Empty Data, Full Conclusions: The Biggest Hole in Modern Sport

Minute 63, the home side switched shape from a 4-2-3-1 to a 3-4-2-1. Seven minutes later they led by two, and within half an hour social media had already named the phenomenon: the tactical revolution of the matchweek. By Wednesday I reopened the pressing heatmap and the chance-creation table to verify it. The most important column — quality chances created after the shape change — was empty. Not zero. Empty. Nobody had filled it in. The headline, meanwhile, had been finished long ago.

That moment is the biggest lesson of my fifteen years in this trade. It is not about one match; it is about the distance between a conclusion and the data holding it up. In football, that distance is usually filled with inspiration. In esports, it is filled with speed. In both places, it produces something more dangerous than a false story: a story that is formally correct and hollow inside.

CONTEXT: WHEN EVERYTHING IS MEASURABLE, EVERYONE IS FORCED TO CONCLUDE

Fifteen years ago, a match analysis could live on the naked eye. You watched, you remembered, you retold. Today, a single football match emits thousands of data points: passes, line distances, expected-goal figures, pressing actions per defensive sequence. Esports goes further — a single game can be logged down to the percentage of resources, the second of cooldown, the single gold transaction. On the surface, this is an analyst's paradise. But the more columns of data exist, the more columns can be empty, and the more columns are empty, the more people want to fill them with guesswork.

Three sources of pressure arrive at once. The first is publishing rhythm: a match ends at 10 p.m. and the report has to be out before breakfast. The second is reader expectation — they watched the game, they want an explanation, not a sentence saying "not enough data." The third is the market. Live match data is sold to betting companies, and that money rewards speed, not caution. I have sat beside the people running data feeds, and what chilled me was not the numbers but the gap between the moment an event happens on the pitch and the moment it becomes money. That gap is so short there is no room left for verification.

In esports the pressure is stronger because the cycle moves faster. A balance update can overturn the champion standings within a week, and every time it does, the analysis world gets another chance to say very certain things about a very small sample. Football has no patch, but it has an equivalent: systemic shocks that change how an entire league operates. Empty stadiums were the largest patch in Premier League history, and we missed the lesson — not because it was hard to understand, but because while it was happening most of us were busy describing it with feelings.

THE CORE: HOW A CONCLUSION GETS MANUFACTURED

Let us reconstruct the process. A team wins three games in a row after changing how it plays. Step one, the writer picks the phenomenon. Step two, the writer finds one metric that rose across those three games — and there is always at least one metric rising, because three games is a random enough sample for anything to fluctuate. Step three, the writer labels that metric with a term that sounds systematic: "high press," "vertical control," "deliberate switch of play." The conclusion is born before the question is ever tested.

The crux lies elsewhere: across those three games, how many situations actually repeated? Three games are not a trend; three games are a band of noise shaped like a trend. That is a line I have written again and again over the years, and it still holds in both sports I follow. Ganking from the left flank: the 4,200-word lesson I wrote in 2026 still holds for modern football — not because the technical content survives, but because the principle behind it does: to call a behaviour a tactic, you must prove it repeats under pressure, not merely that it appeared once and won.

In 2026, at the MSI held in Brazil, the unusual jungle style of GAM Esports under Lê Duy Khánh — known as Levi — once built a 7,000-gold lead over TSM by minute 22. I stayed up all night dissecting fourteen ganks, and what I learned was not "junglers must be aggressive." What I learned was that this style only had value because it was tied to a specific mechanism — the timing of resources on the map, how opponents reacted, and the risk threshold the team accepted. Copy the behaviour without the mechanism and you will fail exactly the same way in a different context.

In football, the cycle of learning a new system is far longer than one matchweek. A team needs six to ten games to get comfortable with a new pressing structure, and during that window results are often worse than the true level. The paradox is that the worst stretch is precisely when the system is being installed. The hasty analyst reads the installation phase as a failure phase, then weeks later, when the team starts winning, reads it as a miraculous revival. Both readings are wrong, because both ignore the only variable that matters: the time a system needs to mature.

In esports that cycle is compressed. A team may have to switch to a new champion pool within days, and the price of switching wrongly is losing the whole phase. This is why I do not trust power rankings published after one week of play. Not because they are wrong on the numbers, but because they are built on a sample too small to separate skill from luck. The meta is not something to chase; it is something to get ahead of — a lesson from the transfer market: good buyers do not buy what is winning, good buyers buy what the new rules are about to favour.

