xG, PPDA and Empty Stadiums: Four Data Tests in Major Tournaments

On the night of the 2026 World Cup semi-final at Luzhniki Stadium, I sat in the stands with a notebook and a pencil. I was not there to cheer for Croatia or England. I was there to count. Croatia allowed their opponent 8.2 passes per defensive action; England allowed Croatia 12.5. I wrote the number down, circled it, and pencilled a prediction: Croatia would win through pressing in extra time. They won 2-1. When the final whistle blew, I did not shout. I reopened my spreadsheet to look for outliers. A win can be explained by dozens of reasons, and the analyst's job is to separate the reasons that actually produced the result. I was born in Vietnam, work in Marseille, and spent most of my career as a transfer-market administrator. In 2026, when Opta first published xG tables for Ligue 1, I was 57 and in no hurry to believe them. I hand-recorded 1,204 shots from 20 clubs over the first half of the 2026-18 season and compared them with actual goals. The correlation coefficient reached 0.84. Colleagues said my reaction was slow. But I need verification before use. In the summer of 2026, I learned to trust something no one had yet named: xG. My method has not changed much since. I ask a question, present the raw data, set it against context, and only then state a principle. I never cite a new metric without stating the sample size, confidence interval, and match context. When a prediction of mine is right, I quietly hunt for exceptions. When one is wrong, I extract a principle from the failure and apply it to a new setting. A major tournament is the harshest test of this method, because national-team emotion compresses everything, and people usually want one single number to believe in. The four tests below are how I keep myself out of that trap. The first test is xG in Ligue 1 in 2026-18. What I learned was not the 0.84 figure but the fact that I had to verify it myself. A model is only trustworthy when you understand what sample it was built from. When I use xG to value a striker, I always attach the verification data; I never use it as a mantra. The same player can post a high xG at home and a low one away; if you do not separate the two contexts, the valuation table will be wrong. In transfer valuation, I also learned that models overrate young potential and underrate dressing-room chemistry. A 19-year-old with a fine xG can fail because he does not fit the dressing room, and no metric measures that. The second test is PPDA at the 2026 World Cup. I tracked 64 matches and counted each team's PPDA. Croatia went deep with impressive pressing intensity, but if you look only at the low PPDA of a semi-finalist, you risk a false conclusion. Croatia won a tournament of low PPDA? Then PPDA is only a letter. It describes pressing intensity, not defensive quality, not the ability to convert chances. To understand Croatia, I had to combine PPDA with distance covered, successful presses, and expected goals. A metric means something only beside other metrics. The third test is empty stadiums in the Bundesliga in 2026-20. My editor assigned me to cover football's restart after the pandemic. I was 60, sitting in Marseille, analysing 81 matches without crowds. Home teams won only 26% of the time, against 43% before the pandemic. An empty stadium is the finest laboratory for a data obsessive. It showed me where home advantage lives: a large part of it comes from the crowd, from pressure on referees, from psychological rhythm, not merely from a familiar pitch. A Ligue 2 club, Le Havre, used this report to lower the price for a young striker who had shone at home. I do not know whether they were right, but the data entered a real decision. The fourth test is the 2026 World Cup. When pundits praised Achraf Hakimi for 142 sprints and 2.3 chances created per match, I dug into the data and found the corridor behind him was empty for 34% of the time. Morocco stayed safe because their centre-backs ran above 31 km/h. I wrote a note warning that the attacking full-back trend only holds if the defence is fast enough. In the match against France, the opponent attacked relentlessly down Morocco's right. The necessary condition held; the sufficient one did not. This is a lesson I repeat often: a system works only when every compensating variable is accounted for. What the four tests share is something major-tournament fans often forget: correlation is not causation. A beautiful metric can be the consequence of something else. A team that presses better does not necessarily win if it cannot convert chances. A full-back who sprints often is not necessarily better if the defence behind him cannot cover the gap. Home advantage does not necessarily come from the pitch if empty stands make it vanish. I am 66, old enough to know a number never tells a story unless you ask it to. For every conclusion, I force myself to write at least three competing explanations before choosing one. This is slow, and in a major-tournament season, slowness is treated as weakness. But I have seen too many models collapse because people rushed to turn a number into truth. In another corner, I have seen club IPO deals turn fan emotion into money. When financial-reporting pressure presses down on sporting decisions, a beautiful metric can be used to sell tickets and shares, not to understand a match. That is why I always separate analysis from marketing. There is another temptation: turning every comment into a report. I avoid it by always starting from an image, a player, a concrete moment, and only then moving to tables. Players are variables, the market is a function, but most of my life has been a constant. That constant is the discipline of verification. I do not worship a single number; if a team wins with low PPDA, then PPDA is only a letter, not yet a truth. Even in esports, where professionalisation is turning players into assembly-line products, I see the same mechanism: individual flair is sanded smooth in digital training. A single mouse click on an esports screen carries the shape of a football pass. At the next major tournament, I will still count, still take notes, still hunt for exceptions after every win. What I want readers to carry away is not a new metric but a question: what sample was this number built from, and what does it leave out? Some matches are won on the pitch but lost on the data sheet — I choose the data sheet, not to deny the win, but to understand it deeply enough for next time.