A Blank Cell Is Not a Zero: When the Tennis Stat Sheet Falls Silent

A Blank Cell Is Not a Zero: When the Tennis Stat Sheet Falls Silent

The stat sheet from the women's quarter-final at Melbourne Park has fourteen rows. Row nine is empty.

No zero. No dash. Just a strip of white sitting between two numbers printed in bold. On the broadcast, a graphics operator filled that strip with a large zero, and within forty seconds the commentary box had agreed that the player had just produced a flawless service set.

I stayed behind after the stands emptied. The raw log from the ball-tracking system showed something else: that blank cell was an on-court data station that dropped out for six minutes, precisely across four decisive service games. Nobody recorded. And because nobody recorded, it became the prettiest statistic of the evening.

Those six minutes are not rare. They are just rarely mentioned.

Based on my experience watching matches at Melbourne Park and Sydney Olympic Park, I have seen the same script many times: a blank cell appears on a stat sheet, and within minutes it is filled with a number nobody verified. The person filling it means no harm. They are simply doing what every data system is trained to do — leave no white space.

Three sources, three rulers

A Grand Slam match generates thousands of data points per hour, and they come from at least three sources that never match perfectly.

The first layer is ball tracking. Hawk-Eye Innovations, launched in 2026, has become the standard infrastructure for recording ball position, speed and trajectory at most major events. The second layer is the human statistician sitting courtside. That person decides whether a shot is a winner or an unforced error. The third layer is the aggregation software that computes derived metrics and pushes them to the broadcast graphics.

The critical point sits in the second and third layers. In the second, definitions are not uniform across providers. There is no single definition of "unforced error" co-signed by the ATP, the WTA and independent statistics outfits. A backhand into the net, struck after the player has been dragged off the court, can be counted as an unforced error in one place and a forced error in another. Two stat sheets from the same match, two different numbers, both "correct" by their own ruler.

In the third layer, when a data field has no value, the software has two choices: leave it blank, or fill it. In a broadcast environment, leaving it blank is not a permitted option. So it gets filled.

What is worth noting is that the machine layer and the human layer do not always thicken together. Since 2026, when the US Open introduced electronic line calling on most of its courts, the volume of ball-position data rose markedly. But the number of on-site manual recorders fell at events staged under restricted conditions. Which means: some fields are recorded densely by machines, while others — the ones decided by humans — thin out.

Three kinds of blank

A blank cell on a tennis stat sheet can carry at least three completely different meanings, and all three print identically on screen.

The first is a cell that was never recorded. The data station dropped, the operator was absent, or the shot landed inside a system handover window. The event happened, but nobody witnessed it as data. This is the most dangerous kind, because it tends to coincide with the moments that matter most — when the court is quietest, when the fewest eyes are watching.

The second is a cell cut by threshold. A provider publishes a metric only when the sample is large enough. If a player came to the net twice all match, the net-performance metric may be withheld as too thin to mean anything. The cell is blank not because the value is zero, but because there is not yet enough to speak.

The last is a cell that is genuinely zero. The event did not occur, and that has been confirmed. This case is rarer than people assume.

All three print identically. And when a stat sheet has fourteen rows, telling them apart requires something that is not on the screen: the raw log.

A number is testimony, not a verdict

"Data whispers. Whoever is willing to listen will hear an entire match." But data only whispers when we know where it was born.

In 2026 I wrote an English-language piece predicting Croatia would reach the World Cup semi-finals, based on their group-stage xG. A group of amateur coaches on Reddit called me a bookworm who did not understand football. Croatia reached the final. After the tournament, a reporter from The Athletic contacted me to ask how I calculated "defensive xG prevented" for defenders. I spent two weeks writing Python, cross-checking against StatsBomb data, and sent back a seventeen-page breakdown.

In 2026 they laughed at my xG. This year they ask me what xG is. But in the gap between those two sentences, something else happened that few noticed: the same match, three different xG models, three different numbers. None of the models is wrong. They simply bet on different assumptions about what a shot from that position, in that situation, is worth in goals.

A number is not the truth. It is testimony about the truth, delivered by a witness with a point of view.

