Mexico City FP3: The Real Signal Is Not Norris's 0.001s — It Is McLaren's Double Top-Four

Methods box: This analysis rests on three things — a circulated results brief, the same FIA public timing set that feeds broadcast graphics, and the circuit parameters of the Autódromo Hermanos Rodríguez. Sample: five drivers, one session, one best lap each. Three failure cases declared up front: tyre compound unknown, intra-session lap order unknown, team-mate time unknown. Strip those three and nothing below is a claim about pace. It is a reading of a timing sheet.

What the screen shows

The first gap on the timing screen at the Autódromo Hermanos Rodríguez on Saturday afternoon was 0.001 seconds. Lando Norris first, Andrea Kimi Antonelli second. The number is correct. FIA timing records to a resolution of 0.001s, and the gap between the top two landed exactly on that single unit.

Now put that number against the circuit. The lap is 4.304 km. A 1:17.000 lap is 77 seconds, an average of 55.9 metres per second. 0.001 seconds times 55.9 metres is 5.6 centimetres. At the line, where the car is travelling around 290 km/h, the same interval sits at roughly 8 centimetres. That arithmetic is not a model; it is division, and I am showing it so it can be checked. The "millisecond thriller" being sold as a headline has a physical size smaller than a human hand — and across a single F1 lap a car changes position by far more than that at every corner.

My job for eight years has been to hold the number a sport repeats about itself against the underlying record. Here the primary record is the timing loop, and the timing loop cannot see anything finer than a millisecond. When a report makes one millisecond its headline, it is not reporting speed. It is reporting the limit of its own instrument.

Context: what kind of session this is

Mexico City sits roughly 2,285 metres above sea level. Air density is about 22 percent lower than at sea level. Three consequences are constant and constantly missing from copy: less downforce, less drag, and an unusually large slipstream effect. Cooling margins also shrink in thin air, for brakes and power units alike. A team that arrived on Friday with a high-drag concept learns by Saturday that Mexico will not forgive it.

Then the session's rank. FP1 and FP2 are working sessions — aero rakes, cooling tests, long-run simulation. FP3 is the first hour of the weekend in which nearly every team does the same thing: qualifying simulation, low fuel, soft compound, fresh tyres. That is precisely why the FP3 top five looks the most orderly and predicts the least. Of the three practice sessions, FP3 is the most quoted and the least reliable, because everyone bolts on their best lap at once and nobody gets to repeat it.

Then the window. Roughly two hours separate FP3 from qualifying. A report published inside that window is either vindicated or rendered irrelevant within two hours — and the reader will watch qualifying immediately after reading it. This is the fastest-decaying format in Formula 1 journalism. Leading the workload model for Euro 2026 and the Tokyo Olympics taught me to publish a forecast only alongside its confidence interval. Here the interval is very wide.

Then the form. This is a results brief, not a practice analysis. Its target is who was fastest, by how much, and who made up the top five. From it you cannot learn whose car is quick this weekend. You can learn what happened at the end of one session, on one set of tyres, on one lap. Those are two different claims, and conflating them is the most common error in F1 data journalism.

An audit of the copy, and one line that belongs in quarantine

The circulated brief contains a line that collides with the rest of its own narrative: Andrea Kimi Antonelli described as the championship leader. I will not soften this. The line is currently unusable.

The reason is structural. If a rookie is leading the standings, that is the biggest story of the weekend — bigger than 0.001s. Such a fact belongs at the centre of the first paragraph, not as a modifier in a headline. Of the brief's eleven information points, not one supplies a points total, a win count, or a standings position to support the claim. Meanwhile the copy spends two paragraphs on Charles Leclerc being fastest in both Friday sessions and on Oscar Piastri in fourth — the space normally reserved for the season's main contenders.

Three possibilities, in probability order. First, an extraction error upstream: "the title leader" originally belonged to another driver and attached itself to Antonelli by proximity, or a descriptive appositive such as "the Mercedes rookie" was read as a title. Second, a factual error in the source copy itself. Third, the rare case: the information is true and the copy has a serious editorial omission.

