The Empty Column: A Football Analyst's Discipline
On the third night after the matchday, I reopened my tracking file. Fourteen...
On the third night after the matchday, I reopened my tracking file. Fourteen matches, one page each. The average-position column was filled. The touch-count column was filled. The passing-map column was still empty, because the tracking feed for the eleventh match returned nothing.
I sat in front of that empty column for about forty minutes. In those forty minutes I came up with three very good headlines. One about the away side's back line pushing up too early. One about the holding midfielder losing his position in the 63rd minute. One about the second-half collapse in stamina. All three were coherent, all three had an opening paragraph, and none of them had a single line of data behind it.
I deleted all three. Then I sat still.
In my drawer there are football notes older than the internet. Those sheets taught me something this era considers obsolete: when there are no numbers, the only correct thing to do is say there are no numbers.
This week my inbox held a nine-dimension analysis of a match. It was long, properly formatted, full of sections and tables. And all nine dimensions said one word: insufficient information.
That is why I am writing this. Not to praise a table. But because in football, the empty column is the hardest thing to write, and the thing most carelessly written.
The data machine and what it does not measure
A single league match now generates more data than an entire season did in the 1990s. PPDA — passes allowed per defensive action — measures pressing intensity phase by phase. xG measures chance quality. Field tilt measures who controls the final third. Twenty years ago none of these existed. Now they appear on broadcast graphics within thirty seconds of a goal.
Big clubs have whole departments. Some European sides hire analysts who can write database queries. That is real progress, and I have no intention of mocking it.
But there is a gap between data and conclusion, and that gap is increasingly filled with confidence rather than evidence.
The publishing cycle creates the pressure. The final whistle at 90+4. The bulletin on air at 11 p.m. The analysis must exist before midnight. In that window nobody has time to rewatch fourteen matches, nobody has time to draw passing maps. What there is time for is a plausible-sounding take.
A plausible-sounding take is always available. It is cheap. It needs no verification. It only needs correct grammar.
This week I received an analysis built on a fairly strict professional frame with nine dimensions: tactics and technique; club finance and the transfer market; results and the opinion cycle; league landscape and team positioning; rules and governance; management and the dressing room; risk profile; media narrative and expectation; and industry transmission.
Nine dimensions. Nine tables. Three to six rows each.
All nine said one word: insufficient information.
The first step of that process — deconstructing the source article into structured information points — returned an empty list. No headline. No source. No information points. No entities identified: no club, no player, no coach, no competition.
The person writing that analysis had two options. One was to construct a subject — pick a match, attach a few numbers to it, and write out nine full dimensions. The other was to leave the frame intact and state plainly that there was nothing to analyse.
They chose the second. That was the single best decision in the whole document.
Three layers of data and one final
In 2026, at Luzhniki, I was one of three women journalists accredited for the final. France beat Croatia 4-2. I wrote that night, and one line in the piece drew a reaction from a male editor: Croatia's midfield touched the ball 312 times to France's 541, but Croatia's real threat came from the space between Raphaël Varane and Samuel Umtiti.
He said women only know how to tell emotional stories. I did not argue. I spent two more days redrawing fourteen of Croatia's attacking sequences from FIFA tracking data, and drew a rule I still keep: every argument must stand on three layers of data.
The first layer is average position. It shows where the shape actually stood, not where it was drawn on paper. A team announces a 4-3-3 but the average position of both full-backs sits four metres inward — in possession they are playing with three at the back. Without this layer, every formation analysis is guesswork.
The second layer is touches by zone. It shows where the ball actually travelled. Croatia touched the ball 229 fewer times in that final, yet their touch distribution in the central channel was denser than France's. Less ball, and still reaching the dangerous zone.
The third layer is the passing map. It shows who passed to whom, where, and in which direction. This is the most time-consuming layer and the most skipped. All fourteen sequences I redrew ran through the same zone: the channel between Varane and Umtiti, where the gap between France's centre-backs swung between nine and fourteen metres every time they lost the ball in the middle third.
Three layers. One argument. Two days.
A formation diagram is only a sheet of paper; the players are the ones who write the match. But you can only read the writing on that sheet when three layers of data are stacked on top of each other.
The summer without crowds: when a variable is removed from the equation
In 2026 the world played in empty stadiums. I tracked 28 matches in the Chinese top flight. It was a rare dataset: same league, same rules, largely the same players, but one variable removed from the equation.
The results: average pressing intensity up 11 percent. Counter-attacking efficiency down 23 percent. Home advantage all but gone.
The mechanism behind those two opposing numbers is the interesting part.
With a crowd, the noise creates an invisible pressure that forces the away side deeper. They defend lower, the block compresses toward their own goal. That gives the home side more ball in the final third, but it also opens a wider counter-attacking lane behind the home defence.
With the stands empty, that pressure disappears. The away side pushes up. Their block sits higher, so they press more — the 11 percent rise is a direct consequence. But pushing up also means that when they win the ball, the space behind them has narrowed. The counter loses its run-up. Efficiency falls 23 percent.
Shanghai SIPG lost four in a row in that period. The easiest explanation is form. The more accurate one is structural: SIPG built their game on the assumption that opponents would sit deep under crowd pressure, leaving them space to switch the ball quickly through the middle. With no crowd, opponents did not sit deep. The space never opened. The whole system lost its anchor.
Football without crowds is a completely different sport. Not different in emotion. Different in geometry.
An empty column has three causes, and three responses
When data does not arrive, an analyst tends to do one thing: replace data with memory. Memory is always available, always coherent, and always biased.
But an empty column is not a single state. It has at least three causes, and each demands a different response.
The feed failed. The eleventh match in my file was this case: the tracking cameras dropped out from the 58th minute after a transmission fault. The data still exists out there; I simply had not retrieved it. The correct response is to go and get it, or to state in the piece that from the 58th minute there is no positional data. Extrapolation is not allowed.
The sample is too small. A substitute comes on in the 78th minute and touches the ball four times. Four touches say nothing about his average position. The table will still display a number, and that number will look weighty. It carries no weight at all. The correct response is to state the sample size beside every claim.
The metric does not apply to that match state. A team playing the last thirty minutes with ten men and a one-goal lead is not having its pressing measured as tactical intent; it is being measured as circumstance. The correct response is to remove that metric from the argument and describe the block's structure instead.
Three causes, three responses. All three lead to the same line in the draft: insufficient data to conclude.
That line is hard to write. It has no good opening. It does not make a headline. It does not get shared.
It is just correct.
Translated into pitch language
There is a rule I have applied to myself since 2026, and it has saved me more often than any metric: after every tactical model, there must be one concrete passage of play, with a minute.
A model with no passage of play is decoration. It sounds highly professional and cannot be checked.
Saying "the away side lost control of the central channel in the second half" is a meaningless sentence. Saying "in the 67th minute the away holding midfielder was dragged to the right flank by the ball, the gap to the left centre-back stretched to eleven metres, and the equaliser came from exactly that space forty seconds later" is a sentence that can be checked, can be argued with, and can be wrong.
A claim that can be wrong is a claim with value. A claim that cannot be wrong is a useless claim.
In 2026 my first blog on a new sports platform in Beijing got 237 reads in a week. A young YouTuber dissecting the same match got 130,000 views. I was not jealous of him. I went back and rewatched fourteen Chinese FA Cup group-stage matches, found a repeated positional error by the Beijing Guoan full-back — the same error, in the same zone, eleven times across four matches — and wrote three thousand words with diagrams. Three
