Tennis Can Measure Everything Except What Decides the Match
**Câu trả lời cốt lõi**: Dữ liệu quần vợt hiện đại mô tả rất tốt nhưng giải thích rất kém, vì các chỉ số tổng hợp bỏ qua trọng số điểm, thời gian giữa các điểm, vị trí tr
On June 8, 2026, on Court Philippe-Chatrier, Carlos Alcaraz walked to the baseline at 0-40 down. Three championship points for Jannik Sinner. The clock had already passed the fifth hour, and the Roland Garros final that day would close at 5 hours 29 minutes, the longest in the tournament's history. In that moment, two weeks of data collected by two analytical teams became as light as a sheet of paper.
Alcaraz saved all three championship points. He won the fourth-set tiebreak 7-3, won the fifth-set tiebreak 10-2, and closed the match 4-6, 6-7(4), 6-4, 7-6(3), 7-6(10-2).
The next morning I sat in front of three screens in a short-term rental in Liverpool and opened the official stat sheet. Hundreds of cells. Average first-serve speed. Points won on first serve. Winners. Unforced errors. Distance covered. All of it full, detailed enough to reconstruct every rally geometrically.
Then I realised the one thing I actually wanted to know — what Alcaraz was thinking during the forty seconds between the third and fourth points of that game — was not in any cell.
A perfect stat sheet. The empty column stayed empty.
A sport measured to the millimetre
Tennis is one of the earliest and deepest digitised sports. Electronic line calling arrived at Wimbledon in the mid-2000s, spread across the majors, and by the 2026 season nearly the entire ATP Tour and all four Grand Slams operated on electronic calling across every surface. Every serve is logged for speed, landing point, estimated spin, receiver position. Every rally is rebuilt as a string of coordinates.
Based on my experience covering ATP and Grand Slam matches over nearly a decade, I can say the volume of data a professional analytics team can access today is many times what it was when I started working with StatsBomb sheets in football. And like football, tennis is falling for an illusion: that measuring more things means understanding the match better.
That illusion has a price.
According to figures published by the US Open organisers, the 2026 men's singles champion's prize money reached 5 million USD, the highest in the event's history. The ATP's sponsorship deals with Saudi funds were announced in February 2026, and the WTA Finals moved to Riyadh from the 2026 season. In March 2026, the Professional Tennis Players Association filed suit against the ATP, WTA, ITF and ITIA. In February 2026, Jannik Sinner received a three-month ban, from February 9 to May 4, after WADA appealed the initial ruling over two clostebol positives at Indian Wells in March 2026.
Every event on that list was recorded, published, debated. Every event generated data. And every event exposed the limits of that data.
At surface level the story is even clearer. From the 2026 season, most Masters 1000 events run twelve days instead of seven or eight. The calendar thickens, the matches multiply, the data multiplies. But the question of whether someone should withdraw from a tournament still depends on waking up one morning with a feeling in the left knee — something no sensor measures.
The three most glamorous metrics and their cost
Every tactical diagram is an orderly lie — I go looking for the truth behind it.
The most abused metric in tennis coverage is first-serve percentage. It appears in every stat sheet, every press conference, every on-screen ticker. But it is an average over a wildly non-uniform set. A player serving 62% in the first set and 71% in the fifth has the same average as one serving 66% throughout — and those are two different psychological states, two different fatigue levels, two different stories.
When Alcaraz stood at 0-40, his match-long first-serve percentage meant nothing. What mattered was his first-serve percentage across the next four points, under that specific pressure, after five hours. The standard sheet does not answer that.
The second metric is the winner-to-unforced-error ratio. It is used as a form gauge, and that is a category error. A winner can be the product of a brilliant shot, or of an opponent guessing the wrong direction, or of an opponent forced into a weak reply after being dragged off court the ball before. The sheet records one plus sign for three different causes. Unforced errors work the same way: an error after a twelve-shot rally signals exhaustion; an error immediately after a serve signals a lapse in focus. Same minus sign.
The third is total points won. It sounds like the fairest measure, which is exactly why it is the most dangerous. Total points assumes every point is worth the same. In tennis, they are not.
Points are not equal
This is where I want to slow down, because it is the foundation of everything else here.
In athletics, every metre is a metre. A 400m runner finishing in 44 seconds and one finishing in 46 differ by exactly two seconds, no more, no less. That sport has no concept of a "significant second". But when I moved to reading tennis matches, I realised I had to carry a different concept with me: the leverage of each point.
A break point at 4-4 in a deciding set is worth something entirely different from a break point at 1-4. A point at 30-30 differs from one at 40-0. Professional data analysts know this and have models for weighting points — usually called leverage or expected points. But those models rarely surface in mainstream coverage, because they are hard to express in a single line.
The result is that audiences read flat stat sheets where every point counts the same, then draw conclusions about a match decided by precisely the points that are not counted the same.
