When the Analysis System Returns Zero

The report ran to nine sections. It had comparison tables, a six-row risk matrix, and even a transmission diagram drawn with arrows made of characters. Every cell read the same: insufficient information. I read it at two in the morning in Milan, with a third coffee long gone cold, and what kept me awake was not what the report contained. It was what the report did not contain: not a single club named, not a single player, not a coach, not a competition, not a match, not a minute of football. The only populated field in the entire document was a two-word label: football. I read it four times, not because it was hard to follow, but because I wanted to be certain my eyes had not skipped something. And I realised I had skipped exactly one thing: the report itself was its own subject. To understand how a document that long can hold a zero, you need to know how it is produced. The system runs in two stages. The first stage takes a raw source — an article, a news item, a match file — and breaks it down into discrete information points: who did what, when, where, and with what consequence. The second stage takes that list of information points and puts nine professional questions on top of it. Those nine questions run from tactics and technique, through club finance and the transfer market, through results and the opinion cycle, through league landscape and team positioning, through rules and compliance, through management and the dressing room, through the risk profile, through media narrative and expectation, all the way to the transmission chain of an entire industry. That is the framework I and many colleagues use to read a match or a transfer. When the first stage returns an empty list, the second stage faces three choices. One: go back to the source and check whether retrieval failed. Two: return an empty result and state plainly that nothing can be assessed. Three: fill the gap with what sounds plausible — a real club, a real player, a real transfer, assembled into a story that reads. The third choice is by far the most common, because it looks useful. It produces something publishable, shareable, sellable. The other two produce something that looks like failure. And in an industry where there is a match every night and every match needs commentary, failure is the one thing nobody wants to sign. The report I read that night took the second choice. It did not salvage, embellish or fill. It stood still in front of the gap and described the gap. That makes it one of the most honest pieces of analysis I have read in years. Start with pure logic and ignore the circumstances. To assess a risk you need three things at once: a named subject, a stated event, and a time horizon. Without a subject, risk has nothing to attach to. Without an event, there is nothing to weigh. Without a time horizon, there is nothing against which to compare probability with impact. In the entire input of that report, all three were absent. So every risk cell had to be empty. Not because the writer was lazy, but because filling a risk cell at that moment would require inventing a subject. And once a subject is invented, the rest of the report arranges itself around that first lie — like a high defensive line arranged around a player who does not exist. Roughly thirty per cent of my work over the years has been reading reports like that, but in the opposite state: crammed with subjects, crammed with numbers, crammed with conclusions, and my job is to find where the filling happened. At this point we need to separate two completely different kinds of zero, and confusing them is the most common error in this profession. The first kind is the zero of a broken system. Data does not arrive, the parser returns empty, the pipeline is severed. That is a technical fault, and the fix is to repair the pipeline. The second kind is the zero of reality. The match genuinely produced no signal. The player genuinely did not touch the ball in that zone. The team genuinely did not press. That is not a fault, it is information — usually the most important information in the whole match. Seen from the outside, both produce the same number. Inside, they are opposite stories. A full-back finishes a match with zero tackles. That number can mean he defended badly, was beaten repeatedly and arrived too late. It can also mean nobody dared attack his flank, so he had nothing to do. The same zero, two opposite conclusions. Ask what the system has hidden before you judge a defender. In my personal database there is a separate file for cases like this. I call it the zero file, and it weighs far more than its name suggests. In 2026, at thirty-six, I published a six-thousand-word analysis of Atalanta under Gian Piero Gasperini. I used GPS data from thirty-seven Serie A matches to demonstrate something I believed was new: Robin Gosens was not a conventional full-back but a number ten operating in the wide zone. The number I leaned on was specific. An average of 21.4 receptions inside the penalty area per match — more than the team's leading striker. I drew heat maps, built pressure charts, added arrows for movement. The piece was republished by an Italian specialist outlet and opened the door to regular work. That is how I obtained press credentials for the 2026 World Cup. But there was one detail I overlooked for three months. It took me three months to realise I had been reading that position wrong. What I measured was not where Gosens stood when the team attacked, but where he stood once the team had won the ball in the opponent's half. The difference sounds small. It is not small. It reverses the role: he was not creating space by moving inside, he was occupying space already created by two central midfielders stretching the opposing back line. Numbers do not lie, but they do not tell the whole story either. In the summer of 2026, when football stopped, I was thirty-nine and slid into a long anxious stretch. For six months I wrote nothing. Instead I sat in a room reviewing 4,500 wide-attacking situations from Serie A between the 2026 and 2026 seasons, hand-drawing thirty-eight pressure diagrams in pencil on A3 paper. 4,500 situations, and one detail changed the way I read a match entirely. By June 2026, with the European Championship underway and me just past forty, I spotted a pattern that had never appeared in my database: Italy's central midfielders Nicolò Barella and Marco Verratti were generating 14.7 passes into dangerous areas per match through triangular movement. Not long balls, not line-breaking passes. A small triangle, repeated, in the same zone, until it became almost a mechanical habit. Three months of isolation, 4,500 sequences, and an answer simple enough to be startling. In July 2026 I was in Moscow for the World Cup semi-final between France and Belgium. I took careful notes. Coach Didier Deschamps dropped the defensive block to an average of just 24.8 metres, and Blaise Matuidi tucked inside to block the passing lane into Kevin De Bruyne. I wrote in detail about space, about the distances between lines, about how Belgium lost their bearings after the break. My piece sank. A colleague wrote only about Vincent Kompany's tears after the defeat, and his piece was shared six times as widely. He had an image. I had a diagram. Emotion is not data noise; it is data that has not yet been decoded. That 24.8-metre distance was real, and it explains why Belgium could not play out. But it does not explain why people read my piece to the end and forgot it, while they remembered his. It took me two more years to understand that