The Data Vacuum: Analyzing Tennis When Every Cell on the Spreadsheet Reads N/A

At two in the morning in a Los Angeles studio, I keep a fourteen-column spreadsheet open on my second screen. Column one is the tournament. Column two is the surface. Columns three through ten are the familiar metrics: service games won, return points won, break-point conversion, first-serve points won, double faults in tie-breaks, winner-to-unforced-error differential, points won when trailing in a game, and average minutes per set. The last four columns are variables I built myself: the gap between expectation and result, three-week match load, rest days between matches, and a column I simply label confidence.

Fourteen columns. Not a single cell held a number.

It was not a broken connection. There was simply nothing to fill in. The match I had been assigned to call was still seven hours away, the player had not confirmed fitness, the tournament had not published a practice schedule, and the point-by-point feed had not opened for that round. In my ear, the producer asked what I had prepared. I told him the truth: one framework, four questions, and no numbers at all.

People imagine tennis commentary as a profession of talking. Most of the job is filling blank cells — and knowing when you are not allowed to fill them.

The measurement era

Tennis entered the measurement era later than basketball or football, but it caught up fast. Hawk-Eye first appeared at a Grand Slam in 2026, initially only to serve player challenges. Within fifteen years it became an automated line-calling system and then a three-dimensional data source describing ball position, spin, bounce point and trajectory. Every rally became a queryable row of data.

Alongside that infrastructure, tournament statistics platforms began publishing point-level, game-level and set-level metrics to the public. By the middle of the 2020s, a viewer in Hanoi could open a phone and see the second-serve points won by a player in an ATP 250 qualifier — something that two decades earlier existed only in a coach's notebook.

Analysis changed shape accordingly. Probability models built on serve and return strength became standard tools. Independent analysts built their own surface-specific Elo ratings. Tactical consultants sat beside players at major events carrying notes on an opponent's tie-break serving patterns.

But the measurement era produced a less-discussed side effect: it created the impression that everything can be known. Audiences grew used to opening a page and finding a number. Journalists grew used to citing one. Bookmakers grew used to pricing with one. And analysts like me grew used to starting every argument from a cell that had already been filled.

When that cell is empty, most of us do not know what to do. We fill it with adjectives. We call a player in form without a denominator. We call a win pivotal without defining where the pivot is. We use a confident tone to cover a simple fact: we are holding nothing.

There are four kinds of emptiness in tennis analysis, and each demands a different response.

The first is emptiness because nothing has happened yet. This is benign; it is only a matter of time, and historical data can build a forecast with an explicit confidence interval.

The second is emptiness because information is withheld. Injury is the classic case. A player withdraws citing a wrist injury or a fitness issue, and that is the end of it — no diagnosis, no recovery timeline, no medical confirmation. Every pre-match model instantly loses value, while the ranking page keeps showing the old number.

The third is emptiness because the sample is too small. A player moves from hard court to clay, plays three matches, wins two. A 67 percent clay win rate looks impressive. Three matches is not a sample. The problem is that leaderboards display it as a fact.

The fourth is emptiness by the nature of the sport. Some things cannot be measured by sensors: the pressure before a serve at 0-40, the silence of a centre court, the feel of the ball on a humid evening. These variables exist, they affect results, and they sit outside every spreadsheet.

These four kinds of emptiness are not the same, but the public response to them is identical: fill them with story.

Injury, withdrawal and an asymmetric information market

Injury information in professional tennis operates like a decentralised market with no ledger. Those who hold information are the player, the coach, the physio team, the tournament representative, and occasionally a well-connected journalist. Those who do not are everyone else — including most bettors and most fans.

The gap produces what I call information lag. A player may have had a sore back for three weeks, still walked on court, still won on experience, and only when losing an unusually flat match did the public learn that he had not served at full intensity for the entire tournament.

Forecast models do not fail because the algorithm is weak. They fail because the input is stale. A model built on a player's seasonal service points won will confidently assume he holds serve 84 percent of the time. If his shoulder hurts and the actual figure that fortnight is 71 percent, every prediction generated from that model is a form of pseudo-scientific fiction.

I once sat in a press conference where a player was asked about his condition. He said he felt fine. Three days later he withdrew before the quarter-final. A younger colleague turned to me and asked whether he had lied. The answer is more complicated. In professional tennis, disclosing an injury is a strategic act with real consequences: the opponent learns where to target, sponsors learn what to worry about, the tournament learns what to plan for. Silence is a rational decision.

That means analysts must learn to work with an input that is deliberately blurred. The only way to keep credibility is to say plainly: I have no fitness data on this player, so every number I give you is a probability under normal conditions, not a prediction.

That is a hard sentence to say on air. It is also an honest one.

