The 37 seconds that cost Norris the win
It took Race Control 37 seconds from yellow flag to VSC after Lance Stroll's brake failure. Ten seconds quicker and every driver would have had the same chance of a cheap pit stop. Instead, race leader Lando Norris passed the pit entry 2.4 seconds before the boards lit up – and, by our numbers, lost the race right there.

The timing data leaves no room for doubt. The yellow flag in sectors 22–23 came 22:36 into the race. The VSC was called at 23:13.7. Norris crossed the line to complete lap 14 at 23:16.1 – he was already past the point where you commit to the pit lane. Antonelli, 6.2 seconds behind on lap 13, made it with nothing to spare: pit-in 11.7 seconds after the signal. Verstappen and Russell followed.
The VSC ended at 25:17 and the track went green at 25:27. Norris turned in at 25:41 – fourteen seconds into green – and paid the full pit loss, plus a slow stop when the front-right wheel stuck. From the lead to fifth; third at the flag once Russell and Leclerc made their stops.
F1.com's post-race piece says Norris had "just passed pit exit to start lap 14" when the VSC was called. The timing says otherwise: it was the end of lap 14, not the start, and it was the pit entry he had passed. The conclusion is the same – he had no choice – but the margin was 2.4 seconds, not a whole lap.

Matt Youson on F1.com wrote on Sunday morning that medium–hard with one stop on laps 18–24 was the main play, that medium–hard–hard (10–16 and 31–37) gave almost the same race time, and that hardly anyone had run the hard tyre in practice. He also flagged that the Madring had "potential for mayhem" with Safety Cars and VSCs.
All of that held, except one thing: the car did not need to stop twice. The nomination was C2 as hard, C3 as medium and C4 as soft. The hard (C2) turned out to be close to indestructible. Leclerc did 47 laps on his first set; Antonelli, Verstappen and Norris 42 each – and the fuel-corrected slope in our data is −0.02 to +0.02 s/lap, which is to say zero degradation. The medium on low fuel, by contrast, went backwards at 0.09 s/lap, and Russell lost 0.1–0.2 s/lap to Leclerc's fourteen-lap-older hards before Mercedes gave up and stopped him again on lap 28. The soft late in the race: +0.2–0.4 s/lap. Exactly the picture Pirelli's Simone Berra described: front-left graining, medium and soft overheating, the hard holding up.

"There was no option really to retain the lead. We needed to accept that the timing was incredibly unfortunate." – Andrea Stella
The only alternative McLaren had was to leave Norris out on his mediums. Pirelli estimated they had ten competitive laps left. We let the simulator test it: Norris stays out until lap 25, everyone else drives exactly as they did. Click a scenario and it loads in Simulate below.
| Scenario | Actual | Simulated | Race time | Comment |
|---|---|---|---|---|
| NOR stays out until lap 25 | P3 | P3 | +0.5 s | Four laps stuck in traffic after the stop. Stella was right. |
| NOR makes it in under the VSC (lap 14) | P3 | P1 | −20.4 s | Keeps the lead with six seconds over Antonelli. |
| LEC stops on lap 40 instead of 48 | P4 | P4 | −4.1 s | Quicker, but the cars ahead are out of reach. |
| RUS skips the VSC stop, one stop on lap 35 | P5 | P5 | +8.5 s | A one-stop on the hard, Leclerc-style, would not have helped Russell. |
Two things fall out. First: Stroll's brake failure was worth a little over twenty seconds to Norris – the whole difference between winning and the podium. Second: once he had missed the VSC, no strategy was worth more than half a second either way. On a circuit where our overtaking threshold sits at 0.8 s/lap and the gaps ahead of him were 10–15 seconds, track position was everything.

The post-race piece says drivers saved "around 11 seconds" by pitting under the VSC. Our measured pit loss under VSC in Madrid is 11.2 s against 24.2 s under green, a 13-second discount – the same order of magnitude. But the formula the library used before the race, based on the field's pace ratio under the flag, gave only a five-second discount. The reason: the ratio was diluted by laps that were only partly run under VSC. That is the single most important lesson from Madrid, because it is exactly the number the reactive-rival model looks at when deciding whether a car should dive in under a VSC. From the next race on, the library uses the measured value whenever there are at least five stops to measure.
The second lesson: tyre life cannot be derived from degradation. The hard had zero degradation and lasted 45+ laps; the medium had almost as little degradation in the first stint but was finished after 25 because of graining. Tyre life therefore becomes its own parameter per compound in the library.

For the first hour of the race we followed the gaps and the undercut cards live. They matched reality – Antonelli's threat to Norris was visible, and the rejoin positions after the VSC stops were right. Then the stream dropped after exactly one hour, a recorder timeout that has now been raised to four. The full race can be replayed in the Pit wall, with the flags at their exact times.
How we did it: lap data from FastF1 (F1 Live Timing). Degradation per compound from green laps without in-/out-laps, fuel-corrected at 0.035 s/lap, stints of at least eight laps. The simulator runs every other driver on their actual lap times and models the changed driver on his own base pace per compound from the race; cars that get stuck behind him lose time under the same traffic rule as the field simulation. Sources: F1.com (Matt Youson's pre-race strategy piece and the post-race analysis of Norris and the VSC), Pirelli (Simone Berra, Dario Marrafuschi), McLaren (Andrea Stella, post-race press conference).
Simulate
Change one driver's pit stops and let everyone else run exactly as they did. The altered driver is modelled with his own pace per compound from the race, degradation, fuel, traffic and the VSC discount – cars that get stuck behind him lose time.
