A Report Full of N/A: The Discipline of Verification in Tennis Data Work

22:40 on a Tuesday, the desk lamp still on in New York. On the screen sits a nine-row table, each row one dimension of a professional tennis review: technical and tactical, data and form, tournament system, tour landscape, rules and governance, team management, risk, media, and industry transmission.

The second column, read top to bottom, looks like an inventory of failure. The phrase “insufficient information” appears seventeen times. No tournament name. No player. No surface. Not a single data point to hold on to.

The deadline is 6 a.m. I have six hours, an empty table, and a professional habit that has hardened into instinct: if I cannot verify it, I do not write it.

WHY AN EMPTY TABLE IS STILL DATA

This is the paradox at the centre of modern tennis analysis, and I have watched it for nearly three decades. Tennis owns one of the densest data infrastructures in sport. Since Hawk-Eye was deployed at the ATP Miami event in 2026, almost every rally at a major leaves a trace: serve speed, landing point, spin rate, movement direction, points won behind the first and second serve, break points saved.

That density comes with a denominator thin enough to be taunting. A best-of-three match runs roughly 150 to 200 points. Best-of-five can reach 250 to 300. Inside that, break points are usually only two to six. Every metric about delivering in the decisive moments therefore stands on very narrow ground.

Then four variables split the sample: hard court, clay or grass; indoor or outdoor; altitude, ball type, temperature; and format. Divide 150 points across four dimensions and what remains is mostly noise. That is why I keep telling editors that a tidy, complete, handsome tennis statistics table can still mean nothing.

The nine-dimension framework turns out to be useful in the opposite way to its design. It does not generate content. It blocks content.

THE STOP THRESHOLD HAS TO BE SET BEFORE WRITING

I set my sufficiency threshold in advance: three independent sources, or two independent datasets plus one video confirmation. For match data, the first source is the official point log, the second an electronic tracking system, the third the video needed to check a specific rally. For market information such as transfer fees or endorsement contracts, the threshold is two sources that state figures plus one indirect confirmation from the management side.

When all nine dimensions return “insufficient information”, that threshold is not being violated. It is running correctly.

Fans watch with their eyes; I watch with a probability distribution. A distribution needs a sample. Without a sample there is no distribution, only a feeling dressed up in terminology.

DENOMINATOR BEFORE NUMERATOR

In the data class I once taught to interns, the first lesson was not calculating serve points. The first lesson was asking: what is the denominator?

A player saves four of five break points in one match. Eighty percent. It sounds like a specialist. With five break points, though, the confidence interval for that figure spans nearly the whole scale. It takes around forty matches before the denominator is large enough to separate a genuinely strong break-point server from someone who got lucky for three weeks.

For first-serve points won, I usually need roughly forty to fifty matches on the same surface before I start trusting a trend. Below that threshold I describe in ranges, not conclusions. “Somewhere between 60 and 65 percent” is far more honest than “he serves brilliantly”.

THE RANKING IS A STRUCTURE, NOT A MEASURING STICK

The ATP ranking runs on a rolling 52-week window. A Grand Slam champion receives 2,000 points. A Masters 1000 champion receives 1,000. The ATP Finals can deliver up to 1,500 points to an undefeated winner. The Masters 1000 events are mandatory for eligible players.

That structure produces points cliffs. A deep run at a Grand Slam can bring in close to a quarter of a year’s total points in two weeks. When the 52-week window rolls forward, those points leave the system all at once, and the ranking position falls while actual form does not fall with it.

I have watched enough cases to settle on one rule: the ranking measures accumulated achievement inside a defined time window, not current level. About eighty percent of the “does this player deserve to be top 10” arguments I have heard are really arguments between two different definitions of the same word.

SURFACE FORCES THE SAMPLE APART

Rafael Nadal won 14 Roland Garros titles. It is one of the most stable facts in the sport, and it is also the cleanest illustration of why a denominator must be split by surface.

A cross-surface composite metric is the average of skills that differ in kind: sliding on clay, staying low on grass, handling high bounce on indoor hard courts. Combining them into a single index and then comparing two players is comparing two mixtures with different ingredients. The result can be arithmetically correct and semantically wrong.

So I always place four lines of notes beside any metric: court conditions, physical state, recent schedule, and the tactical role the coaching team has assigned. Those four lines cost ten extra minutes and save me from conclusions that would cost ten years.

WHEN THE EMPTY TABLE IS ITSELF THE ANSWER

Back to the nine rows. If I fill the blanks with numbers, I have an article. If I leave them, I have a record of process, and a correct answer.

I choose the second, then rewrite the request to the source side: send the original text, or at minimum the tournament name, the player name and a timestamp. A review with nothing to review is not a weak review. It is a review in the wrong format.

THE COUNTERINTUITIVE ANGLE

A common belief runs through sports media: more data means fewer errors. From what I observe, the error rate does not fall. The speed of error rises.

The truth sits deep under the table, where headlines never reach. When everyone can access the same data pool, the edge stops being ownership of numbers and becomes the choice of which numbers to put first. Bias lives exactly there, in the right to select the sample.

One player landed 68 percent of first serves in his latest match and 61 percent across the season. Both are true. A headline needs only one of them.

The nine-dimension framework has its own flaw. A tool that returns “insufficient information” for every case is a tool missing a proper stop condition. It protects the analyst from error and simultaneously blocks the ability to say anything at all. The market forgets nothing; it merely disguises itself as a new summer. But a writer permanently stuck at the starting line never arrives.

WHAT TO WATCH IN THE NEXT ROUNDS

I do not write about tennis; I only keep a ledger of karma from data. That ledger only means something when the table has a denominator, the denominator has context, and the context has a source.

The signal worth tracking in the coming rounds is not a new metric. It is an old habit: before believing a percentage, ask how many points it was calculated from, on which surface, and across how many matches. Whoever answers those three questions with specific numbers is doing the job. Whoever offers only the percentage is selling emotion.