Match load: data that exists but is rarely used

If there is one area where professional tennis has data but rarely uses it, it is match load. Tournaments publish results. Statistics platforms publish minutes played. Very few publish rest days between matches, flight hours between continents, time zones crossed, or training volume between two events. This data exists, scattered across dozens of sources.

In a personal project I ran in 2026, when competition stopped because of the pandemic, I collected data from 312 matches across three major European football leagues in the 2026-20 season, comparing results with crowds and without. The most interesting finding was not about football. It was about the structure of attention: without crowds, home advantage fell noticeably, while total goals rose slightly. The transferable hypothesis for tennis is that a low-pressure environment reduces stress errors but also reduces defensive discipline.

For tennis, that hypothesis still awaits a long enough dataset. It shows something important: the data you need is often not where you are looking. People look for win rates. The answer sits in a flight log.

A player wins a semi-final in Europe on Saturday night, flies to North America on Sunday morning, plays a first round on Tuesday. That is three days with two time-zone shifts, one long-haul flight and one practice session on an unfamiliar surface. His seasonal hard-court win rate still reads 68 percent. The real denominator for that week is three days and a body out of rhythm.

Load does not appear in the statistics table, and that is exactly why it matters.

Surface conversion: when small samples lie

Every season, broadcast programmes spend a few minutes on a comparison table of a player's record by surface. It looks scientific: four columns, four surfaces, four percentages.

Most players contest only eight to fifteen matches a year on any given surface. At a denominator of twelve, a single win or loss moves the rate by more than eight percentage points. That swing is larger than any technical difference we are trying to measure.

In other words, most surface-specialisation debates rest on numbers whose uncertainty exceeds their signal. This does not mean surface is irrelevant. It means we are using the wrong instrument to describe a real phenomenon.

A more honest approach splits by specific condition: bounce height, bounce depth, court speed measured by the tournament's own index, temperature and humidity at match time. A player can be excellent on dry, hot clay and poor on damp clay after rain. Both are logged as clay.

I once spent two days testing this with data from a major clay event. The result was not strong enough to publish, but strong enough to change how I speak on air. Since then, whenever I prepare to call a match on a surface where one player has fewer than fifteen matches across two seasons, I write one line in my notes: small sample, no conclusion.

That line has saved me from several deeply embarrassing calls.

Ranking defence and the rolling 52 weeks

The professional ranking system runs on a rolling 52-week cycle. Points from a tournament are deducted when that tournament returns the following year. This creates a pressure few fans see directly: the pressure to defend points.

This is one of the few areas where public data is fairly complete. Anyone can look up how many points a player will lose in the next four weeks. Very few use it to explain odd on-court behaviour.

A typical case is a top-20 player entering an event below his usual tier, playing with unusual intensity, then losing in the third round through exhaustion. Newspapers call it inconsistent form. The ranking table shows a player inside a points-defence window who must accumulate. Conversely, players entering a period with nothing to lose suddenly play freer tennis. Reporters call it rediscovering themselves. Sometimes that is true. Sometimes it is arithmetic.

The point is not that every result reduces to points. The point is that points structure is an undervalued variable, and ignoring it produces false stories about motivation.

Rules, medical timeouts and the grey zone

Tennis has a relatively clear rulebook for on-court situations. One area blurs considerably: decisions involving health and time.

A player calls the physio. The match stops for a few minutes. The player continues and wins, or loses. Public data records that a medical timeout occurred, but rarely the intervention itself, the actual seconds elapsed, or the condition before and after. For an analyst this is a serious gap, because at the highest level that stoppage is often the inflection point.

I have rewatched a match in which a player lost the first set with a very low second-serve points won rate, called the physio early in the second set, and then won second-serve points at a rate well above his own seasonal average. No data explains it. It could be a technical adjustment. It could be a change in how the opponent returned. It could be the medical intervention. Three hypotheses, one dataset, no way to separate them.

In that situation a professional has two choices. One is to tell a compelling story, attribute some mental quality to the player, and present it as fact. The other is to say plainly that the data does not permit a conclusion and to describe the phenomenon in purely descriptive language. The second is less attractive on television. It is the only choice that lets me look in the mirror after a season.

An accurate description of the unknown is worth more than a false explanation of what happened.

Night sessions and unmeasurable variables

In recent years, major tournaments have expanded evening and night sessions. This brings broadcast revenue and global audiences, and it creates a variable that official data barely captures.

Humidity rises. Temperature falls. The ball travels slightly slower and bounces slightly lower. These shifts are small enough that match-level aggregate metrics barely register them. For players whose serve depends on high trajectory and heavy spin, they can be the difference between winning and losing a tie-break.

I have often sat in a studio as a match entered a fourth set close to midnight local time. On the tracking screen I watched average serve speed fall. On the statistics screen, first-serve points won stayed flat. Two streams of information, one reality.

