either side of the thirds of the race. The best 3 one-stop and best 3 two-stop plans are then re-run at full sample size. Every candidate uses the same list of seeds, so differences between strategies are not caused by different luck.
Each comparison shows a model check: with everyone on their real strategy, how many places the simulated finishing positions were away from the real result on average, and how often the real winner won the simulated races. A large error means the comparison should be read with more caution.
A race is not simulated when the replay could not be faithful: a wet or mixed race, a race stopped by a red flag, a classified driver with no recorded strategy or grid slot, or fewer than 10 finishers. If a race's data is corrected after its simulation was computed, the old result is hidden until it is recomputed. The current model version is replay-2. How closely the model reproduces each real 2026 race is published, race by race, on the simulator accuracy page.
The takeaways at the end of a race page start as drafts generated from that race's own data and simulation — never from other sites' reporting — and are reviewed and edited by hand before they are published.
PITWALL is an independent fan project. The simulator is an analytical model and does not reproduce official Formula 1 team simulations, telemetry or proprietary data. Its results are estimates that can differ from what happened on track.
Spotted a data error? Corrections are welcome through the contact page. More about the project is on the About page.