Antonelli tops FP3 in Madrid, but a 0.1-second gap is not yet proof Mercedes is fastest

The clock in the Madrid control room was ticking into the final minute of FP3 when Oliver Bearman lost his Haas at Turn 10. The car hit the barrier, the session was stopped for the second time that afternoon, and the timing screen froze at precisely the moment it was least representative. At the top: Andrea Kimi Antonelli, 1m 32.797s, set with five minutes left. Behind him, Charles Leclerc, 0.1s back. Then Oscar Piastri third, Lando Norris fourth, Max Verstappen fifth, George Russell sixth. An order neat enough to make me stop. I have sat through enough Saturday afternoons at enough circuits to know that when everything looks too tidy — every team in its expected place — the problem usually lies not in the numbers but in how the session was sliced up to produce them. Today it was sliced heavily. This weekend is the 2026 Spanish Grand Prix, at an entirely new venue. Madrid has taken over from Barcelona after many years, which means nobody on the grid, not even engineers with twenty years in the job, has real reference data. Simulation models are built from satellite imagery, from synthetic data, from shakedown laps. On a track without rubber, with corners that exist only on paper, the order is decided far more by who dares to run and who has enough track time than by the car itself. The broader setting: this is the first season of a new technical cycle. Cadillac is on the grid with two experienced drivers, Sergio Perez and Valtteri Bottas. Arvid Lindblad wears Racing Bulls colours. A regulation reset always reshuffles the pack — history is clear on that — but reshuffling can only be measured with track data, and we do not have it. FP3 produced two red flags. The first for Lewis Hamilton, into the barrier at the final corner. The second for Bearman. Together, roughly thirty minutes of running were lost at a circuit where track time is the most expensive asset there is. The session restarted with fifteen minutes left. That means most teams had exactly one clean soft-tyre opportunity. The teams also made life hard for themselves. Nobody went out for the first five minutes. The front-runners waited longer still — Mercedes and Ferrari both kept their cars in the garage waiting for the track to come in. Sound in principle. But when two red flags arrive, that patience becomes a data debt. Now the numbers, read the way I read injury files: looking for the gaps. An injury record does not lie — only the reader knows how to hide the truth. Antonelli: 1m 32.797s. Leclerc: 0.1s back. Norris: within a tenth. Piastri: immediately ahead. Four cars from three different teams inside two tenths. On a new circuit, on virgin asphalt, the noise band of a free practice session is wider than two tenths. The gap among the top four has not cleared the margin of error. The conclusion that Mercedes is fastest has no basis yet — only the narrower conclusion that Mercedes produced the best single lap in the best conditions. The most important detail the timing screen omits: Antonelli's lap came with five minutes left, when the track had absorbed the most rubber after an hour of fragmented running and a restart. Run latest, gain most, in principle. Leclerc went under 1m 33s the moment the session resumed — same logic. That is standard scheduling, not a discovery. But it means the final order reflects run plans more than car potential. Then Lando Norris. The McLaren driver lost almost all of FP2 to a gearbox problem and completed just three flying laps in FP3. Three. He still sits within a tenth of the fastest man. If I have to pick a single positive signal from the afternoon, I pick this data point — not the fastest lap. A car reaching peak pace on its first laps at an unfamiliar circuit points to a stable aerodynamic platform. But the sample is tiny, and the gearbox fault is a far more worrying technical note than fourth place. Max Verstappen is fifth, and I do not read that as a sign of weakness. He aborted his first soft-tyre flying lap after cutting the chicane at Turn 5. One aborted lap plus two red flags means he never completed a fully representative run. Fifth here may be the consequence of lost running, not of a performance deficit. I leave it as an open question. George Russell, by contrast, is sixth and more than half a second behind his team-mate. Same team, same run plan, same surface. When the dressing-room door closes, I understand that strategy is not drawn on the whiteboard. The half-second between the two Mercedes is the most notable personnel story of the afternoon, and it is written down nowhere. Lewis Hamilton hit the barrier at the final corner. Broken front wing, front-right puncture, and repeated instructions to stop before he complied. Notable: he was on course to beat his team-mate at the moment of impact. Ninth is a lost-running result, not a verdict on pace. But it is a real cost: qualifying preparation at a new venue cannot be bought back with experience. Oliver Bearman crashed at Turn 10 in the final minute. The Haas must now be rebuilt in the shortest possible time. This is the session's highest operational risk: lost data, plus potential exposure on component quotas and post-session parc fermé rules. Arvid Lindblad is the second concern. He crashed in FP2, the car was rebuilt, then a fresh problem appeared in FP3. Two disrupted sessions for a rookie at an unfamiliar circuit is the recipe for a ruined qualifying. Alexander Albon set no time after a rear track-rod failure. Williams lost both data and hardware before qualifying. At the back, Cadillac's Perez and Bottas were twentieth and twenty-first. In a first-cycle season, new and low-budget teams at the rear is the expected pattern, not a scoop. But it is still data: the gap between the front and the back remains wide despite the new rules. Let me be blunt about the limits of this analysis. No long-run data. No track temperature figures. No fuel loads or engine modes. A free practice session, at a new circuit, cut twice by red flags, is a weak evidentiary base for any claim about the competitive order. Data has no gender. Only the reader of data carries bias. The popular reading after this afternoon will be: Mercedes are back, Antonelli is the team's new star, and the season's order is shifting. I think that reading is half right, and the wrong half is the more dangerous one. The right half: Antonelli was fastest in both FP2 and FP3. Continuity across sessions matters more than a single lap, because it suggests the pace is repeatable rather than fortunate. And he was quicker than a team-mate who has won races. The wrong half: the entire timing sheet was set on a rapidly evolving track, after a restart, in a window short enough that anyone off the rhythm dropped out of the top ten. Bearman's last-minute crash may also have cut across almost the whole grid's qualifying simulation, meaning the final order may understate several cars. A timing sheet where everyone sits exactly where they were expected is a timing sheet worth doubting. In my trade we call that a file that is far too clean. The counter-intuitive point worth stressing: if there is a genuine performance signal from this afternoon, it does not sit with the man at the top of the sheet. It sits with the man in fourth, on three laps. And if there is a genuine question about the leading team's strength, it does not sit in the 0.1s gap to Ferrari. It sits in the half-second between the two Mercedes. One more thing about the new venue. Two red flags, several barrier contacts, and a surface not yet swept clean of rubber. This is a high-risk environment, and risk at a new circuit does not discriminate between good and average drivers — it discriminates between those with track time and those without. Qualifying here should be modelled with a higher probability of surprise than at an established circuit. What I take away from the Madrid afternoon is not the name at the top of the timing sheet, but the question about the internal order at Mercedes. If that half-second recurs in qualifying and in the race, it stops being a free-practice detail. It becomes a structural fact about the team. And how will I cross-check it? With three sources: any FP2 long-run times, GPS deceleration data through the slow corners, and the way both cars take their first runs in Q1. A timing sheet tells you who is quick. It does not tell you why.