The Empty Cell in Marseille: The Discipline of Football Data Verification from Ligue 1 to V.League

THE EMPTY CELL IN MARSEILLE Three weeks after the 2026 World Cup final, I reopened the spreadsheet that had travelled with me through the tournament. Twenty columns. One hundred and four matches. And an empty cell sitting on row 1,204. That empty cell held the successful pressing minutes of a midfielder I will not name here. The data source returned a null value. Three options appeared: delete the row, infer from a nearby metric, or type two words into the cell — "insufficient". I chose those two words. Not out of perfectionism. But because I have lived long enough in this trade to watch the life cycle of an invented number. It starts as a harmless guess in a spreadsheet cell, becomes a line in a report, and three months later becomes the basis for a club paying several extra million euros for a player. I am 66 years old, old enough to know a number never tells a story unless you ask it to. That spreadsheet was the product of the summer of 2026, when 48 national teams converged on three North American countries. I followed it from a flat in Marseille's 8th arrondissement, through screens and calls with three colleagues in Montreal, Dallas and Guadalajara. For each match I recorded eighteen metrics. For each metric I noted the source, the extraction timestamp and the sample size. Empty cell number 1,204 was not an accident. It was a rule. MY CAREER BEGAN WITH A NEWSPAPER FOUNDED IN 2026 I entered the profession in 2026, the same year the Independent was founded in London. I was 26, newly arrived from Vietnam to study sports science in France. I had no intention of writing. I simply wanted to understand why a team that ran less could win. Forty years later, I am still answering that question. The tools have changed. When I started, football data meant photocopied sheets: passes, shots, fouls. We counted by hand. We wrote in notebooks. We argued in cafés near the Old Port. By the mid-2010s, football data had become a revenue-generating ecosystem. Opta, StatsBomb, Wyscout, Football Reference — each name a new layer. Then clubs began building their own analysis departments. Then Vietnamese clubs began hiring data staff of their own. Alongside the swelling of data came the swelling of rumour. The two grew together inside the same technical pipeline. A transfer site can pull data from an official provider, blend it with an unsourced social media post, and publish something that reads very convincingly. I have worked as a transfer market administrator in Marseille since 2026. My job is to value players with models. But before a model runs, I must verify the inputs. Most of my working day is spent verifying inputs. SUMMER 2026: I LEARNED TO TRUST SOMETHING NOBODY HAD NAMED — xG In 2026 I was 57. Opta released its first xG tables for Ligue 1. In French analysis rooms at the time, nobody used the phrase "expected goals" in an official meeting. People said "shots", "conversion rate", "chance quality" — but there was no single agreed number. I did not believe it immediately. I hand-recorded 1,204 shots from 20 Ligue 1 clubs in the first half of the 2026-18 season. I logged position, foot, and the type of pass that preceded each shot. Then I compared them with actual goals. The correlation coefficient reached 0.84. That 0.84 was enough to build my own striker valuation dataset. It was not enough to declare xG a truth. Because among those 1,204 shots, thirty-seven cases had xG saying one thing and the goal saying another. Those thirty-seven cases were exactly what I wanted to understand. Colleagues said my reaction was slow. Someone said I had spent six weeks doing what software does in six seconds. True. But software does not know how a shot in the 89th minute, when your team trails 0-1, differs from a shot in the 12th minute when your team leads 3-0. Software returns a number. The reader of that number has to answer the question behind it. From then on I set a rule for everything I write: never cite a new metric without stating the sample size, the confidence interval and the match context. My readers always see verification data alongside the claim, rather than a repeated mantra. WORLD CUP 2026: PPDA IS STILL JUST ONE LETTER Thanks to the dataset built in Marseille, a sports newspaper invited me to contribute during the 2026 World Cup. I was 58, sitting in front of a screen for 64 matches, counting each team's PPDA — the passes allowed per defensive action. In the semi-final between Croatia and England, Croatia allowed England only 8.2 passes per defensive action. England allowed Croatia 12.5. I filed a prediction that Croatia would win through their pressing in extra time. They won 2-1. I did not celebrate. I reopened the spreadsheet and hunted for outliers. Because if I only recorded the correct result, I would be deceiving myself. A correct prediction can come from a good model, or from luck. The only way to tell them apart is to find where the model was wrong. Croatia won a tournament of low PPDA? Then PPDA is still just one letter. Low PPDA said Croatia pressed early. It did not say Croatia had centre-backs quick enough to cover when the midfield pushed up. It did not say Croatia had a goalkeeper with the nerve to face counter-attacks. It did not say Croatia had forwards sharp enough to punish opponents in the space they themselves created. The sentence "Croatia won because of low PPDA" is a convenient shorthand. It omits ninety per cent of the story. A metric only means something when placed beside a counter-metric. That is the second principle I have carried through my career. 