The Empty Report and the Transfer Window: The Craft of Saying 'Not Enough Data'

3 a.m. in Guangzhou. The air conditioning sits at 26 degrees, three monitors are still on, and I have 47 data files open at once. My right knee twinges exactly as I drag the slider to the final frame of the chart. The knee pain taught me how to count, and I have never stopped counting.

But that night I counted something else: the empty cells. Forty-seven files, twelve data fields each, and 61 percent of the cells held no value. I stared at that number longer than necessary. Then I typed into the last page of the report a line nobody in this trade wants to type: "Insufficient information to reach a conclusion."

Six hours later, my editor called. He asked what I had. I told him I had an empty report. He laughed. I did not.

The transfer window runs on cheap fuel

This is the third week of the transfer window. This market runs on a cheap, fast-burning fuel: rumour. A social media account posts one line. Ten minutes later there are twenty articles. An hour later there is a "source close to the deal". By the end of the day that name sits in the projected line-ups of three different clubs. Nobody in that chain lied. Nobody checked either.

I have worked as an analyst since 2026, after retiring at 31 because of my knee. Before that I played badminton, and before that I anchored broadcasts of major events, including the Sudirman Cup. That job taught me something analytics has to relearn from scratch: silence is information too. When an athlete goes quiet at a press conference, that is data. When a coach answers in circles, that is data. The transfer window is a three-month press conference, and most of its content is silence.

The transfer market is just a spreadsheet that knows how to wear a shirt. The transfer fee is the loudest number, but the release clause, the wage schedule and the sell-on clause are the real variables. A 40 million euro transfer spread over four years is a different animal from 40 million paid once. Cash flow is data. Money on the line is the most honest measure of belief, and during a transfer window the odds on whether a player stays or leaves are a better indicator than any "source close to the deal".

In Vietnam, fans follow the European transfer window through one extra layer of mediation: language. An English report gets translated, then summarised, then commented on. With each layer, a conditional sentence becomes a declarative one. After four layers, a rumour with a 30 percent probability becomes a completed deal. I always read the original first.

Four times the numbers spoke, and one time they went quiet

In 2026 I published an analysis of Eran Zahavi at Guangzhou R&F. He scored 27 goals in the Chinese top flight, but his total expected goals across the season was only 21.5. A gap of 5.5 goals. I wrote that the following season he would land around 20. I was mocked. In 2026 he scored exactly 20. The lesson is not that I got it right. The lesson is that I pointed out, before the result arrived, that finishing above xG is not talent — it is statistical noise waiting to be corrected.

If Zahavi is the case where the numbers told the truth, then 27 June 2026 in Kazan is the case where the numbers went quiet. The night South Korea beat Germany, I looked at the screen and saw every probability lying. I had gone back over Germany's pressing data before the match. They managed a PPDA of just 2.3 in the group stage, their back line kept leaving space behind, and their transitions were exactly one beat slower than their opponent's press. The odds on a South Korea win were 10.0. I wrote that South Korea would win 2-0. Kim Young-gwon broke the deadlock in the 92nd minute. Son Heung-min finished it in the 96th. Germany went out in the group stage for the first time since 2026.

I am not telling this story to boast that I was right. One correct match does not prove a model correct. One match is a sample of size one, and a sample of size one has no statistical value.

In 2026 German football returned during the pandemic. I tracked 81 matches without crowds. Home teams won only 28 percent of them, against 44 percent before social distancing. When the stands are empty, I understood that data needs noise to exist. My model collapsed. I refused to publish for two more rounds while I waited for stability. A programmer colleague pushed me. We rewrote the algorithm and added a variable called "no crowd". In June 2026 my prediction run returned 32 percent.

Then came 9 December 2026. A World Cup quarter-final in Qatar, Brazil against Croatia. Brazil generated 2.3 xG, Croatia 1.2. Brazil led in extra time. I put all my faith in the model. Goalkeeper Dominik Livakovic made 11 saves, a World Cup record since 2026, kept out Rodrygo's penalty, and Croatia advanced 4-2 on penalties. I lost a sum that was not small. I wrote "Why xG Is Not the Truth", then started building a goalkeeping framework — something my model had never had a single variable for.

I tell four stories, not to arrive at a conclusion about football, but to arrive at the night of the empty report.

That night I was trying to answer a transfer question. A club wanted to know whether a player was worth the salary on the table. I had 47 data files. I had minutes played, passes, duels won, kilometres run per match. I did not have three decisive things: medical data, the knee scan, and the actual release clause written into the contract. Without those three, any conclusion about the word "worth" is a guess wearing a spreadsheet.

If I had written it, I would have produced a very readable piece. And there is a good chance I would have been wrong.

This trade pays the people who answer

The industry rewards the people who answer. Nobody pays for a report that says "not enough data". I know that better than anyone, because I am the slowest publisher on my team. But twelve years of counting taught me that my job is not to produce an answer. My job is to produce an answer with a checkable basis. Those two are not the same thing, and the gap between them is where people lose money.

There is a common misunderstanding during a transfer window: people believe more information is better. Rumours do not accumulate into truth. They accumulate into noise, and noise has a deeply unhelpful property — it fills empty cells with a value that looks like data but has no provenance. In my trade, an honest empty cell is worth more than a filled fake one.

Correlation is not causation. A striker who scores 20 goals in League A has not proven he will score 20 in League B. League B may demand 8 percent more running, defend on different principles, and organise set pieces differently. There is one comparison I only use when it genuinely illuminates the problem. In badminton, rally length is a decisive variable, and it varies enormously between tournaments. A player with a high scoring rate in a short-rally event will not hold that rate in a long-rally event. Football is the same, only the unit of measurement changes. Moving a player from a low-running league to a high-running league changes the sample, not the colour of the shirt.

I have read scouting reports that drew conclusions from exactly one season. One season is one sample. When sample size equals one, the margin of error equals infinity.

There is something harder still. Data can make people overconfident, and overconfidence during a transfer window is the fastest way to bankrupt a club. A beautiful report is a painkiller. It makes a board feel reassured while the information is still missing. I have seen it happen, and I do not want to be the one writing the prescription.

There is one simple test I apply to every report before I send it: if you delete every proper noun and every competition name, does the conclusion still stand? If it does, it is a principle. If it does not, it is a story.

What I am watching

I will not tell you who to buy in this transfer window. I will tell you what I am watching: those 61 percent of empty cells. Over the next two weeks I will add medical data and contract terms. When the empty rate falls below 20 percent, I will publish. Not before. You may end up reading my work after everyone else. I accept that.