the two questions do not exclude each other, and that an analyst who cannot answer the second will never get the chance to answer the first in front of enough listeners. Back to the gap. There is one area of football where zeros appear so often that we have grown used to them and stopped seeing them: refereeing and video assistance. Take a specific situation. A goal is disallowed after the system draws an offside line less than a millimetre thick. Technically, the system returned the correct answer: the player was offside. Informationally, the system returned a zero: it told us nothing about whether that player gained an advantage, whether it was deliberate, whether it broke any tactical intention. That is the limit of the number in its purest form: a measurement accurate to the millimetre, used to answer a question that is not a measurement question. The millimetre offside line is slowly killing attacking instinct in a way that is very hard to see. Players learn to slow by half a step, learn to keep a safe distance from the last defender, learn to abandon the runs they used to make on instinct. Nobody orders them to. No coach gathers the squad and says: be less bold. It is the natural outcome of a strict system of reward and punishment. A goal disallowed by half a foot is a correct number and a real loss. Both exist at once, and the industry chooses to publish only the first half. There is another kind of zero, more dangerous because it is not empty but looks full. That is the case with cup fairytales. An amateur club reaches the final of a major competition. The statistical system immediately returns a string of positive data: wins, goals, saves, an unbeaten run. Reading that, one could conclude that some system is working. But when I strip back each of that club's matches, a very different structure appears. Most of the wins came from a favourable draw that placed them in an easier bracket, plus one explosive night when every shot found the top corner. That is not evidence of a system. That is evidence of a good draw and one night when everything worked. This confusion appears constantly in my profession, and it fits how humans read data. We look for patterns in noise, and when we find a beautiful pattern we do not want to reopen the question of whether it has any basis. So what makes that empty report rare? Because it runs against the incentive structure of an entire industry. In analysis, the reward does not come from being right. The reward comes from being certain. A piece with a clear argument, a decisive conclusion and a specific prediction will be shared more widely than a piece saying there is not enough data to conclude. Readers want certainty, and the market supplies certainty. Certainty is a commodity with stable demand. That is the execution blind spot. We have taught machines to produce text, but we have not taught them — or ourselves — to stay silent when silence is required. A system that returns honest information can be judged inferior to one stuffed with plausible fabrications, purely because it has nothing to publish. I have seen the consequences of this mechanism many times. An article saying a club is interested in a midfielder, with three unnamed sources, goes on the front page. Another saying there has been no contact at all, based on direct checks, is pushed to the bottom. Accuracy is not the variable that decides how far information travels. The feeling of certainty is. There is a way to test the strength of this mechanism that I use fairly often. Put two descriptions of the same event side by side: one written with a conclusion, one written with the question still open. Give both to non-specialist readers. In most cases, the one with the conclusion is rated better, clearer, more credible — even when its content is more often wrong. To me, that measurement shows the problem is not the tool. Machines do not spontaneously invent data. We ask them to. We ask them to by rewarding only the products that sound certain. So what should be done with gaps like these, in practical terms? Three things, and I say this from experience of getting all three wrong myself. The first is to label the zero. Whenever a data field is empty, state clearly whether it is empty because the system failed or because reality produced no signal. The two demand different actions. The first demands re-running the pipeline. The second demands rewriting the hypothesis. The second is to keep the analytical framework intact even when every cell is empty. This is the easiest place to go wrong. When every answer is that nothing can be assessed, the natural reflex is to dismantle the framework and write something gentler. But keeping an empty framework is more useful: it shows exactly which kind of information is missing, and therefore points to what to do next. The third is to accept the social cost of returning a zero. This one is hardest and has no technical solution. I come back to the 2026 Gosens lesson, because it connects directly. If I had had the courage then to publish an early version of the piece with a cell stating clearly that the player's role could not yet be concluded, I might not have spent three months walking in the wrong direction. But if I had published a cell like that, the piece might not have been republished, and I might not have had press credentials to go to Moscow in July 2026. That is a very concrete price of honesty in analysis, and I do not want to pretend it is cheap. Sometimes saying you do not know yet costs you an opportunity, and that loss is real, not an abstract trade-off. But if you say you know when you do not, what you are building is not credibility. It is a debt that comes due on a day you did not choose. A heat map shows position; an intention map shows thought. The second does not exist in any data package I have ever bought. That is why I begin every analysis by asking myself: if I had never watched this match, what would I invent? The question sounds pointless, but it has a concrete effect. It forces me to see the easiest story first, the story anyone would write, so that I can set it aside and look for the rest. And if you are wondering why I have spent an entire article on a report that contains nothing, the answer lies in that nine-dimension framework itself. That framework is designed to answer every question about football. It is not designed to say that it does not know. But on that particular run, it said it did not know — nine times, across nine different dimensions, with the same level of honesty. To me, that is the most credible output an analytical system can produce, because it proves the system is capable of refusing itself. Systems without that capability will always return an answer. And always returning an answer is the precise definition of a system that cannot be trusted. The next match is this weekend, and I will sit in front of the screen again with a blank sheet of A3. On it I will write three lines in advance: what I think I know, what I think I have measured, and what I cannot yet assess. The third line is the longest, and each time it grows a little longer. The test for next time is not whether I predict the score correctly. It is whether I notice, before the match ends, that a cell in my framework is empty — and whether I have the nerve to leave it empty. Next time you read an analysis of your team, look for the cell that says the author cannot yet conclude. If there is no such cell in the whole piece, what you are reading is not analysis. It is a product designed to look like analysis, and it is selling you a certainty its author does not have.