Ferrari's FP3 1-2 at COTA: why Leclerc's medium-tyre lap is the number that matters

Timing screens at COTA only start telling the truth in the final 20 minutes of FP3. Before that came installation laps, aero-rake runs and procedural mileage. Then Lewis Hamilton set 1:34.893 on the soft compound, the only sub-1:35 lap of the session, with Charles Leclerc second on 1:35.127 and on the medium tyre. Ferrari locked out the top two, and most readings of that result skip the part that matters.

A word on data quality first. The session record carries no fuel loads, no engine modes, no upgrade specification and no long-run simulation. In 14 years of watching Formula 1, I file single-session reports like this in the lowest-information category. Everything below is form, not forecast.

Why COTA splits the field

Sector 1 rewards power, traction and aero efficiency: the launch, the uphill run to Turn 1, the high-speed esses. The second half punishes stiff, low cars over a bumpy surface. Teams lower and stiffen the floor to buy sector-1 speed and accept plank-wear risk. At the 2026 United States Grand Prix, both Hamilton and Leclerc were disqualified after inspection for exactly that trade. This time the trace shows up at Turns 9 and 10.

The 0.234-second gap is not a 0.234-second gap

The gap between Hamilton and Leclerc reads like an ordinary 1-2. The two cars were on different compounds. Soft-to-medium single-lap delta at COTA conventionally runs to several tenths; I have no confirmed data for this weekend's compound set and track temperature, so that is a convention-based estimate. A medium-tyre lap only 0.234 seconds off the best soft lap implies Ferrari held genuine margin. Leclerc's medium-tyre lap is the most technically significant number of the session, because it works as both a single-lap indicator and a race-strategy read. A team splitting compounds across two cars in FP3 has session capacity to spare. Ferrari also swept all three sessions, and spending FP3 on compound mapping rather than a pure qualifying simulation is the mark of a confident garage.

Session rhythm matters too. The opening 20 minutes produced no meaningful order; it was installation and aero measurement. All the strategic value sat in the last 20 minutes.

Red Bull is quickest where it is hardest

Red Bull led sector 1. Max Verstappen moved from 1:35.702 to 1:35.061. That profile fits a low-drag trim: strong on the traction-limited opening sector, comparatively weaker through the high-load direction changes. It is an efficiency trade, not a deficit.

The hole in the record changes the reading: the compound of the 1:35.061 is not stated. On softs, Red Bull trails by 0.17 seconds. On mediums, Red Bull is the pole favourite. One missing detail, two opposite conclusions. I log both scenarios before writing, because forcing a result into a single model is the mistake I have made most often.

Isack Hadjar's balance complaints in the second Red Bull point to divergent setups, likely one experimental car, so Verstappen's headline lap may not represent the optimised package. Verstappen also reported unsettled downshifts on team radio. That is a setup problem, not a concept problem: engine-braking mapping, differential entry settings, rear ride-height interaction with the surface, all tunable inside parc fermé limits.

The white line decides the order

Lando Norris lost a 1:35.2 at Turn 10. Andrea Kimi Antonelli lost roughly a 1:35.0 at Turn 9. Leclerc lost several laps, also at Turn 10. Three teams, two corners, four representative laps deleted.

The grey zone is not where the light fails. It is where the race is most real.

With the top four teams separated by two to three tenths, one track-limit excursion is worth more than any setup gain. This is circuit-level risk, not driver-level risk, which raises the probability of a scrambled grid. Leclerc owns both the best medium-tyre lap of the session and the worst track-limit record of it.

Here I have to argue against myself. The Mercedes half a step behind narrative contradicts the session data: Antonelli found a tenth on Leclerc's benchmark before his lap was deleted at Turn 9, a lap good enough for provisional P1. Mercedes' raw single-lap pace is closer than P4/P5 suggests. Same for McLaren: Norris's deleted 1:35.2 would have sat around the second row, and the MCL40 being described as difficult to tame implies a narrow setup and tyre-temperature window. Narrow-window cars lose more in qualifying, where there is one push lap and no time to build temperature, than in race conditions.

Pierre Gasly finished ninth at 1.000 seconds, which keeps the front-to-midfield stratification at roughly one second.

The contrarian angle

The easiest thing to write after FP3 is that Ferrari have found the formula. That is unproven. There is no long-run data, no race simulation, no fuel figures. A model built on 20 minutes of single-lap running carries an error bar wider than the gap between four teams.

Second: if the white line is the main variable, the team that wins qualifying at COTA may be the one willing to give away a little peak speed to keep laps legal. Risk management beats peak optimisation here.

And a warning: do not impose a collapse scenario on any team because they lost one good lap in FP3. I have done it, and the data punished me.

Four checkpoints before Q1

I will watch four things. The deletion count at Turns 9 and 10; if it persists, the grid will drift away from the pace order. Red Bull's sector-1 split; a continued lead there with time lost in the second half confirms the low-drag trim. Leclerc's first push lap into Turn 10, as a signal of whether he will trade a little speed for a legal lap. And the compound of Red Bull's fastest lap, which decides who the real pole candidate is.

My theorem does not predict the champion. It predicts who collapses first. Every new contract is a hypothesis, and qualifying is the experiment. I do not believe in titles. I believe in the system that produces them. After FP3 at COTA, Ferrari hold more data than anyone else; whether they hold more pace is still open.