The Silent Blank: The Biggest Trap in Sports Analytics

The Silent Blank: The Biggest Trap in Sports Analytics

In June 2026, I sat in a windowless room at the training centre of a Championship club, staring at a report that had just slid out of the printer. Twelve columns of numbers. Four blank ones. Four empty spaces sitting side by side like bars of music someone had torn a few beats out of.

The assistant manager picked up the sheet, skimmed it, put it back down and said: “No red flags.” I stayed silent. He had just read silence as innocence.

Those four blanks were pressing intensity after loss of possession, reaction time after long passes, line-breaking passes, and backward-pass rate under pressure. The club’s cameras had not been recalibrated for the new training pitch. The raw data arrived unreadable. The software, having nothing to compute, printed empty cells.

To a coach who has forty minutes to make a decision, a blank cell and a green cell are the same thing: reassurance.

Missing data and safe data look identical unless you have been taught to tell them apart. That is what I carried with me for the next six years, after leaving English football and giving most of my time to tennis — a sport where every point is recorded, and therefore where every gap becomes more dangerous.

---

A whole system built on the assumption that data is always full

Professional football and professional tennis share an eerily similar data architecture. At the bottom are capture devices: optical cameras, radar, in-shirt sensors, Hawk-Eye. In the middle is the parsing layer, where raw data becomes named fields: first-serve percentage, points won on second serve, break-point conversion, distance covered. At the top sits the analyst, the person who reads those fields and turns them into an actionable story.

This architecture has one fatal weakness. It assumes the middle layer always finishes the job. When a field is not filled, the system does not raise an alarm. It simply leaves it empty. And at the top layer, an empty field does not shout. It stays quiet.

Tennis has its own version of this disease. Since Hawk-Eye Live replaced line judges at many events, an entire data category has vanished from the record: disagreement data. Every time a line judge once made an error and was challenged, we gained a data point about the limits of human perception on court. Now we have a system that is almost never wrong, and therefore we have lost the ability to measure error. Automation erases mistakes, but it also erases the traces of mistakes, and those traces were once part of the truth.

There are things data never touches — like the way a stadium breathes. But there are things data touches and then lets go of, and that is where the danger lives.

I once saw this at a far larger scale. In the summer of Russia, silent keyboards typed out a symphony of data. The 2026 World Cup took me to Moscow as an analyst for a sports outlet. The quarter-final between Russia and Croatia at the Fisht Stadium in Sochi on 7 July 2026 was one of the matches I tracked most closely in my career. I recorded that the hosts ran 148 km in total, roughly 12 km more than their own group-stage average. I wrote a long piece about that physical sacrifice and predicted they would collapse in extra time.

I was right about the collapse. Croatia won 4-3 on penalties after a 2-2 draw. But my piece drew twenty-three reads, while a colleague’s emotional article about fighting spirit was shared thousands of times. That night I sat alone in my hotel and asked myself a different question from the one I expected. I did not ask why people do not read numbers. I asked why a correct number carries less weight than a correct emotion.

Russia taught me that silence is also the deepest layer of data. But it took four more years, in Qatar, for me to understand that silence comes in two kinds: silence because there is nothing to say, and silence because we refused to go looking.

---

Four blanks I once misread

Case one was a seventeen-year-old boy. In 2026, while working as a data consultant for Liverpool, I ran an expected-goals model on the U23 group and found an anomaly: a young forward whose penalty-box touch rate was 30% below average, but whose expected goals per shot reached 0.42. Rhian Brewster, just back from injury. I recommended the coaching staff promote him to train with the first team. Many objected that my model was too theoretical. In a friendly against Tranmere Rovers, Brewster scored twice from three shots.

What I did not disclose in that report was a hole in my model. The column on Brewster’s positioning intelligence was empty for the first three matches. I filled it with intuition, and my intuition happened to be right. A correct conclusion built on an empty cell is still a correct conclusion — but it cannot be repeated, and in my trade, what cannot be repeated cannot be trusted.

Case two was the season without crowds. In 2026, as European football shut down, a Championship club asked me to report on performance in empty stadiums. I analysed five hundred matches and found home teams lost only 0.18 expected goals per match without supporters. The more valuable finding lay elsewhere: trailing teams tended to play long balls seven minutes earlier than usual. The coaching staff adjusted their pressing plan accordingly and took eight of twelve points that June.

When the stands are empty, numbers begin to learn how to sing. But I must be honest: of those five hundred matches, thirty-seven lacked positional data in the first half. I excluded them from the sample. Had I not, the result would have differed. I never disclosed that thirty-seven in the report sent to the club, and I still wonder whether hiding it was a polite form of lying.

