The Transfer Window Taught Me to Say: Not Enough Data

In more than two decades of working with numbers, I have learned something more uncomfortable than any formula: there are days when data refuses to speak, and the analyst's proper job is to stay silent rather than plug that gap with a conclusion that simply sounds good. One evening I sat in front of the screen, rebuilt every pass of a match, poured the numbers into the model, and got back a nearly empty table. Not enough shots for expected goals to say anything. Not enough matches for a pressing metric to tell a real system apart from a lucky night. I typed a line into my notebook: "Sample too small, no conclusion." The next morning I still published the piece, but it was a piece explaining why no conclusion should be drawn. The first xG table I wrote by hand on a bus, back when nobody called it data. I counted every shot, every position, every angle, only to answer one simple question: does this team really play better than its results. That experience taught me that data is first of all a defence against emotion, including my own. The transfer window is the season when data gaps appear most densely, and also the season when people tolerate them least. Every day dozens of names are attached to dozens of clubs, each rumour pushed along by a line about a "source close to the deal", and fans drown in a stream of information with no labels. A friend who works as an intermediary once messaged me: "A club asked about him, publish it for me." I asked exactly three questions back: what are the terms, what is the fee, who confirms it. There were no answers. That rumour died the way it was born. In this trade I set my own reliability ranking. News from an official club statement is tier one. News from a registered release clause is tier two, because that is a number that can be verified. News from an agent is tier three, because an agent has a very clear motive to inflate the price. And news from an "unnamed source close to the deal" is the lowest tier, because it cannot be wrong, cannot be right, it exists only to fill a hole. A decent transfer article has to state which tier it stands on. But what keeps me up is not the rumour; it is the conclusions built on a sample that is far too small. A striker scores three goals in four games and is instantly called "the signing of the century". A defender makes two mistakes in two rounds and is instantly labelled "finished". I look at the model and see the opposite: those three goals may come from shots whose scoring probability was only a few percent, meaning he is scoring through luck more than skill, and every lucky run ends one day. Conversely, those two mistakes may come from a defensive system stretched out of shape, where the individual is merely the last man to carry the blame for the whole unit. That is why I always publish the sample size alongside. One match says nothing. Five matches begin to hint. A full season is enough for me to bet. Readers have the right to know how much data I am leaning on to say a sentence, and I have a duty to tell them when that number is not enough. The world saw Croatia as an underdog; I saw them as a chain of coefficients nobody had dared to exploit. At the 2026 World Cup, while everyone was praising the possession sides, I counted Croatia's pressing metric under Zlatko Dalić, with Luka Modrić in midfield, in the match against Argentina and found it lower than even the teams regarded as kings of the ball. They did not keep the ball for the sake of keeping it; they pressed the opponent the moment they lost it, and pressed in an organised way. I wrote a piece predicting they would go deep, and was laughed at. But I dared to write because the sample was large enough, and because I knew exactly what I was measuring. I once also saw in a twenty-year-old winger at Song Lam Nghe An, Phan Văn Đức, an expected-goals-per-match figure higher than the foreign strikers, even though he had only scored a handful of goals. I wrote that he would become a national-team pillar within three years, and was mocked for being lost in numbers. But the sample was already enough, and time answered for me. In 2026 the stands were empty, but every pass still fell into the model's cell, and I understood that data never keeps company with a pandemic. When the leagues paused, many turned to light entertainment writing to pass the days. I dug back through ten years of domestic-league data and found a pattern: clubs that changed the person at the top mid-season tended to slump markedly in the following run of matches, because governance turmoil seeps down into the dressing room. A club director called me afterwards, thanking me for helping him delay a hasty decision. Old data, placed in the right context, can save a new decision. In the transfer window I read one type of contract with special care: a loan with an obligation to buy. On the surface it is a gentle deal, but the word "obligation" turns it into a debt with a timer. The small club takes the player, pays his wages, gives him a chance, and at the end of the season must buy him at a number fixed in advance, while the power to set the price sat with the big club from the start. I have seen clubs forced to sell a cornerstone just to balance a purchase they never truly wanted. The number on the balance sheet is always more honest than the promise on the coaching bench. I also raise a big question over deals built around a player returning from an anterior cruciate ligament injury. My data shows this group tends to peak a season later than expected, and the hardest thing to measure is not the knee but the head. The fear of re-injury makes a player hesitate in the decisive challenges, and my model has no column in which to enter that fear. I can only note it down, as a variable I admit I cannot yet measure. Even VAR, which seems to bring clarity, is another source of noise. It does not remove the argument; it moves the argument from the pitch to the review room and pushes it into the grey zone of the law. An offside by a few centimetres, an accidental handball, is enough to turn a clear result into a debate longer than the match itself. When data arrives late and arrives through human judgement, I learn to wait for the official announcement rather than trust the first frame. The point I want to argue against the crowd sits here: most wrong conclusions in a transfer window do not come from a lack of data, but from a fear of the gap. People feel uneasy having to say "I do not know", so they fill it with a correlation and call it causation. A team spends a lot and then wins, and people say money buys victories. A player changes his shirt number and then shines, and people say the number suits his fortune. Correlation is merely two things appearing together; causation is one thing producing the other, and to prove the second I need more than a coincidence. My job is not to always have an answer. My job is to know when the answer does not yet exist, and to say so clearly. Fans watch the move; I watch 22 numbers in motion, and wait patiently for them to tell a different story. Sometimes that story is: there is nothing to tell yet. My model does not cry, does not celebrate, but after every match it owes me a lesson. The biggest lesson of this transfer window is not which name will land where, but whether the writer has the courage to say "not enough data" before the market pins a label on him. The transfer window truly ends only when the last number is verified; the rest is just noise.