The Empty Dossier and Three Field Verifications

July 2026, the second-floor meeting room of the U19 training centre in Guangzhou. I opened a forty-page opponent dossier, counted eleven statistical tables, and stopped at the tactical notes section — every cell was empty except for one line scribbled by a former assistant: not yet updated. The first leg was forty-eight hours away.

That night I did not reopen my laptop. I pulled the footage of the opponent's last three matches, slowed down every pressing sequence in the final third, and counted by hand. Two successes in eighteen attempts. No table in the dossier contained that figure, and it was the only thing that changed our approach before kickoff. We won the return leg two nil.

Nine years later I still keep the habit: whenever a dossier looks too complete, I go looking for the empty space inside it first.

Performance analysis in Guangzhou has run on a fairly standard pipeline in recent years. Fixed cameras record the match, software segments events, a tagging team assigns labels, and then the analyst — usually the last person to touch the file — assembles everything into a story. Every layer can fail. A camera drifts off axis, software misreads a defensive sequence as a serve, a tagger changes shift mid-match, and the analyst receives a product that looks remarkably tidy.

The problem is this: an empty file and a wrong file do not cause harm in the same way. A wrong file will be argued with, because it contradicts what people remember seeing. An empty file contradicts nothing, so it gets filled with assumptions, and assumptions are never audited.

Across eighteen years of watching this industry, I have seen the pipeline grow thicker in data and thinner in field notes. A single match now generates hundreds of metrics. But metrics are defined in advance and matches are not. The gap between those two things is exactly where the analyst is supposed to sit.

In Vietnam, where I started, field conditions forced the opposite discipline: no cameras, no taggers, so every conclusion had to pass through the eye. In China I met the inverse situation — enough tools to avoid looking at all. Both extremes produce the same error: believing the table is the match.

Three times in my career I have had to prove otherwise.

The first was the U19 match I just described. After the manual count, I redrew all ninety minutes on the tactics board and cross-referenced it with our GPS positioning data. The forwards were not running in the wrong direction — they were running four to six metres off the central axis. An error that small gets flagged by no software, but it was enough to put every line-breaking pass into an opponent's feet. I proposed that the head coach narrow the distance between the two wide midfielders from twenty-eight metres to twenty-two. Those six metres were not a minor adjustment; they were the entire difference between a pressing system with an exit and a pressing system that seals its own exits. The return leg finished two nil.

The second was the summer of 2026. I was invited into an opponent-analysis group for a local television channel during the major tournament in Russia. Germany met South Korea on the twenty-seventh of June; Germany lost two nil and went out. The whole country talked about the shock. I spent three days reviewing every midfield sequence and counted nine occasions on which Mesut Özil drifted into exactly Toni Kroos's zone, stripping the midfield of structure at its lowest layer. At the same time, both fullbacks pushed an average of sixty-three metres forward with nobody covering the space behind them. Two parallel forward runs with no cover behind them are a funnel, not an attack. My three-thousand-word piece was read by few, but the club's head coach printed it and pinned it to the meeting-room wall.

The third was 2026, when the stadiums stood empty. I was then a mid-level staffer in the analysis department of a Guangzhou club. Six matches without crowds, four of them with possession above sixty percent, one win. At first I blamed expected-goals drift. But on review, opponents were pushing noticeably higher — with no crowd noise to slow their first step. Our build-up lost its structure at the very first layer. I built a comparison table of forty-five indicators between the attended season and the crowdless season, and the pressing-distance metric deviated by thirty-two percent. Context is not stored in a spreadsheet, but it changes the meaning of every row in the spreadsheet.

Those three episodes taught me the same lesson in three different ways. The first was a lesson about resolution: raw data was fine enough to show the error but not fine enough to show the cause. The second was a lesson about layers: a problem at the structural layer can never be explained by data from the outcome layer. The third was a lesson about measurement conditions: the same metric measured in two contexts is two different metrics.

When the data stands still, I start reading the gaps between the numbers again. Those gaps usually have a concrete shape — a patch of ground nobody occupies, a running rhythm half a second late, a distance between two teammates that widens minute by minute.

Translated into table tennis, the language becomes even clearer. The first three shots decide most of the structure of a rally, yet most statistical tables record only the outcome of the final shot. A player misses not necessarily because the technique is wrong; often the racket meets the ball three to five degrees off the body line, or the gap between the feet opens half a beat earlier than the hip rotation. Three degrees and half a beat never appear on any aggregate chart. Based on my experience watching matches, most arguments about a player are arguments between two people looking at two different layers of the same movement.

That is why I always split analysis into three layers. The foundation layer is raw technique — grip, hip rotation, contact point. The middle layer is structure — spacing, height, rhythm. The top layer is in-match decision-making. A conclusion is only trustworthy when it states which layer it stands on, and the reader needs to know which layer they are reading.

Here is the counterintuitive part: an empty dossier is more dangerous than a wrong one. A wrong dossier has objectors. An empty one gets filled, and the person filling it does not know they are filling it.

The natural reflex when a file comes back empty is to re-run the pipeline, refresh the data, push the timeline. In most cases that is the right operational decision and the wrong analytical one. If the pipeline broke at the tagging stage, re-running it produces a newer file that looks fuller while carrying the same systemic fault — except now nobody suspects it any more. There are cracks that never show on a tactics diagram, and they tear an entire campaign apart.

In Guangzhou, where a new metric is introduced almost every week, I learned to separate fashion from foundation with a single question: will this metric still hold next month, once opponents have reviewed their own footage? A metric that survives that question is foundation. A metric that only looks good for one round is fashion. A system runs well only when the pieces inside it are not cracked — and the piece most likely to crack is always the one nobody bothers to check.

One thing is rarely said: sometimes the correct action for an analyst is to refuse to deliver the report. Not from lack of effort, but because a report pitched at the wrong layer will shape training for weeks, and the cost of fixing it far exceeds the cost of waiting two more days.

Before the next match, I will run the check in reverse order to habit: who was the last person to touch this file, under what conditions did they measure, and what is certainly absent from it. Those three questions cost less than a meeting, and they usually reveal more than the table does. Numbers do not lie, but they keep quiet about the part that matters most. The job of the person sitting beside the table is to hear that silence before it turns into a wrong decision.