In the 2026 Roland Garros final, Sinner played better than Alcaraz for most of the first two sets. He controlled the rhythm, served accurately, and pulled the match onto his preferred track. Look only at the aggregate numbers after two sets and you see a very clear picture. The problem is that tennis is not scored by the sum of sets, but by specific points at specific moments.
Three championship points in the fourth set were the three highest-leverage points of the entire tournament. No aggregate metric reflects that.
Gap one: twenty-five seconds nobody reads
The 25-second serve clock arrived on the ATP Tour in the 2026 season. It is one of the most structurally consequential changes to modern tennis data, and almost nobody exploits it.
Every changeover between points produces a measured interval. It is recorded. It exists in the system. And it is the most honest physiological fingerprint a player leaves on court, because it is not chosen by anyone — it is the result of how long the body needs to return to serving state.
A player taking 18 seconds early and 24 seconds in the fifth set is telling you something first-serve percentage cannot. A player suddenly stretching to 25 seconds after a 30-shot rally is telling you something about their tolerance threshold. A player holding 16 seconds across five hours is telling you something about a physical base no column accounts for.
I tested reading this data across several big matches in the 2026 season and found a fairly stable pattern: between-point time rises before the technical metrics decline. The body gives up the rhythm before the shot gives up its accuracy.
That is the kind of signal a documentary writer covets, because it can be told as a visual story. It is also the kind of signal a standard stat sheet erases, because it has no slot in the traditional analytical frame.
Gap two: the feet are not in the model
Electronic line calling records every player's position at every moment. That data exists. But what gets published to the public is almost entirely ball speed, landing point and direction.
There is one variable almost never fed into mainstream analytical models: where the receiver stands on a second serve.
This is where most of the modern tactical war is fought. A player standing two metres behind the baseline and one standing four metres behind can post identical second-serve return win rates while playing two different sports. The closer player accepts risk to seize control; the further player concedes control to reduce risk. Aggregate metrics cannot distinguish the two choices, while an opposing coach distinguishes them instantly.
I remember an evening at a hard-court Masters event, sitting beside an analyst working for a top-20 player, asking what kept him up at night. He did not mention first-serve percentage. He mentioned an opponent changing second-serve return position in the middle of the third set without changing his win rate — meaning the opponent was experimenting, preparing for something coming in the fourth.
A stat sheet cannot see preparation. It only sees outcomes.
Gap three: intent
There is a behaviour in professional tennis no metric captures, which is why it is the most argued-about: the decision to concede an opponent's service game in order to invest in your own.
A player is in the third set, leading 4-3, opponent serving. He can choose to go all-in and risk the expenditure, or to play within himself, hold rhythm, and prepare for the decisive service game. Both choices are rational. Both produce rallies that look identical on a stat sheet.
You cannot distinguish a missed return from fatigue from a missed return from intent. You cannot distinguish a serve lost to a lapse from a serve used as an experiment. The system records both the same way.
This is why I tell my editors that any tennis analysis built purely on data has a hole in the middle. That hole has a name: motive.
The lesson of the pressing scanner
In 2026 I was 18, a first-year student in Liverpool, and I built a YouTube channel for tactical analysis. I used data to argue that Roberto Firmino was not a false nine in the ordinary sense but a "pressing scanner" — a player whose value lay in breaking the opponent's build-up structure, not in goals.
My video showed Firmino making 23 pressing actions in a Champions League match against Manchester City, far above the average for a winger in the same game. It ran twelve minutes. It caused an argument. Some called me a tactical vandal. It reached 40,000 views in a week.
What I learned from that was not that data has power. What I learned was that data only has power when a hypothesis stands in front of it. Had I started by counting pressing actions and then gone looking for a story, I would have had no video at all. I started with a counter-intuitive question — if Firmino does not score, why does Liverpool play better with him? — and then used data to answer it.
That method transfers to tennis, and it is why I moved to this sport.
Modrić and the limits of fitness forecasting
The 2026 World Cup taught me that arrogance is an own goal nobody saves.
Before the Croatia-England semi-final I wrote a prediction that Croatia would lose for lack of young legs, having played two consecutive 120-minute matches, with an older average squad age. I had the numbers to back it: distance covered, minutes played, rest days.
Croatia won 2-1. Luka Modrić played the match of the tournament, not by running more but by running in the right places. He reduced distance and increased the quality of every movement.
I did not take the piece down. I hosted a livestream to analyse my own mistake in front of about 300 viewers and posed an open question: is fitness really more important than positional intelligence? The debate ran two hours.
The lesson I carried into tennis is this: every fitness-based forecasting model assumes energy is spent evenly. The best players in the world do not spend evenly. They spend deliberately. And intent, as established above, is not in the model.
The biggest gap: injury data
If there is one area where tennis is deceiving itself most, it is injury.
