A 0.2-Second Sensor Lag at San Siro and the Trap of the Empty Data Table

In the summer of 2026, at the Milanello training centre, I was handed the task of auditing the motion dataset from 20 Serie A matches of the 2026-17 season. The summary table in front of me was tidy to the point of suspicion. Milan generated 1.85 xG per match at San Siro and 1.02 xG per match away, while the actual goal counts from the two sets of matches were identical. A hurried reader would write it up immediately: the team attacks better with a home crowd and was merely unlucky in front of goal. I nearly signed off on that assessment. What held me back was a detail nobody had bothered to check. Every build-up sequence starting from the goalkeeper carried a timestamp slightly out of step with the video. I replayed each sequence, cross-checked each frame, and found the culprit. The sensor planted in San Siro's south-west corner was lagging by 0.2 seconds. Two tenths of a second is not enough for the human eye to catch live. It is just enough for every short pass from the goalkeeper to be logged at the wrong moment, with the error cascading through the entire sequence that followed. We recalibrated the hardware and wrote a 14-page internal report. Vincenzo Montella used the findings to shift more circulation to the right flank. Milan won five of their last eight matches and sealed a Europa League place. I retell that old story because it is repeating everywhere. Today, every match in Serie A, the Premier League, the V.League or a provincial youth tournament can be captured by multi-camera systems. Training shirts carry GPS chips, accelerometers, positional data down to hundredths of a second. Clubs use it to pick line-ups, value players, decide transfers and explain failure to a board. Nobody objects to any of that. The problem sits elsewhere. A measurement system is only as trustworthy as its last calibration. The 0.83 xG gap between home and away at Milan in 2026-17 was a beautiful number. So beautiful that had I presented it in a coaching meeting, nobody would have asked a follow-up. People like metrics that agree with their instincts. Full stands, a confident team, high xG. That chain of logic is self-satisfying, and that self-satisfaction is the most dangerous thing in an analysis room. Cross-checking against the video, I found the system's record did not match what happened on the grass. A pressing escape was tagged as a long pass when it was in fact three short passes in a row. A counter-attack was filed under aerial duels. Those small errors added up into a systematically distorted picture, and that picture went into the meeting as evidence. From the training ground in Milan to the esports screen, the law of the gap holds the same. In more than forty years in this trade I have reported live from 406 consecutive Grands Prix. On a race track, one misreading speed sensor is enough to send engineers into the wrong pit window, and the mistake only surfaces once the tyres are dead mid-stint. Data does not protect itself from the person reading it. Every tracking number belongs on the operating table, not on the altar. On 27 June 2026, in Kazan, I sat in the commentary box and posted a line at minute 70. Germany's defensive line was pushing an average of 68 metres high, pressing had failed 17 times, and South Korea had already produced 12 counter-attacks. Without dropping the block, the goal would come from an aerial situation. In the 90+3rd minute, Kim Young-gwon scored exactly that. Germany left the World Cup bottom of their group. Every collapse has a precondition; few people are willing to look at it beforehand. But that is the easy part. The hard part is this: a back line pushed 68 metres high can still be harmless if the midfield covers on time. Data describes distance, not timing. The gap between centre-back and goalkeeper looks like a long vertical rectangle, but that rectangle only becomes a fatal door when the central midfielder fails to drop. The same quantity, two different outcomes, depending on something no table measures. The Germans that year forgot that football never forgives complacency. They had the data, the specialists, the budget. What they lacked was one person willing to ask: when was our sensor last calibrated? That is the counter-intuitive view I have held throughout my career. When an analysis department presents a clean table, my first question is not about its content but about the conditions it was produced under. Who placed the cameras. Who labelled the events. Who decided a passage of play was a pass rather than a clearance. Those choices are human, and they shape conclusions that travel under the name of objectivity. More dangerous than wrong data is empty data. When an analytics system returns an empty table, most people's reflex is to treat it as good news. No red flags, no injuries, no problem. But an empty table usually means the data extraction process broke somewhere, and broke in silence. A broken pipeline still emits a file with the correct structure, the right fields and the right format, just missing content. Downstream systems accept it, validate it, and pass it on. Nobody raises an error, because there is nothing to raise an error about. In football, this kind of failure happens far more often than outsiders imagine. A match with no GPS data because the devices were not started correctly. A training session half-recorded because the connection dropped. A player missing from a report because his identifier was mistyped at the start of the season. The report still looks clean. Conclusions are still drawn. And a personnel decision is made on a foundation of nothing at all. I saw a case like that at Milan. During the 2026-17 season, several youth-team sessions were under-recorded because the receivers were not synchronised with the server. Reports still came out every week. For months, nobody noticed. When it surfaced, the coaching staff had to rebuild the entire physical assessment of that group from scratch. If you run a football club, the thing to do sits elsewhere: put one question at the top of every report. If this file were empty, would we know? For clubs in the V.League, where analytics budgets are thin, the risk runs higher. A team may have only two people handling data, filming, labelling and writing reports. When the workload outgrows the staff, the first thing cut is always the cross-check. The table still appears, but nobody verifies it. And a coach who trusts that table will pay for it in points. An empty stadium does not kill a match, but it takes away something no metric can measure. A sparse ground changes the rhythm of a game in ways no camera records: instructions become audible, players hesitate a fraction longer in duels, referees handle incidents differently. Read only positional data and you will describe the shape of the match correctly while missing the reason it took that shape. I have kept one habit since my early years in the job: whenever I receive a dataset, I go looking for the places it says nothing. A gap in a table always means something. It may be an unlabelled passage of play, a dead sensor, or an event the collector never imagined. Reading gaps is harder than reading data, because no software does it for you. Looking at the rest of the season, I will track one specific thing. For every club I get to observe, I will cross-check the published tracking report against video from at least two matches. If the number of mislabelled events exceeds one per cent of total events, every derivative metric should be suspended and audited. It is the simplest test available, and also the one fewest people are willing to run. Data only tells part of the story; the rest lies with those who know how to listen.