The offside line and false precision

The same error appears where few expect it: the offside line.

In football, an offside position is determined by projecting lines from a video frame onto the pitch. That frame is captured at a fixed rate — commonly 50 frames per second. Each frame is therefore 0.02 seconds apart.

A player sprinting at 30 km/h covers about 8.3 metres per second. In 0.02 seconds he moves roughly 16.7 centimetres. That is the distance between two frames. Everything that happens between those frames goes unrecorded, and the system must interpolate.

So when a goal is disallowed because a toe crossed the line, what is being announced is not a fact accurate to the millimetre. It is an inference built from data far coarser than that. This is false precision: a number that looks more certain than what it actually knows.

And the consequence is clear. When the intervention threshold is set at "clear and obvious error", every error below that threshold goes unrecorded. But unrecorded is not the same as nonexistent. The silence of the VAR room gets read as "no error", when it actually means "error below the intervention threshold". Those two sentences are far apart, and they print identically.

Tennis players live inside the same silence. When a player calls for a medical timeout, that is data. When he does not, it may be data, or it may just be a blank cell.

Home advantage and the forgotten variable

In June 2026, when the Bundesliga returned to empty stadiums, I was running match-prediction models for a data consultancy in Sydney. My model priced home advantage at 0.45 goals per match, based on several seasons of data.

After nine rounds without crowds, that number fell to 0.08.

A magazine asked me to write a piece explaining "football without crowds". I declined and asked for three more weeks of data. When the article ran, I opened it by saying that I had been wrong: I had treated the crowd as a constant when it was a variable. Mis-specifying one variable is like losing your bearings for an entire year.

What is worth noting is that not everyone was wrong in that way. Some analysts read the disappearance of crowds as "noise" — they saw the model grow less accurate, concluded the data had become noisy, and carried on with the old model. They had a blank cell where the crowd variable should have been, and they filled it with the word "noise". Then the following season brought crowds back and everything lined up again — but for the wrong reason.

The instinct to fill the gap

Data science has a term for this: imputation. When a cell is blank, you can replace it with the mean, with the nearest value, or with a value inferred from other variables.

In medical or meteorological data, imputation has its own discipline and is reported transparently. In broadcast sports data, it happens silently, and nobody logs it.

The problem with imputation in sport is not that it distorts the numbers. The problem is that it manufactures an event nobody observed. When you fill a zero into a player's service column, you are asserting that the player committed no double faults during those six minutes. Nobody witnessed that. But from that moment on, the assertion is inherited by every model, every column, every graphic.

A subtler version of the same error is reading a blank as "stable". This is the most common trap. With no data, people default to assuming nothing changed. But absence of evidence is not evidence of absence. An empty data table tells you that you know nothing — it does not tell you that everything is normal.

The principle holds wherever records exist. In a testing file, the absence of an adverse finding does not mean everything is clean. It only means the file is thin. A season missing detail is like a match missing stoppage time: it does not make the match shorter, it just makes you understand less of it.

In March 2026, Ashleigh Barty retired while ranked world No. 1, at the age of 25. For months afterwards, every Australian ranking projection and every Grand Slam forecast had a blank in the first row. And that blank was quickly filled with names chosen by feel rather than by indicator.

What to watch in the next round

If a blank cell is a kind of data, it can be measured.

The signal I will be tracking in the coming period is not a new metric but the blank rate of each data provider. When that rate drifts above a provider's own baseline, it signals that the manual recording layer is thinning for some reason — staffing cuts, event conditions, or a process change. And when the recording layer thins, the stories told afterwards go wrong in a predictable direction: they come out cleaner than reality.

That is why I keep the habit of logging the data version for every piece I write, along with the retrieval date. Not to look careful, but because if three months from now a blank cell has been filled with a different number, I need to know who filled it, when, and on what basis.

Before you trust a number, ask where it was born. With a blank cell, the question matters more: was it never born, or did someone fill it in?

Next season, when the stat sheet shows another white space, I will not write a zero into it. I will go looking for the six missing minutes, and see who else is looking.