The last straight in Baku: 14 km/h with the pedal down
George Russell drove the final 900 metres of the race with the throttle pinned and lost speed all the way – 322 km/h became 308 at the line. Max Verstappen held 327. The gap was 0.196 seconds. Further back, P7 was decided by four hundredths on the same thing: who had electrical energy left when the straight ran out. Here is the last lap car by car – and how it came to that.
A 2026 car has 350 kW of electrical power to add to the engine, but only for as long as the battery lasts. Run out mid-straight and the car slows even though the driver has the pedal to the floor – that is what clipping means. We read the telemetry for all fifteen cars on lap 51: top speed on the final straight and speed across the line.
| Car | Finish | Top → line | Loss |
|---|---|---|---|
| Russell | P1 | 322 → 308 | −14 |
| Verstappen | +0.232 | 331 → 327 | −4 |
| Hadjar | +10.7 | 325 → 269 | −56 |
| Leclerc | +14.1 | 325 → 279 | −46 |
| Antonelli | +0.372 | 331 → 288 | −43 |
| Hamilton | +22.4 | 328 → 285 | −43 |
| Lindblad | P7 | 345 → 312 | −33 |
| Ocon | +0.046 | 327 → 321 | −6 |
| Bearman | +0.78 | 335 → 320 | −15 |
| Sainz | +0.49 | 342 → 332 | −10 |
Russell and Verstappen. Both peak about 900 metres from the line, Russell at 321–322, Verstappen at 331. Then the curves part: Russell loses a km/h every hundred metres at 100 per cent throttle, Verstappen holds 325–327 and only lifts in the final metres, alongside and realising it is not enough. At the line the difference is 19 km/h in speed and 0.196 s in time. Russell said afterwards that he lifted for the yellow after Bottas's crash and that the car then had "no turbo". The trace matches the description. Why the energy did not come back after the lift we cannot see – battery state is not in the data F1 publishes.
Lindblad and Ocon, 0.046 s. Lindblad was the fastest of anyone on the straight, 345 km/h, to keep the Haas behind – and was down to 312 at the line. Ocon crossed at 321 and closing. Thirty metres more of straight and P7 would have been Haas's. That is energy management in a sentence: Lindblad spent everything to defend, and it was enough by four hundredths.
Antonelli and Leclerc. Both empty over the final stretch. Antonelli 9 km/h faster at the line, 0.37 s late. Leclerc did what Russell did: got to the flag on what was left.
Hadjar lost 56 km/h from peak to line but had ten seconds in hand – Red Bull may simply have let him roll home.
Russell took pole by 0.837 s, the biggest margin of the year, and won the start. Then he pulled away: 5.2 s on lap 10, 8.6 on lap 20 and 11.2 s ahead of Oscar Piastri on lap 29, the last green lap before the Safety Car. The lead was built on pace, not tyre choice. Russell's median over laps 11–29 was 1:47.50 on the medium. Verstappen on the medium ran 1:47.80, and the three quickest soft starters – Leclerc, Piastri, Hadjar – 1:47.82 to 1:47.88. Nobody on the soft was quicker than Russell on the medium.
Behind Russell the race stood still. Piastri, Verstappen and Hadjar ran in a train from lap 4 to lap 29, about a second apart, and nobody got past Piastri. Red Bull let Verstappen through past Hadjar on lap 7 – Hadjar had been ahead since the start – but that only swapped two cars behind the same cork.
Alex Albon hit the barrier at Turn 6 on lap 30. Safety Car. Everyone at the front stopped on laps 30–31 at the cheap SC price; Red Bull double-stacked without a hitch. At the lap-35 restart Verstappen passed Piastri, while further back Franco Colapinto hit team mate Pierre Gasly, who collected Lando Norris – Safety Car again, laps 36–38, and another free stop for everyone.
The Safety Car did not decide the race – Russell led comfortably before it and held on after it. It made Friday's tyre question irrelevant. Everyone stopped on the same laps, nobody attempted an undercut, nobody ran a different plan. The pit window we had calculated never opened. We cannot even measure the green-flag pit loss in Baku 2026: only two green stops were made all race.
