When the Algorithm Forgot the Dressing Room

On the fourth Wednesday of November, I sat in the familiar cafe near Nyhavn harbour, opened my phone and read the club's registered matchday squad. His name was not on it. I read it three times, as though reading it again could change the words on the screen. Four months earlier, I had been the one who placed his dataset on the meeting table, pointed at the green columns, and said he was the missing piece. He ran an average of 11.8 kilometres per match. He recovered the ball 6.2 times per match. Those numbers were so beautiful that I forgot to ask one very simple question: who would he actually talk to in the dressing room?

I remember that morning, cold and clear, the kind of Danish cold that makes you believe everything can be measured, even homesickness.

It happened during the summer transfer window. A Superliga club was looking for a defensive midfielder. The budget was tight, the wage bill had hit its ceiling, and the coaching staff wanted someone who could play immediately, without a long adaptation period. I was tasked with building a screening model from data across leagues in Africa, South America and Eastern Europe. My model placed heavy weight on two variables: distance covered per match and ball recoveries in the opponent's half. To me at the time, those were the two indicators that most directly reflected the intensity and will of a defensive midfielder.

He was 22 years old, 1.86 metres tall, from an academy in Dakar. Across his last 34 domestic league matches, he averaged 11.8 km covered per game, 1.3 km above the average for midfielders in his position in the Superliga. His 6.2 ball recoveries per match placed him in the top 5 percent of the entire dataset I had collected. His duel success rate was 58 percent. Placed side by side, those numbers painted an almost perfect portrait: a tireless sweeping machine, fierce in the tackle, always appearing exactly where the ball fell.

I presented in the scouting meeting. In the room sat a veteran scout I deeply respected, nearly twenty years older in the profession than me. He listened as I talked about the model, the weights, the reliability of the sample. Then he asked a question I considered an obstacle at the time: “Have you watched him play in a real match, under real pressure?” I answered that I had watched seven matches, and that the model had accounted for all of them. He simply nodded and said nothing more. I thought I had won that argument. I was wrong.

What I want to say is not that the model was wrong. The model was right. What it could not calculate lay somewhere else entirely.

The contract was signed in early July. The fee was modest, with bonuses tied to appearances, the kind of structure Nordic clubs favour: low risk, high potential, and a sensible release clause should the player shine. In August he played three friendlies. The data still looked good. He ran, he tackled, he recovered the ball. But there was one detail the spreadsheet did not display: he barely spoke to his teammates in training.

Based on my experience covering matches and training sessions over many years, I began to notice the things that lie outside the spreadsheet. In one tactical session, the coach asked the whole team to move as a block and hold the distance between lines. He ran a great deal, but he ran with the instinct of a lone player: always toward the ball, rarely toward the position his teammates needed him to fill. In one defensive transition, when the left back pushed high, he did not drop to cover. He stood in the middle of the pitch, eyes fixed on the ball. His distance covered in that session was still outstanding. But the space behind his back was something no metric could measure.

That was when I understood something I had known but never been willing to admit. Screening data measures individual capacity, but football is an interdependent system in which a player's true value is only defined when he is placed beside the other ten. Distance covered told me he worked hard. It did not tell me he ran to the right place. Ball recoveries told me he won the ball. It did not tell me he knew whom to pass to once he had won it.

I rewound the footage of his seven matches in Senegal. This time I did not look at him. I looked at the spaces around him: whether his teammates had to compensate for him, whether he dragged the whole shape with him, what he did after each phase of play. It turned out that in his old league he played in a system that allowed him to hunt the ball freely, and his teammates had grown used to covering for that freedom. In Denmark, where positional discipline is the foundation of everything, that freedom became a hole. The viewer sees the goal; I see the sequence of events before the goal — and this time, that sequence led me to an uncomfortable conclusion about my own work.

I had once written that the same metric, viewed through two different sporting cultures, reveals the cultural layer hidden inside it. In Japan, where I grew up, collective discipline is taught from childhood, and a player like him would have been moulded from his very first sessions. In Denmark, people respect individuality more, but they also demand the ability to position yourself within a system. He came from a third football culture, one that prizes individual skill above structure. That difference was not in my model, and that is why I failed to anticipate it.

That was not his fault. It was the fault of the person reading the data.

In most transfer models I have worked with, there is a systematic bias: we overvalue young potential and undervalue dressing-room chemistry. A 22-year-old with beautiful numbers is always more attractive than a 29-year-old with ordinary numbers who understands his role within a collective. The model measures the first. It is blind to the second, because integration, a shared language, and the ability to endure the loneliness of being far from home are not in any database I have access to.

I thought about myself. I left Japan for Copenhagen at 22. I too was once someone who ran a lot but did not know how to talk to anyone. Fortunately, I did not carry the burden of a contract. He did. And PPDA cannot measure the heart, but it points to where the heart is beating — except that this time, I read the wrong place.

Four months after signing, his name disappeared from the registered list. No official announcement, no press conference, just a quiet absence on a November afternoon. He trained separately with a fitness coach. I heard he still ran his full 11.8 km every session. That distance, sadly, now led nowhere.

In telling this story, I do not want it to sound like an indictment of data. Data is not guilty. The way I read it is guilty. I turned a screening model into a verdict, and I forgot that behind every data row is a person with a childhood, a family, a fear I will never touch. I do not believe in luck; I believe in what luck conceals — and what was concealed this time was a young man trying to learn a new language while the whole world looked only at his columns of numbers.

There is a line I wrote in my journal that night, under the yellow light of a small apartment: “Data only recounts the past, while football lives in the future.” I still believe that. But I also began to understand that if data can only recount the past, then I must use my own eyes to look into the future — into the gaps the spreadsheet does not draw. And if data cannot measure the heart, at least it must point me toward where to look.

In this transfer window, I have changed how I work. Whenever a name rises out of the model, I spend at least three evenings watching him play without looking at the analytics screen. I watch how he reacts when a teammate makes a mistake, when the referee gives a wrong decision, when the team falls behind. Those moments are not in the model, but they tell me more than any metric.

People ask whether I regret persuading the club to sign him. I do not regret it. If I had not been wrong, I would never have seen the gap in the way I read data. Scouting will not progress if people like me keep presenting perfect but soulless models.

What I learned, after all of it, is an uncomfortable humility. I am a data monk, and I thought I could tell every story with cells of numbers. But when the match ends and the stadium empties, only one thing remains that I have still not learned to measure: the heart of a person standing alone, far from home, who can run any number of kilometres and still never reach the place where someone is waiting for him.