Race engineer's decision surface. Forecasts when tyres fall off, when undercut windows open, and which call a strategist would make in the next eight laps — derived from FastF1 telemetry, packaged into a cockpit a non-engineer can read in three seconds.
Real pit-wall flow isn't "recommendation up top, evidence below." A team principal won't trust a call they haven't already half-formed in their head from the data. The page reads top-down as: where am I (status strip) → what's the evidence (race pace, stints, qualifying read) → what's the call (the hero card at the bottom). The recommendation lands as confirmation, not ambush.
Each channel serves one purpose so the eye learns the mapping after one card. Hue separation matters most for the two drivers — they share the same panel and overlay each other on every chart.
Dark mode wins on saturation-heavy displays in dim rooms (Bloomberg, mission control). Light mode wins on sustained reading of dense prose — relevant for the recommendation "why" lines. The toggle persists via localStorage. Driver brand colors blend 45% toward black in light mode because the bright teal that pops on near-black washes out on near-white. The HARD-tyre swatch flips from FIA-white to dark gray on white backgrounds for the same reason.
The Decision Queue sorts by urgency. The #1 item gets the hero treatment — large title, dominant accent border, max contrast. The rest stack under "Next up" with smaller cards. Three equal-sized cards force the user to read all three; one hero delivers the answer in a single glance, with secondary calls available without competing for attention.
Most tools show you what happened on a lap. This one explains why a car–driver pairing is fast or slow — the corner families a setup rewards, where the lap time actually lives, and what to do about it. Built to be read in 30 seconds by someone who has to make a call.
This page is currently showing representative sample data for a Hamilton (Mercedes) vs Verstappen (Red Bull) style head-to-head — generated to demonstrate the product, not real timing data. Nothing here should be read as actual F1 telemetry yet.
To load the genuine article, run the included pipeline on your own machine (where the F1 timing API is reachable) and drop the result next to this file:
pip install fastf1 → python fetch_race_data.py --year 2021 --gp "Abu Dhabi Grand Prix" --a HAM --b VER
That writes race_data.json. Reload this page (served over http, or locally via
python -m http.server) and the badge flips to LIVE,
every panel re-renders from the real lap. Corner pace, the track delta map, and the fit profile
are computed directly from telemetry; the written verdict is a transparent heuristic over those
measured deltas — labelled as inference, never dressed up as fact.