Official data struggles to fill this kind of gap. A sensor measures ball speed; it does not measure how a body feels after four hours. An algorithm counts racket contacts; it does not count the times a player chose not to attack for reasons she herself would struggle to explain.

I am not arguing that intuition should replace data. I am arguing that data should be read alongside an awareness of its limits. When a metric stays stable while my eyes see decline, the likeliest explanation is that the metric is measuring the wrong thing.

The media, the analytical darling and the expectation loop

Every season produces a few analytical darlings: young players with standout metrics, pushed by statistics sites, embraced by media, and within months treated as though their progress were a law of nature.

I helped create one. In 2026, in a sports channel's analytics room, I watched a young striker's footage fourteen times, dug into expected-goals data, and found that his no-backlift finishing style produced an unusually high conversion rate. I wrote a long piece, published it on the channel blog, and was handed the lead commentary slot for his next match.

That article launched my broadcast career. It also taught me a lesson I only fully understood years later: when you place a name on the altar of data, others will worship it with you, and when it fails, you are the first person asked why.

The analytical darling eventually has to stand on his own feet. I say that on air every time a young player is pushed too fast. Not to diminish him — to remind everyone that denominators change and opponents read data too.

The expectation loop is mechanical. A young player wins a few matches. Metrics rise. Media writes. Bookmakers adjust. He becomes the favoured side in every match. Opponents prepare specifically for him. The next result falls short. The story pivots instantly from generational talent to not ready for the big stage. Throughout the entire loop, no new data was generated. Only expectations moved.

From raw data to data as a product

A structural shift will shape how we watch tennis in the coming decade: data is moving from an internal tool to a commercial product.

Fifteen years ago, point-level data at a tournament was the property of the organiser and its technology partner. Today it is packaged and sold to media platforms, bookmakers and independent analytics firms. Player tracking data, once used only for broadcast, is becoming an input for injury-prediction models.

The upside is transparency. The downside is a widening gap between those who have data and those who do not. A low-ranked player without an analytics team cannot know that over the past three months her second-serve points won fell ten percentage points in third sets. A top-10 player knows, and knew two weeks ago.

This asymmetry is rarely covered, because it generates no story. It only generates results.

Over the years I have learned to read small rate changes in unheralded players as traces of investment. When a world number 80 suddenly improves break-point conversion across three consecutive tournaments, it is rarely a miracle. It is a new coach, a new data specialist, or a team that found resources.

Numbers are seasoning. People are the meal. But in a sport where prize money is distributed by round reached, access to data becomes a competitive advantage measurable in currency.

The contrarian angle: value lives where nothing can be measured

Most people in this profession believe the edge lies in having more data. I believe the opposite is true most of the time.

When data is abundant, everyone has the same information. Models converge on the same numbers. Bookmakers price almost perfectly. The marginal edge between a great analyst and an average one shrinks to invisibility.

When data is empty, the reverse happens. Most people fill the gap with feeling, rumour and attractive narrative. A small minority say: I do not know. Over the long run that minority is more accurate, precisely because it refuses to forecast without a basis.

This is emotionally difficult. Audiences want answers. Broadcasters want answers. But silence is not the absence of an answer — it is the answer, for those listening.

I paid to learn this. At a World Cup, I gave a safe penalty-shootout prediction; the scoreline landed, but I knew I had avoided a concrete number out of fear. Afterwards I rewatched the entire tournament, logged every phase I had misjudged, and built a spreadsheet comparing my predictions with reality. It was not flattering. It showed that most of my error came from filling gaps with confident tone.

Since then I state confidence levels. Seventy percent. Sixty percent. Forty percent, and here is why I am unsure. A spreadsheet does not know what longing is, and we should stop pretending otherwise.

There is another temptation worth naming. When data is thin, analysts drift toward nostalgia, comparing today's players with those of twenty years ago and concluding that the older generation was tougher, more disciplined, more complete. Those comparisons usually ignore that playing conditions, equipment, sports medicine and calendar density have changed so much that the two eras no longer share a measuring stick. Every time I am about to write such a sentence, I ask what a twenty-year-old reader would see in it. If the answer is an old man defending his own time, I delete the sentence.

What to watch from here

Over the next three months, three variables matter more to me than the ranking table.

First, rest days between matches for players inside the top 20. As the calendar tightens, recovery gaps will show on court before they show in the standings.

Second, third-set second-serve points won among young players. This is the clearest separator between someone pushed upward by data and someone with the technical base to absorb pressure.

Third, the quality of injury disclosure. If tournaments keep publishing the minimum, every analysis must carry an uncertainty warning. That may be good for the profession.

A quiet summer turns records into orphaned numbers. But silence is also an opportunity, if those of us in the trade dare to say we do not yet know. The analytics market will not reward the loudest voice. It will reward the one that speaks accurately about its own limits — and stands firm while every cell on the spreadsheet is still blank.