Case three was Qatar 2026. Japan beat Germany 2-1 on 23 November 2026, then beat Spain 2-1 on 1 December 2026. I missed both. I went back through my own data and found why: I was so focused on the big nations that I ignored scouting data from Japan’s pre-tournament friendlies. That is an attention-priority error, and in analysis, attention priority is a form of missing data we create ourselves.

Case four is tennis, the case I think about most. On the ATP and WTA tours, statistics are near-perfect for main-draw matches on Hawk-Eye courts. Step one pace outside that, and the gaps open immediately. Challenger events in Asia often lack full positional data. Qualifying matches sometimes leave only a scoreline. And Vietnamese tennis — where Lý Hoàng Nam and Nguyễn Thùy Linh once reached the world’s top few hundred — publishes almost no serve and return data at a depth sufficient for opponent analysis.

What does that mean? It means a Vietnamese player walks onto court against an opponent whose second-serve points won when trailing is unknown to the entire team. Not because the data does not exist. Because nobody collected it.

There is a subtler kind of blank: the one created by the rules themselves. A retirement mid-match, a rain delay, a set cut short by darkness — all produce truncated datasets. Average those truncated tables without checking and you are comparing a player who completed three sets with one who played seven games. The percentage still returns a number. That number is meaningless.

---

The trap is that we believe in emptiness

There is a rule of thumb I learned from English analysts older than me: correlation is not causation. Everyone knows that one. Very few know the second: the absence of evidence is not evidence of absence.

In tennis, this appears where few expect it. A player posts an unusually high tie-break win rate in a season. The table shows it. The table does not show that he played only seven tie-breaks that season — a sample far too small to mean anything. The average is correct. The conclusion drawn from it is not.

The 2026 Wimbledon final between Carlos Alcaraz and Novak Djokovic is an example I often use with younger colleagues. Alcaraz won in five sets, but he lost the first set 1-6. For those thirty minutes, every newsroom had already written the story: Djokovic would win. The data never said that. The data said only that a very small sample was leaning one way.

The same applies to injuries. When a player withdraws from an event, the tournament page usually leaves the reason blank. That blank is quickly filled with speculation: wrist injury, burnout, personal matters. The International Tennis Integrity Agency, established on 1 January 2026 to replace the Tennis Integrity Unit, works on the opposite principle: no comment on open cases. Its silence is read as an admission, when it is only a procedure.

And here is the part that worries me most, the risk I believe is the largest of this decade in professional sport. Large language models are being introduced into analytics departments to summarise reports. Such a model receives a table with four empty cells and writes: “No significant risk indicators detected.” Syntactically, that sentence is correct. Professionally, it is a bomb. A system that cannot distinguish “no risk” from “no data” converts ignorance into confidence, and that confidence spreads faster than any error.

The frightening part is how quiet the process is. Nobody is fired over an empty report. No headline is written about an unfilled data field. The disaster only surfaces when a player suffers an injury the model called clean, or when a player is beaten by an opponent whose scouting file contains not a single line.

I am too old to believe in miracles, but young enough to know which miracles can be measured. And I have been in this trade long enough to recognise that the greatest catastrophe does not come from wrong numbers. It comes from empty cells read aloud in a confident tone.

---

What I could be wrong about

I may be wrong that this is widespread. Perhaps most professional analytics departments already have null-value checks in place, and what I witnessed was merely carelessness in a few places. I have no data to prove scale.

I may also be wrong to weight this more heavily than model bias. A biased model can still be useful if you know its bias. An empty cell tells you nothing at all, not even that it is empty.

And perhaps I am wrong to expect so much of readers. Data needs a coat of story to reach a reader’s heart, something I learned after the summer in Russia. But there is a limit: when the coat of story covers the tear in the data, it is no longer a coat. It is a curtain.

---

A signal for the next round

If I ran an analytics department at a tennis federation, I would start with something small and deeply uncomfortable: every report sent to the coaching staff would carry a mandatory declaration line stating what we do not know. Not a disclaimer at the foot of the page. A line directly under the title, in bold, listing the fields still empty and the reasons why.

It sounds like a pointless bureaucratic ritual. But I believe it changes how a team thinks. When you are forced to write down what you do not know, you stop reading a blank cell as a green one.

For Vietnamese tennis, that signal is more concrete still. Before dreaming of a prediction model, first collect basic serve data for thirty domestic players across three different surfaces. Every dataset is a garden — the farmer plants questions, the harvest yields contracts. But a garden only grows if someone is willing to count every seed, including the rotten ones.

On an Anfield night, I stopped counting numbers to listen to the ghosts whisper. Those ghosts do not speak with the numbers I have. They speak with the numbers I lack. And perhaps, across thirty-eight years of watching this industry, what I have really been hunting is not the formula for a victory, but the formula for recognising what I do not know.