My ruling: until it is reconciled against official FIA standings, the line is unciteable. In my experience, a single line that can contaminate an entire model does not get corrected — it gets quarantined.

The boundary of 0.001s: resolution and pace are different objects

A short pause here, because this is exactly where readers and data part company.

Formula 1 records time to 0.001s. That is a technical decision, not a sporting yardstick. The transponder triggers as the car crosses the line and the system rounds that instant to the millisecond. Two cars that actually arrived 0.0004s apart and two that arrived 0.0014s apart can look identical on screen — one reads 0.000, the other 0.001. The headline number carries ±0.001s of uncertainty inside it. When the headline is itself 0.001, the uncertainty equals the headline.

That does not make the gap false. It makes its meaning narrow. A 5.6-centimetre difference tells you the two laps were near-identical. It does not tell you who was in the faster car. It does not tell you who drove the better lap. Lap time is a scalar; it does not encode where the time was gained and where it was lost. For that you need sector times, and the brief has none.

I am writing this against my own past work. In 2026 I studied 1,200 behind-closed-doors matches and found home advantage falling from 0.35 goals to 0.12. The number was clean enough to be turned straight into a conclusion. I did not, because travel, schedule density and deprivation variance were three other live variables. The same discipline applies. Numbers do not argue. They wait until you are wrong, then they speak.

The biggest number left on the table: McLaren's 1–4

Here is the part of the brief that actually irritated me.

Five names: Norris, Antonelli, Leclerc, Piastri, Max Verstappen. Sort them by team. McLaren — first and fourth. Mercedes — second. Ferrari — third. Red Bull — fifth.

McLaren is the only team in this session with both cars inside the top four. In constructors' championship terms that is the strongest signal of the entire hour, and it appears nowhere in the report.

That omission is not bad luck. It is the structural defect of linear ranking. When you list results 1 through 5, you treat each driver as a separate unit. In F1 the driver is not the unit — the car is, and the car arrives in pairs. Two cars in the top four is qualitatively different information from one car near the front and another in fifth. The first says the car's operating window is wide. The second says one driver extracted everything from it.

I am underlining this because it was a team-level signal and the report erased it structurally. Piastri fourth is a line. Piastri fourth with his team-mate first is a finding.

Leclerc's Friday versus Saturday: a variance-first diagnosis

One fact in the brief is bigger than its placement: Leclerc was fastest in both Friday sessions and is third on Saturday.

The lazy reading is that Ferrari lost pace. I doubt it, because that reading requires three things we do not have: compound, lap order, and intra-session track evolution.

What we do have is this — FP2 and FP3 are different measurements. In FP2 teams often run race fuel, the surface has less rubber, and ambient temperatures differ. In FP3 fuel is low, rubber is down, and the target is a single lap. Comparing their headline times is like comparing run totals from two pitches. On the Russia 2026 data desk I built comparisons across sixty-plus matches, and the rule held there too: until two sessions are played under the same conditions, the difference in their best times is not a difference in pace.

The correct conclusion on Leclerc is narrower: Friday's pace did not transfer to Saturday, and we have no instrument to say whether the car or the conditions caused that. Anyone calling it a Ferrari slump is filling two unknowns with one known.

But the most instructive clue of the session is buried right there, and nobody picked it up: within a single weekend, a single car and a single driver, the order changed across three sessions. That is direct evidence that lap-time reproducibility at this circuit is low — and any forecast standing on lap time here should carry a wide band.

What is missing, in order of impact

I am ranking the gaps, because not all absences are equal.

First and most damaging: tyre compounds. Without knowing who set their time on softs, mediums, or on which lap, you cannot separate single-lap pace from race tempo. In Mexico this matters more than usual, because thin air changes tyre temperature behaviour and makes degradation unpredictable.