2026: EMPTY STANDS ARE THE FINEST LABORATORY FOR A DATA OBSESSIVE In 2026, football restarted after the pandemic. My editor assigned me to the Bundesliga because of my World Cup 2026 data work. I was 60, sitting in Marseille, analysing 81 matches played in empty stadiums during the 2026-20 season. The result: home teams won only 26 per cent of matches. Before the pandemic, that figure was 43 per cent. Empty stands are the finest laboratory for a data obsessive. In a normal laboratory you want to change one variable and hold everything else constant. Football does not grant you that. You cannot separate the crowd from the referee, from the pitch, from the fixture list. But in 2026 the pandemic did it for you: everything stayed the same, only the stands were empty. I wrote a report titled "Empty Stands Kill Home Advantage". In it, I separated home and away metrics for every team and every player. Then something happened I had not foreseen. A Ligue 2 club, Le Havre, used that report to negotiate down the price of a young striker whose pre-pandemic home record was outstanding. Their argument was simple: half his goals came with a home crowd, and a home crowd is not always there. I had produced a tool. Someone else used it their own way. That was my first lesson in the responsibility of a data professional. A table of numbers can become leverage in a negotiation. From then on, I state the limits of every report I write, at the very top, in bold. QATAR 2026: NECESSARY AND SUFFICIENT CONDITIONS My empty-stadium report reached Canal+, so in 2026 they sent me to Qatar for the World Cup. I was 62, sitting in the stands at Al Thumama, and for the first time in my life I could see my own data with my own eyes. The subject I was tracking was Achraf Hakimi. Pundits praised him for 142 sprints and 2.3 chances created per match. Those numbers were correct. But when I dug into positional data, I found something else: the corridor behind Hakimi's back was empty for 34 per cent of match time. In other words, for more than a third of the match, the zone this full-back vacated had nobody covering it. Morocco still stood firm, because their centre-backs ran above 31 km/h and covered in time. I filed a warning note: the fashionable high full-back only holds if the defence has enough speed. The note drew little attention. Against France, the opposition attacked Morocco's right flank relentlessly. A tactical system only functions when necessary and sufficient conditions appear together. A high full-back is the necessary condition. Fast centre-backs are the sufficient condition. Since then I no longer praise a new tactic without examining the compensating variables. In every analysis I list the necessary conditions, the sufficient conditions and the breaking conditions. THE TRAP OF THE EMPTY CELL Back to the August 2026 spreadsheet and empty cell 1,204. There is a paradox in data work: the empty cell is always the cell under pressure to be filled. Readers want an answer. Editors want a long enough piece. Search algorithms want a full page of content. Nobody wants to hear "insufficient". But in football, those two words are often the most accurate answer available. I once saw a transfer report build an entire player profile on eight matches, three of which the player entered in the 78th minute. The sample size was eight. Actual minutes played were under 500. The confidence interval was close to zero. The report was still read. Still cited. Still used as the basis for a proposal. Three reasons explain why empty cells get filled unconsciously. First, narrative bias. The human brain hates a vacuum. When data is missing, it automatically constructs a substitute story. That story is usually more coherent than reality, and therefore more believable. Second, time pressure. An analysis piece can wait three days. A transfer bulletin cannot. When the clock runs, people choose available data over correct data. Third, market incentive. A player is valued higher when more positive data is published. An agent understands this. A journalist understands this. And sometimes both understand it at once. I am not saying every rumour is fabricated. I am saying every rumour needs a sample size. And that sample size belongs in the first line. NGUYỄN QUANG HẢI, PAU FC AND THE PRICE OF A RUMOUR In 2026, Nguyễn Quang Hải moved from Hà Nội FC to Pau FC in Ligue 2. I live in Marseille, more than four hundred kilometres from Pau, and I followed that deal with the eyes of a data professional. It was a clean example of how the European market prices a Southeast Asian player. It was also an example of how rumour works in the opposite direction. Before the contract was signed, dozens of articles linked Quang Hải with clubs in Belgium, the Netherlands, Japan, Korea and Thailand. Each article had a number. None explained where the number came from. What is notable is that most of those numbers were not technically wrong. They simply lacked context. A salary of "around 300,000 euros a year" sounds very specific, until you learn that in Ligue 2 the average first-team salary is much lower, and most income comes from image rights and bonuses. A few years earlier, Nguyễn Công Phượng had moved to Sint-Truiden in Belgium, and Đoàn Văn Hậu to Heerenveen in the Netherlands. Each trip produced a wave of numbers. When the players returned, the numbers vanished, and nobody went back to check whether they had been right. That is why I always ask three questions before using any transfer figure: who is the source, when was it published, and is the number the total contract value or merely the transfer fee. With Vietnamese football, the third question matters most. A deal "worth one million euros" may include the transfer fee, agent fees, three years of salary, a signing bonus and other add-ons. Take only the first figure and drop the rest, and you have manufactured half a truth. Half a truth in football is more dangerous than a lie, because it comes with statistics attached. V.LEAGUE AND THE QUESTION OF WHO OWNS THE DATA I once sat with an analyst working in V.League and heard how domestic clubs have begun hiring data staff. The numbers are