The Empty Report and the Discipline of Two-Source Verification

On the night of 11 July 2026, at the Luzhniki Stadium in Moscow, I sat in the corner of a coffee shop on Dagu Road in Shanghai with a lined notebook and a pencil. Croatia versus England, a World Cup semi-final. Kieran Trippier put England ahead from an early free kick, and for more than an hour afterwards Gareth Southgate's side held more of the ball. In the 68th minute, Ivan Perišić equalised. In the 109th minute, Mario Mandžukić ended it. In my notebook I had twelve Croatian passes driven straight through the central corridor, against six from England. That night I wrote a two-thousand-word piece. It received thirty-seven reads.

Seven years later, a nine-dimension analysis landed on my desk. All nine dimensions were empty. No tournament name. No team. No player. No date. No source. A fully constructed skeleton, with a heading for every cell, with tables, with an entire risk-warning section — and not one piece of data inside any of them. I read it three times, then did what I always do. I went to cross-check. There was nothing to cross-check.

From one semi-final to one hard rule

In 2026 I was a first-year economics student in Shanghai. That semi-final taught me something I have never forgotten: possession share is the most misleading metric in football. The team with more of the ball is not necessarily the team controlling the match. Croatia let England have possession and drove the ball straight through the middle, at double the opponent's rate according to my notes. From that point on, I never used raw possession or raw pass volume as a primary argument. I moved to event-level data, and I set myself a hard rule: every conclusion needs at least two independent sources.

When the pandemic shut down global football in 2026, I used the gap to learn Python and build a database of 1,540 matches from Europe's top leagues and every World Cup from 2026 to 2026. In the pandemic, I built an empire out of numbers nobody was watching. It still stands.

I built an index I call defensive compression, combining PPDA with the location of the first contested ball. Running a backtest across 58 rounds, I found that Leicester City of 2026/16 — the side the press called a miracle — actually ranked third on that index. N'Golo Kanté and Jamie Vardy did not produce a miracle. They operated a system organised to a degree that is hard to believe. The piece drew 2,300 reads, and a football scout left a comment confirming the method had value.

The Euro tournament played in 2026 taught me the second lesson. My model produced a top four: Italy, Spain, Belgium, France. Italy won, their first European title in 53 years, with Gianluigi Donnarumma named player of the tournament. But the model also predicted France would meet Italy in the final, and Switzerland eliminated France in the round of 16 on penalties. I wrote a supplementary piece about the error. Variance is not the enemy — it is a mirror held up to the arrogance of a prediction.

At the 2026 World Cup in Qatar, I tracked every Morocco match. Against Spain, Morocco's PPDA was 7.7, the lowest at the tournament, while their centre-backs made 33 clearances inside the box and Yassine Bounou saved in the shootout. The piece reached 150,000 reads on Weibo and put me in my current role as a data analyst.

Why I refused to fill in the skeleton

A serious sports data report, whether on football or esports, cannot begin with a conclusion. It begins with a chain of identification: which discipline, which version, which tournament, which team, which player, which date. The nine-dimension skeleton I received that day was built on exactly that logic — a patch and meta layer, a tournament format layer, a team and player layer, a regional landscape layer, a club finance layer, a rules and governance layer, a risk profile, a public narrative, and an industry transmission layer. Every layer requires a named entity.

Without a game title, the patch layer is completely inert. You cannot distinguish a minor numerical tweak from a mechanic rework. You cannot say who benefits, who suffers, or whether a champion's win rate has shifted. In esports, where the patch cycle is measured in weeks rather than seasons, that ambiguity is far more expensive than in football. Esports is not slower than football — it simply runs on a different clock.

Without a named tournament, you cannot say anything about upset probability. A single-elimination format and a double-elimination bracket produce two entirely different distributions of outcomes. Without a named transfer, you cannot estimate the trade-off cost of assembling a roster.

And here is the most important part, the part that many reports pass through a system without anyone stopping them. An empty evidence base is not a conclusion — it is an unfinished state of analysis. When a compliance checklist has no subject in scope, the correct output is no subject in scope, not no risk present. The absence of information about unpaid wages is not evidence that wages were paid. Those two statements differ logically, and in this profession they differ in consequence.

Data does not lie, but it learns how to hide the most important thing. An empty cell in a table does not tell you the club is fine. It tells you that you have not looked hard enough.

The only anchor of an empty report

After reading that analysis three times, I extracted exactly one genuine finding, and it concerned no team at all. It concerned process: the information-extraction step returned empty, including fields that should have been populated automatically, such as the domain label. When every field is empty at once, the probability is high that the defect sits in the processing stage rather than in the source document. This is the kind of conclusion an analyst can reach without knowing which teams are playing, because it rests only on the structure of the data itself.

I did not fill in the blanks. I could have done so very easily. I know enough about the meta of most major titles to write something that would read as entirely plausible about a particular team being disadvantaged by an update. But when I look back at myself in 2026, I see a first-year student writing from his own handwritten notes, read by thirty-seven people, who still documented his method. I do not want to become the person who walked away from the very principle that built his career.

Every figure on a transfer sheet is a confession by an executive, and every empty cell in an analysis sheet is a confession by the analyst.

The counter-intuitive angle

Most debates about the quality of sports analysis revolve around wrong data. Wrong data gets caught, argued over, corrected. It leaves a trace. It has an opponent to fight.

Empty data does not. An empty report passes through a system in the form of a document that was checked with nothing found. It carries the shape of diligence. It has tables, a risk section, the correct formatting. No reader would assume it contains nothing, because it is presented as though it contains everything. In a decision pipeline, that is the most dangerous class of error: a safety certificate issued for an entity that was never identified.

The second counter-intuitive point concerns the audience. Fans come for conclusions. They want to know which team is stronger, who will win, whether this patch breaks the meta. A valuable analyst is measured by the ability to refuse a conclusion when the evidence is insufficient. Fans remember the goals; I remember the probability before the goal happened — and sometimes that probability is recorded as undetermined.

The third counter-intuitive point sits in the power structure of the industry. In both football and esports, the game publisher is both the rule-maker and a party with commercial interest, and no independent arbitration mechanism stands above both roles. An analysis of governance that cannot name the rule system, the league, or the organisation can conclude nothing about compliance. It can only record that the question has no subject to address. A process that finds no violation does not mean no violation exists.

Signals for the next cycle

Three signals I will track in the coming cycle, and all three belong to method rather than to results.

The first is the possibility of re-extraction. If the source document is still available, re-running the first stage with forced entity recognition — discipline, organisation, individual, tournament, date — would almost certainly recover most of the lost structure. An analysis is only as strong as its evidence base, and that base can be repaired.

The second is the null rate across an entire batch. One empty report is an accident. Three empty reports in the same batch is a systemic defect, and systemic defects have to be fixed at the root rather than the branch.

The third is how the system receives null results. An analysis with no entity in scope should be flagged as not yet analysable, not filed as checked and cleared. One season is a statistical sample. A decade is evidence.

I still keep the notebook from that night at Luzhniki. It reminds me that an analyst's value lies not in the number of conclusions produced, but in the number refused. Thirty-seven reads in 2026 did not make me quit. A nine-dimension empty report this year will not either. All I have to do is record, clearly and with attribution, that this time the data did not arrive.