Empty Files and the Algorithm's Illusion: Why Tennis Still Mismeasures Player Load

Third set, 4-4. The player sits down on the bench after a change-of-direction rally, his right hand on his calf. The broadcast camera zooms in, and on the stadium's big screen the stat board keeps ticking over: first-serve speed averaging 191 km/h, 11 winners, 4 double faults, 78 percent of first-serve points won. Not one line says he has spent 11 hours and 40 minutes on court across nine days. Not one line says he has had his ankle re-taped four times that week. And not one line mentions that three weeks earlier he withdrew from an ATP 500 in Europe for a reason the tournament recorded in its official notice as "undisclosed injury."

I was sitting in row twelve, stand B. About fifteen minutes later the medical team walked on. One and a half minutes. Two minutes. Four minutes. The crowd started whistling, and I understood why — they paid to watch tennis, and he was trying to hold on to a calf. When he stood up and walked to the baseline, the stat board on the screen switched to a new frame, and it still had nothing to say about what had just happened.

That is why I started keeping notes. Every time a player goes down, the public argues about the calendar. I want to argue about something else: what we are not measuring.

A circuit with no silence in it

Professional tennis runs on a calendar with almost no genuine break. Four Grand Slams spread from January to September. Nine ATP Masters 1000 events and eight WTA 1000 events wedge in between. The ATP and WTA Finals close the year in November. Davis Cup and Billie Jean King Cup stretch to late November, sometimes into December. Add the Olympics once every four years, and a 250 and 500 tier that every player must enter to defend ranking points.

Three surfaces swap places within five months. Hard courts in Australia and North America, clay from April to early June, grass for exactly four weeks, then hard again, then indoor. Every surface change is a re-adaptation for the muscle and tendon system: a different foot contact point, different bounce, different reaction time and, most importantly, a different way of braking.

A top-10 player can play 60 to 75 singles matches in a season. That figure excludes doubles, excludes Davis Cup, excludes end-of-year exhibitions. At a Grand Slam, a four- or five-hour match is routine; each hour of it is equivalent to hundreds of high-intensity accelerations and decelerations.

The structure of this sport differs from football in one fundamental way. No club manages a player twenty-four hours a day. Tennis players are independent contractors. They pay the salaries of their coach, fitness specialist, physiotherapist and personal physician. They decide which events to enter. And they are the ones responsible for whether their physical condition is fully disclosed.

That gap is where the problem lives.

Four governing bodies, four data silos, nobody joining them up

Football has something tennis does not: a centralised, mandatory injury surveillance system with unified standards. The UEFA Elite Club Injury Study has run since 2026, collecting data from Europe's leading clubs under a single set of definitions: what counts as an injury, what counts as time lost, from which point recovery is measured. Because of that, comparisons are possible across seasons, competitions and countries.

Tennis has no equivalent in open form. Four groups govern the sport: the ATP runs the men's circuit, the WTA runs the women's, the ITF runs Davis Cup, Billie Jean King Cup and the Olympics, while the four Grand Slams are operated separately by four national associations — the French Tennis Federation, the All England Club, the United States Tennis Association and Tennis Australia. Those four groups use four different medical systems, four different record-keeping methods, and are under no obligation to share data with one another.

The result is a paradox. The sport with some of the richest broadcast data on the planet lacks the most basic medical data.

We know a player's first-serve speed to the nearest kilometre per hour, but we do not know how many hamstring strains he has had in three years.

Medical timeouts and retirements are recorded, but fragmentarily. A retirement is counted as a loss, not as a medical event. A pre-tournament withdrawal is logged in a few words such as "undisclosed injury." A six-week absence is not coded into anything queryable.

I ran into exactly this trap years ago, at a much smaller scale.

In 2026, as a third-year sports analytics student, I interned at the Paris FC youth academy and was assigned to review the U19 medical files. I found an eighteen-year-old midfielder named Lucas Moreau with three hamstring pain episodes across fourteen matches, still starting every week. I charted injury frequency against training intensity, and the model returned an 87 percent risk of muscle tear if he continued at that load. The coach reluctantly gave him a week off. Lucas avoided a serious injury and scored twice in his next three matches.

What I learned was not in the 87 percent figure. It was that if I had not opened those medical files, nobody on the coaching staff would have known about the three hamstring episodes. They were buried in a document nobody was tasked with reading.

I found the gap not in the player's body but in the way we measure it.

Summer 2026 brought me to a similar conclusion at a larger scale. When Germany crashed out in the World Cup group stage in Russia, most coverage blamed the coach's tactics. I went the other way: I cross-referenced the physical records of the players who started all three matches and found a clear pattern — one of them was covering only about 68 percent of the distance he had covered the previous season, while still playing the full 90 minutes every game. Forcing an unrecovered player onto the pitch is not a tactical decision. It is a medical error disguised as a tactical decision.

Germany collapsed because physical warning signs were ignored for months, not because a tactical shape was wrong.

In 2026, when football shut down during the pandemic, I proposed building a model of injury-recurrence risk after an interruption, based on data from previous disrupted seasons. I collected 1,200 medical records from five clubs. The result showed muscle tears rising 23 percent in the first four weeks after football returned. The cause was not that players had weakened during the break. The cause was that every forecasting model had been built on a baseline that had disappeared.

Those three episodes taught me the same lesson, and it applies directly to tennis.

When the dataset is empty, the algorithm returns low risk

This is the point I want to dwell on, because it is the least discussed.