still small. The budgets are still thin. But the direction is right, because once a club owns its own data, it no longer depends on numbers supplied by others. That is the difference between a football economy that consumes data and one that produces it. A consuming economy buys metric tables from abroad, reads a few summary lines, and makes transfer decisions. A producing economy records every phase of its own domestic league, builds its own standards, and knows exactly how many kilometres its own player ran in the second half when it was raining. The difference is not money. It is habit. In Vietnam, youth academies have already done part of this work. They record physical metrics, height and speed for each age group. But most of that data is never connected to first-team decisions. When a young player is promoted, his data is usually left behind at the academy. That is the biggest gap in the region's football. Not a shortage of machines. A shortage of pipelines carrying data from where it is produced to where decisions are made. CLUB IPOS AND THE PRESSURE OF THE BALANCE SHEET Meanwhile, in Europe, another trend runs in parallel: clubs listing shares on stock exchanges. When a club goes public, it must publish quarterly financial statements. It must explain to shareholders why revenue fell. It must prove every euro spent had a reason. That sounds reasonable. But football runs on the rhythm of a season, not the rhythm of a quarter. A club may need three years to develop an academy generation, while shareholders grant them three quarters. That pressure reaches the pitch in very concrete ways: a manager must pick the player with resale value rather than the player who fits the system. A sporting director must buy a name rather than a position. A team must win a friendly to protect its media index. Data, in that environment, becomes decoration. People select metrics to justify a decision already made, instead of reading metrics to find a decision. A model bent to serve a conclusion is no longer a model. It is a record of minutes. ESPORTS AND THE SMOOTHING OF INDIVIDUALITY There is another field I track with the same principle: esports. As competitive esports professionalised, teams began hiring analysts. They measure every click, every reaction millisecond, every unit of distance travelled on the map. A player can be judged by the accuracy percentage of each shot. Those metrics are useful. They also create new pressure: the smoothing of individual play. A footballer can win a match with a touch that is suboptimal by the metrics but opens space for a teammate. An esports player can do the same. If coaching relies only on metrics, you produce uniform machines who never err and never create. A click on an esports screen carries the shape of a pass. Both are decisions taken in a very short window. Both can be misread if you look only at the final outcome. Good data is not abundant data. Good data helps you see what you had not seen, even when it contradicts a beautiful conclusion. ACADEMIES, POTENTIAL AND DRESSING-ROOM CHEMISTRY For years, European transfer models have weighted youth heavily. A 19-year-old with good xG is valued above a 28-year-old with the same figure, because people believe he can still improve. That is true in many cases. But models often omit a variable that cannot be measured: dressing-room chemistry. I once saw a club pay over twenty million euros for a young midfielder with every metric in his favour. His sample was 34 matches in a league of lower intensity. After the transfer, he played 11 matches in his first season. No model predicted that. But a scout sitting in the stands could feel it: how he reacted when substituted, how he spoke to former teammates, how he sat alone on the team bus. I am not proposing we abandon models. I am proposing we put them in their proper place. A model answers "what can this player do". It does not answer "what will this player do inside a specific dressing room". Two different questions. And the second decides most of a transfer's success. ONE EMPTY CELL, THREE CONDITIONS Back to empty cell 1,204. If I were asked to draft a standard for football data professionals, I would write only three lines. One: always state the sample size before stating the conclusion. Two: always state the source before stating the number. Three: allow yourself to write the words "insufficient data". These three lines need no expensive software. They need one habit: tolerating the discomfort of not yet having an answer. That discomfort is the hardest thing to teach in this trade. Beginners assume their value lies in answers. In twenty years as a transfer market administrator, I have learned it lies in knowing when not to answer yet. THE NEXT LOOP The 2026 World Cup closed with 104 matches, more than any previous edition. More matches means more data. But more data does not automatically raise the quality of analysis. It only makes errors travel faster. The 2026-27 season has already begun in Europe. Ligue 1, the Premier League, the Bundesliga, Serie A and La Liga are all running. And in Vietnam, V.League 1 is entering its closing stretch. Before every match I still do exactly what I did on my first day: open the spreadsheet, check the source, mark the empty cells. There are matches won on the pitch but lost on the data sheet, and I choose the data sheet. Not because the data sheet is always right. Because the data sheet can be wrong transparently. A missed shot can be hidden by a lucky goal. A misplaced pass can be rescued by a sprint. But if you record both, you know where you stand. What I hope for most this season is not a new model. It is an old habit: stating the sample size before stating the conclusion. Players are variables, the market is a function, but most of my life has been a constant. And that constant, after fifty years, is still the two words sitting in an empty cell: not enough.