The same error appears where few expect it: officiating. There is no empty data column here, but there is a larger gap — the gap between the law and how the law is read. Referee-assistance technology was introduced as a machine for eliminating error, yet its central clause — "clear and obvious error" — is an ambiguous sentence. Clear to whom, obvious to what degree, and why is the same collision read differently in two matchweeks? The space for subjective judgement in referee-assistance technology is larger than people think, because the intervention threshold is itself a choice, not a measurement. Every time a referee decides not to intervene, that is not the absence of judgement; it is a judgement.

I bring this detail in because it is the perfect example of the central problem. We have cameras, drawn lines, slow motion down to the millimetre, and still a gap that data cannot fill. The honest professional says: this part is judgement, and I chose this, for this reason. The hasty professional says: the data proves it. But data proves nothing when the intervention threshold is set by human beings.

Another example comes from the market. Live data supplied to betting companies is the darkest side effect of sport's digitisation — not because the data is bad, but because it creates a system in which the uncertainty of a match is turned into a product. A shot off the post, a red card in the 89th minute, an injury nobody anticipated — all become variables in a model, and the model does not need to be right; it only needs to be faster than the bettor. When data becomes a commodity traded by the second, the incentive to verify disappears. Nobody pays for an empty column.

And this is where I want to speak plainly about my own trade. We have built an ecosystem in which saying "I don't know" is treated as weakness. But in every serious analytical process — from a club's data room to a simple information pipeline — there is a mandatory rule: when the input is empty, the output must be "insufficient information to assess." Not a guess. Not filling the gap with a story. Exactly a dash, with a reason. I once watched an internal analysis collapse completely because the input data layer was empty, and instead of stopping, people simply kept writing — every conclusion block correctly formatted, every heading present, and nothing underneath holding it up. That report looked perfect. It lacked one thing only: the truth.

One more example shows how a small sample can mislead in both directions. At the 2026 World Cup finals, only 3 of 28 penalty shootout kicks were taken as chips, yet their success rate was 100%, against 78% for conventional strikes. Read the number alone and you conclude the chip is the optimal choice. But a sample of three kicks is not enough to conclude anything about a rate. What it actually shows is a different mechanism: players took the chip only when they had already read the situation and held a psychological edge. What decided success was not the kick but the conditions under which the kick was chosen.

That rule sounds dry, but it is the foundation of everything I believe. An analysis is worth reading only when it delivers at least one thing the reader did not know — not a new feeling, but new understanding. If the input holds nothing new, the output cannot hold anything new either, and the only honest move is to say so.

THE CONTRARIAN PART: DON'T USE CAUTION AS AN EXCUSE TO NEVER SAY ANYTHING

There is a trap on the opposite side, and it is no less dangerous. If every small sample is dismissed, we will miss exactly the most important moments — because every big change begins with a small sample. A new champion appearing for the first time at a regional event, a strange shape used for one half, a young player scoring in his first three games: these are signals, and real signals often arrive before the data is thick enough.

So how do you tell them apart? With three questions, and I use exactly these three for both football and esports. First, does this phenomenon have an explainable mechanism, or only a result? Second, does it repeat under higher pressure, or does it appear only against weak opponents? Third, if the rules of the game change, does it still stand? A signal that passes all three questions is worth an investment of belief, even on a thin sample. A signal that passes only the question about results is worth a footnote, not a headline.

The problem is that most sports content today is written for the moment, not for the mechanism. We reward the people who speak with certainty, because certainty sells better than caution. But over the long run, the people who dare to say "I don't know this part" are the ones who keep their credibility longest. Based on my experience tracking matches, a colleague once told me I looked at an athlete like a metric, not a person. That was true, and it forced me to change. But it also taught me the reverse: empathy cannot replace verification, and verification does not exempt you from empathy. A correct number without a heart is cold. A good story without data is hollow. Both are defective products.

The last point, and the hardest: the beneficiaries of hasty conclusions are not only the writers. Clubs benefit when an analysis says they are improving. Sponsors benefit when their story is told as a revolution. Betting markets benefit when people believe everything can be predicted. Hasty conclusions are not a random mistake; they are a system with motives. A transfer is not a transaction, it is a draft: a way of seeing the future through the meta lens — and in a draft, you do not buy what is winning, you buy what will win when the rules change.

CONCLUSION: WHAT IS WORTH WRITING NEXT

What I carry away from every rereading of my own data is not a formula but a habit: before writing a conclusion, check whether the data column actually holds anything. If it does not, write exactly two words — not enough — and save the space for next time. Readers do not need us to be certain. They need us to be trustworthy. And trustworthiness, in sport as in every analytical trade, is built from one thing only: saying correctly what you know, and saying honestly what you do not. Next time you see a bold headline about a tactical revolution after three games, ask yourself whether the most important data column has been filled in. If it has not, perhaps the most interesting thing has yet to happen — and that is exactly when the real story begins.