At the lap-36 SC stop, with fifteen laps to go, Mercedes and Red Bull fitted the soft (C5) while Hadjar, Leclerc, Hamilton and Piastri took the medium (C4). Median lap time in the final stint, laps 40–51:
| Driver | Tyre | Median |
|---|---|---|
| Verstappen | Soft | 1:45.14 |
| Russell | Soft | 1:45.16 |
| Antonelli | Soft | 1:45.80 |
| Hadjar | Medium | 1:45.85 |
| Leclerc | Medium | 1:46.04 |
| Hamilton | Medium | 1:46.48 |
Hadjar was 2.6 s behind Russell on lap 39 and 10.7 s behind at the flag – 0.7 s/lap. That is roughly the one-lap soft advantage we measured in practice (1.0–1.1 s over one lap, equal on long runs), stretched across a short stint on a light car. With thirteen laps left and no wear to speak of, there was no reason to take the medium. Ferrari did anyway – which is why Leclerc and Antonelli were fighting for P4 on empty batteries instead of Leclerc sitting safe.
Degradation. We wrote on Friday that neither soft nor medium showed measurable wear and that track evolution beat degradation. The race confirms it: the fuel-corrected within-stint slope is 0.00 s/lap for both compounds, and Russell did 31 laps on a used medium without losing pace.
Equal on long runs – half right. Soft and medium starters ran within 0.4 s/lap of each other in the first stint, and the quickest car was on the medium. That matches the practice data (1:50.03 vs 1:50.04). But that comparison mixes tyres with drivers: the quickest car was Russell's. Our separate race model (see What we cannot say below) takes out each driver's own pace and the common fuel and track effect, and puts the soft 0.31 s/lap quicker than the medium (range 0.22–0.43). So the two compounds were equal for the cars as they were lined up, not for the tyre itself.
Antonelli. From P16 our three plans gave P9, P6 and P3. The result was P5 – with two free stops and three cars falling away in front. We wrote that a Safety Car rewrites all three and that Antonelli was the driver with the least to lose by waiting for one.
The starting tyre. The simulation chose the soft for Russell because the soft was a second quicker over one lap and equal on long runs – so take the second at the start. Mercedes reasoned the other way round: on a track where nobody could overtake (our threshold 0.88 s/lap) and where the soft was no quicker in race pace, there was nothing to gain from a soft start, but everything to gain from having new softs left for the end. Russell and Antonelli started on three-lap-old mediums and had new softs for what became a fifteen-lap sprint. We counted the tyre advantage lap by lap and did not count which sets would be left when something happened. We will.
"The race will be decided on the straight and in the pit window." Half right. It was decided on the straight – by the battery, not the wing – and the pit window never came.
Our standard What-if simulator gives no answers for Baku, so we publish no ordinary scenarios. Everyone ran exactly two stints in which compound and race phase coincide: whoever started on the medium finished on the soft and vice versa. The standard model then cannot separate the tyre's contribution from fuel mass, track evolution and the free pushing after a restart.
We have since built a separate model for that problem. It estimates each driver's base pace, a common lap-by-lap effect for fuel and track, and the tyre difference, all at once. On that model the soft was 0.31 s/lap quicker than the medium (range 0.22–0.43), not the 0.7 between Russell and Hadjar in the final stint. Most of the rest was the drivers' own pace: in this model Russell was 0.6 s/lap quicker than Hadjar on equal tyres. We use it for one question only.
FIA and Pirelli are discussing softer tyres that wear faster, to force a second pit stop. We asked the opposite: what if the mandatory stop was scrapped and the tyre rules otherwise left alone? We re-ran Baku with the two who fought for the win.
The assumptions. Tyre wear over a full race cannot be measured here: nobody ran more than 34 laps on one set. So we test a range: - Medium: loses 0.012 or 0.02 s per lap of age and lasts all 51 laps. 0.012 is what this year's other smooth tracks showed (Montréal, Madrid, Monza). Within Saturday's stints we measured no wear at all, so 0.012 is on the cautious side. - Soft: loses 0.02, 0.03 or 0.04 s/lap. It wore 0.047 in Madrid and 0.050 at Monza, almost nothing in Montréal.