A risk-assessment model works on a simple principle: it looks for patterns in past data. If a player has a history of lower-back pain, the model flags it. If he has had three hamstring strains in two years, it flags it harder. If he has had ankle surgery, it factors that into the long-term coefficient.

But when the file is empty, the model does not raise an alarm. It returns a low score. It files the player in the safe group.

The absence of identified risk is not the same as the absence of risk.

In tennis, empty files are common. A twenty-two-year-old who has just broken into the top 50, never had surgery, never had a significant withdrawal — on paper he is a perfectly healthy specimen. But that empty file might simply mean he has never played enough matches for a problem to surface. It might mean he concealed something. It might mean a hamstring strain at nineteen, at an ITF Challenger, that nobody bothered to record.

The professional structure of tennis creates an incentive to conceal. Players pay their own costs. They live on prize money and sponsorship, both of which depend on playing often enough and regularly enough. Declaring a hamstring problem to a tournament's medical staff can lead to monitoring, further questioning, a recommendation to rest — and resting means losing points, losing money, losing position.

Nobody in this ecosystem has an incentive to say "I am not fit."

Compare that with football: a player at a major club has training load measured daily, periodic ultrasound scans, a doctor and a fitness specialist who see him eleven months a year. A tennis player has a small personal team, and that team is paid by him. The conflict of interest sits inside the structure, not inside any individual.

So when a player goes down in the third set, the right question is not "what happened today."

An injury is a story — but that story begins long before the player falls.

What the stat board does not measure

The reflex response to every wave of retirements is to blame the calendar. I think the calendar is part of the problem, but not the largest part. The largest part is that we are measuring the wrong thing.

In football there is a metric I have always opposed: distance covered. It is packaged as a measure of effort, flashed on screen as proof of commitment. But a player who runs twelve kilometres, seven of them ineffective, not involved in any phase of play, still generates a beautiful number. Distance cannot distinguish between running to create space and running after a ball that has already passed you.

Tennis has its own version. Serve speed, winner counts, double faults — these are what go on the big screen. These are what spectators remember. And these are the things that barely measure physical load at all.

The real load of a tennis match sits somewhere else: the number of short accelerations, the number of decelerations, the number of abrupt changes of direction, the number of jumps landed on one leg, the number of split-steps into a serve. A three-hour singles match can contain thousands of changes of direction. Each one places an asymmetric load on the hamstring, the knee ligaments, the Achilles tendon. None of them appears on the stat board.

And nobody aggregates them across consecutive tournaments to see which way that data sequence is trending.

I call this the cumulative-load blind spot. A player can play four matches in five days at one event, fly to another continent, play three matches in four days, then fly again. That total volume is recorded nowhere. But the body records it.

A risk model saves nobody; it only tells you where to look.

Other factors go unmeasured too, and they tend to be treated as a player's private business. The psychological pressure of defending a large block of points over a few weeks. The strain of constant travel, hotel living, changing diets by city. Sleepless nights after a five-set loss. In 2026, a former Grand Slam champion withdrew from a major citing mental health, and the public reaction at the time showed that the sport still lacked a language for that kind of load.

None of this appears in any model. Yet all of it feeds directly into reaction time, into the ability to stay balanced on landing, into the decision to push one more step in a decisive rally.

The truth is that the body cannot distinguish between pain from mechanics and pain from stress. It only knows it is being asked for more than it can supply.

The limits of the model and the limits of the person

I want to be clear about one thing, because I have been wrong often enough to know I am not entitled to overconfidence.

Risk models do not predict injury. No model does, and anyone claiming otherwise is selling you something else. A model only establishes that within a set of similar cases, a certain pattern occurs more often than chance would suggest.

In 2026 I published a forecast about a player I believed would develop a hamstring problem within three months. He then played fourteen consecutive months without injury. I reviewed my entire methodology and found a flaw in how I handled the age variable: I had grouped everyone aged twenty-nine to thirty-four into a single band, when the decline in recovery capacity across those years differs sharply. Since then I have split age bands by single year.

The lesson was not to abandon the model. The lesson was to record where I went wrong, and fix it. I keep a separate ledger of my incorrect diagnoses — my "football case-file" — and it includes public corrections.

That brings me to a conclusion about tennis that I consider more important than the calendar.

The problem is not that there are too many matches. The problem is that we are pushing players through a dense calendar while collecting the bare minimum of data about them. A sport that does not know exactly how many hamstring strains its players have accumulated is in no position to declare that the calendar is the cause, or to identify the solution.

I do not believe in luck; I believe in numbers that have been verified. And when the number does not exist, I am not permitted to draw a conclusion. I am only permitted to say that we are missing data at precisely the most dangerous point.

What should happen next

If I had one request to make of the four bodies that govern tennis, it would not be to shorten the calendar. It would be this: build a shared injury registry, with one common set of definitions, mandatory reporting, and a mechanism that protects players from having their disclosure affect their competitive entitlements.

That sounds expensive. But this sport spends hundreds of millions on broadcasting three hundred frames per second and on ball-tracking systems. An injury registry costs far less, and could extend the careers of the very people who generate value for the whole system.

Until that happens, every time a player sits down on the bench in the third set with a hand on his calf, we will get the same familiar debate about the calendar, a few beautiful stat lines, and one empty file.

Data never lies; only the way we read it is wrong. But before we can misread it, we need something to read.

That night, leaving the stadium, I walked past the interview area. The player was standing there, smiling, saying it was only cramp and he would be ready for the next round. Nobody asked how many hours he had played in nine days. Nobody asked why he had withdrawn from an ATP 500 three weeks earlier. He walked away normally, and in the tournament's file that night, his name still sat on the list of healthy men.