A Tennis Report With Zero Numbers, and Why That Is Real Data

This morning I received a nine-layer tennis analysis package. It had a technical table, a data table, a points table, a risk matrix, an industry transmission diagram. Forty rows.

Not a single number.

Not a wrong number. No number at all. Every cell read "N/A – insufficient information". The technical layer had no player. The data layer had no first-serve percentage. The tournament layer had no week, no surface, no round. The commercial layer had no broadcast deal, no prize money. The governance layer had no governing body.

It is the most honest tennis report I have read this season. And I am writing this to explain why.

My job is to read sports data chains and find where they break. Nine years in, I have learned something few people want to hear: most of what gets called "analysis" in your feed is not analysis. It is a model fed a headline, filling in the body by itself.

A proper tennis analysis system has to pass through nine layers: technical, data, tournament, tour landscape, rules and governance, team management, risk, media narrative, industry transmission. Each layer has mandatory inputs. The data layer needs ranking points, a 52-week window, and a win-loss sample with opponent quality. The tournament layer needs an event name, a week, a surface, a draw shape. The governance layer needs to know which body holds jurisdiction — ITF, ATP, WTA, the Grand Slam board or a national association — because the disciplinary reach of those four is not the same.

The package I received this morning had none of it. And instead of raising an error, it returned a fully structured document, one "N/A" per cell. In engineering this is called silent failure — broken without a sound.

The key detail: the domain label survived. The system still knew this was tennis. What vanished was the extraction step. The node that reads the source document returned empty — most likely a paywall page, a bot-blocked page, or body text sitting in a node the parser never touches. The package's own diagnostic says as much: the failure sits in extraction, not classification.

This is not a rare incident. It is the most common incident in the sports content chain, and the most dangerous — because it does not corrupt the output. It only empties it.

Now the part worth discussing. If a system like this is forced to "produce a result", what does it generate?

I have seen it. Three times this year.

The first time, the model picked a real player, assigned him a "left-handed baseliner" style, then analysed the backhand weakness of a man who never appeared in the source document. The second time, it constructed a final at an ATP 250 in a week when no event was held. The third time, it wrote a passage about "points-defence pressure" — clean grammar, clean terminology, and no time anchor whatsoever.

The trap sits here: the ATP points system does not permit that ambiguity to exist. Points roll off on a 52-week cycle. A Grand Slam champion receives 2,000 points; runner-up 1,200; semi-final 720; quarter-final 360; round of 16 180. A Masters 1000 awards 1,000 to the champion, 600 to the runner-up, 360 to the semi-finalist. An ATP 250 awards 250.

Put the last three numbers side by side and the hole in every "he's finding form" piece becomes obvious: three ATP 250 titles add up to 750 points — more than the 720 a Grand Slam semi-final is worth. A player farming small events consistently can sit above a man who reached the final week of the biggest tournament of the year. The ranking does not measure level; it measures schedule.

This is where I use my usual move: crossing the data. Not player against player — points structure against season structure. A player entering a points-defence window is effectively playing two matches a week: one against the opponent in front of him, one against himself twelve months ago. The media only sees the first.

From my own experience watching matches, the clearest signal of that phase is not in the score. It is in a player serving more conservatively in the decisive service games, choosing topspin over flat, cutting risk at 4-4 instead of accelerating. Outsiders call it "loss of form". People in the trade call it counting points.

And here is the bridge to football. In the transfer window, people argue about transfer fees. In tennis, they argue about ranking points. Both make the same error: measuring an asset by its list price rather than its cash flow. Transfers are not mathematics, but mathematics explains why people lose their minds over them.

Then the part no "N/A" table ever touches: protected ranking. A player out long-term with injury can enter events using the ranking frozen at the moment of injury. The mechanism saves careers, but it creates another kind of noise: a world No. 90 in the system might be a former top-20 player in recovery, or someone who has never been inside the top 50. The ranking cannot tell them apart.

Japan did not play beautifully — they simply exposed a formula the rest of the world ignored, and that formula always starts with counting the things worth counting rather than the things easy to count.

If I had to narrow this entire piece to a single variable, it is the points-defended rate: points earned again inside the same tournament window last season, divided by points that needed defending. Above 1.0 is genuine ascent. Below 0.6 is decline the ranking has not yet priced in. That number exists in no official points table — you have to build it yourself, event by event, week by week.

Which is also why an empty report has value. It does not give me that rate. It simply tells the truth that it does not have it.

The counterintuitive angle: the sports industry is not short of data. It is short of permission to say "I don't know".

Every link in the sports content chain is optimised never to return empty. Fan pages must publish daily. Newsletters must carry a take. Models must produce a pick. Nobody pays for a blank page, so the system learns to fill — and fills with whatever looks most like truth: a real name, a real tournament, a number in the correct format.

The paradox: the only thing in this industry that cannot be bought is a blank cell. You cannot argue an "N/A" out of being wrong. You cannot place an ad on it. You cannot make it amplify the mood of a crowd.

I was wrong about school football data, and that was the most precise finding I have ever had — because that time I published the model first, the data answered later, and the data said no. Had I waited for enough numbers before speaking, I would have learned nothing.

My debate room collapsed in 2026 for the opposite reason: too many opinions, too few blank cells. Forty-seven people, seven topics, nobody willing to say "I have no numbers here". An analysis group does not die of a shortage of ideas. It dies of a shortage of people willing to stop.

So when you read a tennis analysis this week and it flows smoothly, never once hesitating — ask: where is the blank cell in this piece?

If there is none, you are reading a product optimised never to admit a gap. And the only thing it measures is its own production speed.

If there is one, keep reading. That is where real data is breathing.

I believe in data, but I believe more in the mistakes data cannot measure. And a table of nothing but zeroes, this time, measured more than most points tables I have seen all season.