A green-flag stop costs 25.8 s. To pass in Baku a car needs to be 0.88 s/lap quicker, otherwise it stays stuck behind.
Test 1: no Safety Car, one medium against two softs. We ran it both ways round. First Verstappen does 51 laps on one medium while Russell does two soft stints, then they swap. Each time the soft runner stops on the best lap. Swapping the drivers separates the strategy from the driver: Russell's base pace was worth 13–15 seconds over the race. What is left is the value of the strategy itself:
| Wear medium / soft (s/lap) | Faster strategy |
|---|---|
| 0.012 / 0.02 | No stop, by 9 s |
| 0.012 / 0.03 | No stop, by 15 s |
| 0.012 / 0.04 | No stop, by 21 s |
| 0.02 / 0.02 | Two softs, by 3 s |
| 0.02 / 0.03 | No stop, by 3 s |
| 0.02 / 0.04 | No stop, by 12 s |
On Baku's smooth surface the no-stopper wins five scenarios out of six. Two softs pay only if the medium wears as fast as the soft. In the actual pairing Russell wins both ways round, apart from two cases. When Russell is on the softs and they wear 0.03 or 0.04 (medium 0.012), Verstappen wins by 0.6 and 5.7 s.
Test 2: Saturday's race, Safety Cars included. The Safety Car and VSC periods run laps 30–38 as they did on Saturday. Russell starts on the medium and never stops, not even when the whole field pits. Verstappen starts on the soft and changes to new softs once under the Safety Car.
| Soft wear (medium 0.012) | Verstappen stops lap 31 | Verstappen stops lap 36 |
|---|---|---|
| 0.02 s/lap | Russell by 1.5 s | Verstappen by 1.8 s |
| 0.03 s/lap | Russell by 1.4 s | Verstappen by 1.2 s |
| 0.04 s/lap | Russell by 1.6 s | Verstappen by 1.9 s |
Russell inherits the lead when everyone else pits, then sits on a medium more than 40 laps old with Verstappen right behind on fresh tyres. It all comes down to the restart: - Stop on lap 36: the softs are new enough for the pace gap to reach Baku's overtaking threshold, and Verstappen gets past. - Stop on lap 31: they are five laps older and Verstappen stays stuck a second behind to the flag.
With the medium wearing 0.02, the lap-36 stop wins by 6 s. Within our margin of error it is a coin flip.
The tyres. On Saturday Russell and Verstappen used three sets each. In every test here the no-stopper uses one set and the one-stopper two: three sets for the top two instead of six. More strategic choice, fewer tyres.
The caveats. - Assumed wear: The results hold only if the tyres behave as assumed. Nobody ran anywhere near 51 laps on one set in Baku. - Margin of error: Replaying the real race, the model finishes about 1.8 s off the real result, so differences under 2 s are a draw. - Soft starts: Both drivers actually started on the medium. For a soft start we only add the soft's normal pace gain on lap 1, nothing extra for the launch itself. - Heavier car: Soft wear on full tanks may be higher early on than we assume. - Two cars only: Test 1 has just Russell and Verstappen on track, so no traffic from the rest of the field. - Verstappen's pace: Verstappen spent much of the race in dirty air, so his pace may be underestimated. At 0.2 s/lap quicker than our estimate, Verstappen wins five of the six cases where Verstappen runs the medium and takes Test 2 by 6–13 s with the lap-36 stop.
How we did it: Lap data and car telemetry from FastF1 (F1 Live Timing). Speed on the final straight measured backwards from the finish, throttle from the same channel; battery state is not in the published data. Medians on green laps excluding in- and out-laps. Degradation per compound from stints of at least six laps, fuel-corrected at 0.035 s/lap. Pirelli nomination Baku 2026: C3 hard, C4 medium, C5 soft. Sources: F1.com (race report, 26 September 2026, Russell's quotes), FIA timing, Pirelli (Simone Berra).
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.