Nine Chapters, Not One Fact: The Data Verification Gap in Professional Sport

MILAN — On a March morning, the file on my screen ran to nine chapters. Every heading was present, every table carried its structure, every section sat in the order any coaching staff would want. I read from the first line to the last and found not a single hard fact: no team, no driver, no lap number, no Grand Prix, no timestamp. Every cell read “insufficient information”. Yet the closing section was immaculate, complete with a risk rating and a single line stating that no issues had been found.

In forty-one years on this beat I have read thousands of technical reports. This was the first time I held an empty report packaged as a clean one. A void result and a no-fault result are entirely different things, yet on paper they look identical. And I remembered the line I keep repeating to younger colleagues in the studio: data only tells part of the story; the rest lives in whether anyone knows how to listen.

Over the past decade professional sport has moved in one unmistakable direction. In Formula 1, each car pushes hundreds of telemetry channels every second: tyre temperatures at all four corners, brake torque, energy deployment maps, steering angle, the gap to the car ahead. In football, optical tracking systems log the position of twenty-two players and the ball at twenty-five frames per second, turning every passage of play into a set of coordinates. Metrics such as xG, PPDA, counter-attack counts and high-intensity running distance have left the analysis room and walked onto the broadcast.

Vietnamese fans now read an xG table after every V.League round the way they read the scoreline. Clubs at home are hiring independent analysis providers, buying tracking hardware, building their own video rooms. That is real progress, not a passing trend. But while analysis budgets have risen, the budget for verifying that same data has barely moved. Everyone wants one more metric; very few want to pay for the question: what was this measured with, when was it measured, and what is the error margin?

I have covered Formula 1 since 2026 and I once sat on the coaching staff at AC Milan. Two different environments, one shared illness.

In 2026, while working with the AC Milan coaching staff, I was asked to audit the movement dataset covering twenty Serie A matches from the 2026-17 season. The first pass looked superb as a narrative: Milan’s expected goals at San Siro stood at 1.85, against 1.02 away from home. Nearly double. Add a few lines about home-ground psychology and you have an analysis that sounds entirely reasonable.

But the actual goal counts at home and away were almost identical. When a team creates twice as many good chances and scores the same number of goals, either it is desperately unlucky or the data is broken. I chose the second option and began cross-checking video against raw data, frame by frame.

The fault sat in the south-west corner of San Siro: a sensor running 0.2 seconds late. That delay is not enough for the naked eye to catch on replay, but it is enough to shift every goalkeeper-initiated build-up off its tracking axis. An entire season of Milan’s home data was pushed toward the touchline while the ball itself stayed in the middle of the pitch.

I wrote a fourteen-page internal report recommending recalibration rather than another software purchase. Head coach Vincenzo Montella used the finding to shift the emphasis of ball circulation to the right flank. Milan won five of their last eight matches and secured a Europa League place. One sensor off by 0.2 seconds, and a whole season changed direction.

Every collapse has a precondition; it is just that few people bother to look beforehand. Here the precondition was not on the grass. It was inside a box bolted to a stand.

A year later, at another tournament, I learned a second lesson. At the 2026 World Cup, during Germany against South Korea on 27 June 2026, I posted a short analysis at the 70th minute: Germany’s defensive line was holding an average of 68 metres high, their pressing had failed 17 times, South Korea had already produced 12 counter-attacks, and unless the block dropped deeper the goal would come from an aerial situation. In stoppage time, Kim Young-gwon scored exactly to that script.

Thousands of accounts piled in to mock me for turning emotion into arithmetic. What I took from it was not that I had been right. What I took from it was how to write. Left on its own, the figure of 68 metres means nothing to a reader. It has to be translated into a spatial image: the distance between centre-back and goalkeeper stretched like a vertical rectangle, the defensive line like a zip burst open all the way to the valve box. From then on I dropped bare numbers and moved to describing structure.

Every tracking figure belongs on the operating table, not on the altar.

Back to the empty nine-chapter file. On the surface it did not look like a failure at all. It had the structure, the headings, the sequence. It is precisely that formal completeness that makes a skimming reader assume everything has been checked. In the data-publishing business this is the most dangerous class of error, because it makes no noise.

I call it the format-completeness illusion. A report with no data but nine full chapters is harder to catch than a report with three lines and a mistake. The three-line report gets challenged immediately. The nine-chapter report gets filed, cited, and eventually used as the basis for a real decision.

The instinctive reaction when data goes wrong is to blame the algorithm. I think that is the wrong place to put the question. An algorithm never claims to be right; it simply calculates from whatever it is fed. The fault sits at the human checkpoint, where somebody should be asking a question before the data moves on.

In the sports data pipeline there are usually three gates: collection, processing, interpretation. The fourth gate, verification, barely exists. At many clubs nobody holds the job title of source checker. At many broadcasters nobody is accountable for validating a metric before it goes to air. The result is that an empty dataset can travel the entire pipeline without meeting a single barrier.

What worries me more is how we read files like that. A void result is routinely understood as “no problem found”, and in operating culture “no problem found” is nearly synonymous with “everything is fine”. Those two states are worlds apart. The first means there is no data to speak. The second means there is data, and the data says nothing is wrong. On the report page, both arrive as blank cells.

A contract only looks good on paper until somebody tries to fit it into a running system. That is true of a player, and equally true of a dataset you have just bought.

Next week, when a metrics table is put in front of me, I will ask three things: what was this measured with, at what timestamp, and who checked it. If there is no answer to the third question, the rest is decoration. Anyone in this trade should audit their own numbers before asserting them on air. And in an industry that runs on faith in data, the most frightening thing remains a blank cell presented far too neatly.