F1's Empty Spreadsheet Problem: Before the 2026 Rules, Data Discipline Is the Real Seat

Last week a report landed on my desk. Nine analytical sections, three tables, a risk matrix. Not one hard fact. The title was blank. The source was blank. The list of information points was blank. The field meant to carry the entities instructed that they be identified from the information points above, and above there was nothing at all. This was not a report. It was an empty form that somebody had filled in with great confidence. In thirty-one years of this work it has proved the most useful document I have handled, because it showed plainly the thing every paddock hides: we do not lose to a shortage of data, we lose to the appearance of it.

Next season the technical foundation of Formula 1 changes. The power unit splits roughly half and half between combustion and electrical output, the fuel becomes fully sustainable, the cars get lighter and narrower, and active aerodynamics returns. Alongside that sit the cost cap and the tiered aerodynamic testing restrictions. Together they produce one effect: less running at the track, more simulation in the factory. When the number of wind tunnel hours you may spend in a season is fixed in advance, the advantage goes to the team that can falsify its own model fastest, not to the team that accumulates the most data.

One point needs settling, because the paddock is confused about it. The volume of simulation is not a competitive advantage; its auditability is. The 2026 cars will be machines nobody has prior experience of, with no aero mapping, no tyre degradation history, no proven undercut window. A team that says its model indicates something is saying nothing. A team that says which run number the model rests on, what the sample size is, and what the model cannot see, is the team that moves first.

First failure: an empty payload. The structure of the report exists; the content does not. In F1 its direct form is the upgrade report that lists the new floor specification, the carbon lay-up and the wind tunnel hours, and contains not a single lap-time delta. I have read reports with eleven pages and zero numbers. A template is not a fact.

Second failure: a circular dependency. That document told the analyst to extract entities from the information points above, while above there was nothing. This is not merely a technical slip, it is a trap. When an analyst has a slot to fill and no material to fill it with, the analyst invents the material. Strategy debriefs do this daily: the data says the undercut works, and when you ask which data, the answer is the data. A model with no name is not a model, it is an opinion.

Third failure: lost provenance. That report carried no outlet, no date, no retrieval timestamp. In driver-market reporting this is fatal, because the source is the value: an established paddock journalist and an aggregator cannot be read at the same weight. On the technical side the same rule governs wind tunnel correlation. Track data agrees with it is a claim, and until the run number, the ambient temperature, the tyre compound and the track evolution are written down, it is not a verifiable one.

The teammate comparison remains the only clean control in the paddock, because the car is identical on both sides of the garage. Max Verstappen, Lando Norris, Charles Leclerc, Fernando Alonso: in every argument about them, that control is the thing least often invoked. And when even the control arrives without provenance, the argument stops being data and becomes taste.

Now the part that bothers me most, which is the language of the conclusion. That report said, everywhere, that no risk had been identified. That is not the same statement as no risk exists, and the paddock merges the two every week. When a team announces that its upgrade showed no risk, the useful question is not what the risk was. The useful question is what the team actually looked at.

One example I keep returning to, because it is about hazard rather than arithmetic. Force India's fourth-place finishes were filed as overachievement for years, until you price the debt and the administration that followed it. The Buddh International Circuit was filed as a fan failure, until you set the roughly 95,000 race-day crowd of 2026 against a gate that had fallen to about 60,000 by 2026, and against the promoter's balance sheet. There is a ledger behind every lap time, and any result read without it is incomplete.

The received wisdom is that whoever holds the most data wins in 2026. My reading runs the other way. Under a cost cap and a testing ceiling, the binding constraint is not the quantity of data but its reproducibility. A team able to disprove its own model quickly will outrun a team that spends a season proving it right. A second inversion: both media and fandom mistake precision for proof. A gap of 0.04 seconds, a three-race trend, a probability taken to three decimal places. The tools will produce those digits whether or not the data has earned them. I keep a rule. The sample size below which I refuse to rank anything gets stated in the copy, not hidden in a footnote.

And a third inversion, this one against myself. Contrarianism hardens into a brand. When the unexpected answer becomes your identity, you settle the conclusion before you open the dataset. So I keep a running tally of the occasions when the paddock consensus was right. It gets published annually, whether or not it flatters me.

One further point carries the real lesson of this episode. The correct reading of an empty report is not that no risk was found; it is that no analysis was possible. Those two conclusions sit at opposite ends of the scale, and organisations that treat them as equivalent end up mistaking their largest weakness for their greatest strength.

What to watch at the start of 2026 is not the count of upgrades a team brings. It is whether the team has written the line stating what its model cannot see. The team that writes that line will be the team that corrects fastest. And one question the first five races will answer for me: when a team's simulation fails to match the track, will it call the data wrong, or the circuit?