F1's data rooms are losing the ability to verify themselves

Fourteen pages landed on my desk in Milan: headings, tables, fields waiting for numbers, and inside them blank space. The person who sent it had run exactly the process the whole paddock runs, and the process returned an empty payload. In my trade that document has a name: the clean sheet. A wrong measurement can be argued with. A clean sheet gets believed, because nobody thinks to re-check a document that has already been printed and stamped. I looked at it for about three minutes before going on air. Beside it sat a free-practice timing sheet, a few raw telemetry traces, and a young colleague waiting for me to say something absolutely certain about the fastest car's pace. The temptation was simple: fill the three empty fields from memory, and the broadcast would run smoothly, nobody would notice, until the second car pitted half a second earlier than projected and the whole model collapsed. That is why this piece is about Formula 1 in general rather than about any single Grand Prix. Formula 1 today is a sport run on hundreds of sensor channels per car — teams have publicly cited a figure around 300, and it has only gone up. Every lap generates lap time, sector times, top speed, GPS position, the temperature of all four tyre casings, pressures, fuel consumption, steering angle, brake force, and dozens of secondary variables engineers build on top. That produces a four-link chain: the sensor measures, the calibration system interprets, the model computes, and finally the story is told on air. The chain is only as strong as its weakest link. The trouble is that the first three links are nearly invisible to viewers, while the fourth is exposed to millions of people every weekend. A broadcast graphic says Driver A is 0.4 seconds faster in sector two, and the audience assumes that is physical truth. In reality, sector two is usually interpolated from fixed timing loops buried in the track surface, and interpolation always carries error. When someone states that a driver lost the brakes at turn seven, the first question has to be: which sensor produced that, at what sampling rate, and who calibrated it. Start with the first link. Official F1 timing relies on a transponder on the car triggering loops beneath the asphalt at defined points. Lap time is therefore highly trustworthy, accurate to thousandths. Sector time is a different animal: only a few real loops exist, and the rest is reconstructed algorithmically from the car's own telemetry. Two drivers can be compared through two different frames of reference, and nobody annotates the graphic to say so. GPS is the next weak link. A sampling rate of 5 Hz or 10 Hz sounds generous, but at 300 km/h each sample is up to eight metres apart. On sections with elevation change, tree cover, or a steel grandstand nearby, position error can reach several metres — enough for an algorithm to assign an inside line to the wrong phase of a corner. I have seen ideal-lap traces that looked beautiful, and when checked against the video the car was running a full metre off the drawn line. The drawing was prettier than the reality. Tyre degradation modelling is where self-deception comes easiest. To compare laps from two different moments, engineers must subtract the fuel already burned — roughly eight per cent of car weight at the start of a race — then add the track rubbering in, then subtract air and track temperature. Those three corrections combined can exceed the difference you are trying to prove. A linear degradation model only holds for the first twenty laps or so; after that the tyre enters a non-linear region, and every linear forecast becomes an illusion. Every tracking figure belongs on the operating table, not on the altar. Moving from the tyre model to the strategy call, error multiplies. The undercut calculation needs two inputs: your own lap pace and the circuit's pit loss — the time surrendered by stopping versus staying out. Pit loss sits around 20 to 25 seconds at most circuits, and teams compute it per car, per stop, because pit box position and entry angle differ. If your model is off by 0.4 seconds and your rival's margin is 0.3 seconds, you call it wrong. Get it wrong once, lose a position. Get it wrong twice in a race, lose the weekend. At the top of the stack sits wind tunnel to track correlation. Under the Financial Regulations and the aerodynamic testing restriction mechanism, wind tunnel hours and CFD runs are allocated in reverse order of the previous season's standings: the last-placed team gets more development time than the champion. That mechanism only means anything if tunnel data converts into real lap time. When a new component goes on the car and delivers nothing, the first question has to be: what did the tunnel say, what did the track say, and who checked that both were measuring the same thing. I learned this lesson through one specific error, and it came from football, before I settled fully into the paddock. In 2026, while working within the AC Milan coaching staff, I was assigned to audit the movement dataset from twenty Serie A matches of the 2026-17 season. Expected goals at San Siro read 1.85 at home against 1.02 away, yet actual goals scored were level. The gap was too large to explain through form. Cross-checking the footage, I found a sensor in the south-west corner running 0.2 seconds late, which pushed every goalkeeper-initiated build-up out of alignment inside the model. My fourteen-page internal report recommended recalibration. Vincenzo Montella used the finding to shift more circulation to the right flank; the team won five of the last eight matches and secured Europa League qualification. There is one more layer of data that no sensor records: the voice on the radio. When an engineer reads a lap half a second slower than projected and his voice drops on the final word, that is a signal. When a driver stays silent too long before answering, that is a signal too. Data only tells part of the story; the rest lives in whether people know how to listen. In 41 years of watching this sport, I have never seen a model forecast hesitation. The counter-intuitive part: more data has made the paddock less accurate, not more. Sensor channels have grown exponentially while verification capacity has grown in a straight line — and in the cost cap era, the roles that do verification are the first to be cut. A team can run three hundred channels and keep only two people with enough hours to cross-check them. The gap between the volume of data and the number of people checking it is exactly where error is born. And the most dangerous error is the one that looks reasonable. An absurd measurement gets challenged immediately. A tidy measurement that fits the story the paddock already believes travels straight from the engineering office to the broadcast, to the newspapers, into a sponsorship contract, without a single check. Every collapse has a premise; few people bother to look early enough. The clean sheet I held that day was, in the end, safer than a page full of unsourced figures — because blank space forces people to stop. In this regular season, with a fresh dataset and a fresh conclusion released every weekend, the worthwhile investment sits in the process of asking again, with sensors a distant second. What does this data measure, who calibrated it, and has it ever been checked against a second source. From the training ground in Milan to the telemetry room in the paddock, the rule of the gap holds. The question I keep for the next round: if a team had to discard half its channels, which team would notice first — and by how many laps ahead of the rest?