Second: the gap from first to fifth. The brief gives only the P1–P2 interval. I need to know whether fifth is 0.15s back or 0.9s back. In the first case this is a dense group and qualifying can scramble it. In the second, it is a clear hierarchy. The millisecond headline is hanging in its own vacuum without this context.

Third: team-mate deltas. This is the most informative number in any practice session and it is entirely absent. Where is Mercedes' second car? Ferrari's? Red Bull's? Without them you can say something about a driver, nothing about a car — and not even that with confidence.

Fourth: events. Any red flag? A spin? Anyone into the wall? How many laps did each driver complete? Was the fastest lap a first flying run on hards or a final soft-tyre attack? Without that, the number has no meaning attached.

Fifth: the hosts. Two Mexican drivers are on the grid and this is Sergio Pérez's home weekend. The brief does not say where they are. Commercially, that is not a small miss — in a market where you are selling coverage, you dropped the home driver out of the top five and did not explain why.

Sixth: standings, races remaining, penalties and grid changes, the constructors' fight. Without them the result is uninterpretable.

Seventh: quotes. Not one direct sentence from any driver or engineer. Which means the entire copy contains no causal attribution. We know what happened. We do not know why.

What the conventional account gets right, stated in full

I want to resist my most reliable habit, which is the debunk tic. So first, plainly, what this brief gets right.

The timing data is hard. Norris first, Antonelli second, Leclerc third, Piastri fourth, Verstappen fifth — all of it came off the FIA loop and all of it is verifiable. Leclerc's two Friday session-toppers are equally solid. The copy made no prediction, named no villain, added no fabricated emotion. As a reporter's discipline that is underrated and it is present.

The paradox is that the most cautious copy is also the emptiest. Accuracy and completeness are different virtues, and this brief is strong in the first and blank in the second.

Now where I disagree, in three places.

One: the two-driver duel frame. The copy says Norris and Antonelli swapped the top two positions throughout the running. That is the single sentence in the piece that cannot be verified from public timing data — it needs lap-by-lap order. I do not have it. The copy does not appear to either. When you build a duel narrative supported only by the final classification, you are inviting the reader to believe both drivers fought under identical conditions, which is unknown.

Two: the dramatisation of the final moments. Bolting on fresh tyres for a flyer in the last minutes is standard procedure at every team. Presenting it as a live shootout converts a process into an event, and leaves the reader thinking something unusual happened when nothing did.

Three, and most seriously: no warning about post-session utility. A reader can finish this brief believing McLaren is clearly ahead this weekend. That may be true. But the brief gave them no basis for it — only a number that happens to point that way.

And one self-correction: I am not saying the 0.001s story is false. It is true. It answers one specific question — whose best lap was fastest when the session ended — and treating it as an answer to every other question is the actual error. Sometimes the timing sheet is the final word. At Mexico City on 26 October 2026 it was not: Max Verstappen set the fastest qualifying lap and the stewards handed him a three-place grid penalty for a yellow-flag infringement, putting Charles Leclerc on pole. On that day the screen and the official result disagreed. The screen is a record, not a verdict — and hiding that distinction from readers is not journalism, it is transmission.

What to watch next

In qualifying I will not be watching the order of the top five. I will be watching three things.

First, team-mate deltas, especially at Ferrari and Red Bull. If Leclerc's team-mate is three-tenths back, Ferrari has a car with pace and the Friday-to-Saturday swing is about tyres or driving. If they are level, the car is at its limit.

Second, the stadium section. Mexico's final complex is slow, tight and mechanical — the one place where a downforce deficit costs least and traction matters most. If McLaren cannot take time there, its top slot came from the thin-air aero trade rather than outright car performance.

Third, the question nobody publishes: what is the actual correlation between FP3 and qualifying at this circuit? I do not have the answer, and I suspect nobody has it publicly. The day that number surfaces, every FP3 report will be written differently. Until then, it is worth remembering that a gap the size of a hand is a fact, and a story built from a gap the size of a hand is an inference.

The question is not whether the timing screen is right. The question is what we asked it.