My view, after years of reading medical reports and tracking calendars: scheduling density is the single biggest cause of injury, and no medical team saves a player who has to play two matches a week for ten straight months.
But the data to prove it is the most closely guarded in the whole system. Tours publish calendars. They publish prize money. They publish sanctions. They do not systematically and transparently publish injury rates by minutes played, by surface, by age, by cumulative schedule load.
A player withdrawing with a "wrist issue" might be resting preventively, hiding something worse, or choosing events that fit season goals. The public bulletin records all three identically.
The consequence is that every debate about whether someone should play or rest happens in the dark, with figures selected by the parties themselves. Fans get a simplified version: player X is fragile, player Y lacks hunger.
The real problem sits in a packed calendar designed to maximise broadcast revenue.
The counter-intuitive angle: the analytics industry is selling the wrong product
I do not sell predictions; I sell hypotheses. There is an ocean between the two.
The modern sports analytics industry, at most levels, sells predictions. Data sites give you win probabilities. Pundits give you outcomes. Models give you percentages. Everything points at one goal: saying what will happen.
But prediction is the most perishable product in this entire industry. One miss and it loses value. And in a sport where three points in the fourth set can overturn five hours, the miss rate is always high enough to collapse any model.
A hypothesis is different. A good hypothesis does not need to be right to be valuable. It needs to explain. It needs to point at a mechanism. It needs to make the reader see the match differently afterwards.
This is why I write documentaries in a prediction-collapse-reflection structure. I put forward a strong hypothesis, let it collide with reality, then analyse the gap. That gap is always the most interesting part.
It is also why I think most current tennis analysis wastes its own potential. We have the data to tell stories nobody has told. We use it to predict who wins.
Arguing against myself
There is an obvious objection to everything above, and I want to raise it before someone raises it for me.
If data is insufficient, and if the gaps matter more than the filled columns, am I not advocating vagueness? Am I not offering an excuse to ignore numbers and retreat to the gut feeling of a television viewer?
There is some truth in that objection, and I accept it.
But there is a difference between "data is insufficient" and "data does not matter". Data is the foundation. Without it, all analysis is storytelling. What I object to is stopping at data and declaring the work finished.
Take the 2026 Roland Garros final. Without data I would not have known Alcaraz changed his contact height on second-serve returns in the fourth set. Without data I would not have known Sinner's between-point tempo rose sharply from the eighth game of the fourth set. Data is what walks me to the right room. It just cannot open the door.
What I propose is a clearer division of labour: data describes and locates; humans interpret and hypothesise. When the two roles blur, you get analysis that reads beautifully and says nothing new.
The industry blind spot: data buyers without data readers
There is a structural problem rarely discussed: academies, national federations and Davis Cup and Billie Jean King Cup teams are buying more data than ever, but they lack people able to read it under live conditions.
A data sheet is only worth something when it reaches a person who can decide within thirty seconds. In a tennis match, thirty seconds is all you have between games. If your analytical process needs fifteen minutes to produce an insight, that insight arrives after the match.
The best teams in the world therefore do not hire the best analysts. They hire the fastest communicators. The gap between those two profiles is far larger than academies assume.
I have seen this from the other side: working with documentary producers taught me that a good idea not pitched in thirty seconds does not get funded, however deep it is. Tennis runs on the same law, except the penalty is not a rejected budget but a lost service game.
What will change in the next few seasons
Three directions will reshape how we read tennis over the coming seasons, and all three involve filling the gaps described above.
First, between-point time will become a standard metric rather than a by-product. Once tournaments recognise it as the earliest indicator of physical decline, it will appear in broadcast graphics the way first-serve percentage does now.
Second, positional data will be exploited at tactical rather than descriptive level. We already have coordinates for every player in every rally. We do not yet have models that tell the story of how a player restructured his movement across sets.
Third, and the one I care about most, is transparency in injury data. If tours published cumulative minutes linked to injury rates by age group, the scheduling debate would change in kind. It would move from opinion to evidence.
None of the three requires new technology. They require decisions to publish what is already measured.
What remains after all of it
There is a reason I chose tennis after years in football, and it connects directly to this piece.
Football is a sport of systems. You can hide a weak player inside a well-organised pressing block. You can win a match without anyone playing well. Structure can compensate for the individual.
Tennis cannot. On court there is no teammate covering for you. When you are at 0-40 and have to serve, every system vanishes, and what remains is one person, one racket, and one decision.
Data can tell you who has the best break-point save rate of the season. It cannot tell you what that person is thinking at the baseline.
That is why tennis remains the sport I want to write about. It is the sport where every model must finally bow to a moment no model contains.
A note from the drawer of unfinished ideas
In 2026, when stadiums stood empty, I began a project called Arena Ghosts: recording the sound of three amateur grounds in Liverpool — wind, a rolling ball, players shouting in a space with no crowd.
I enlisted two friends. Two months later I abandoned it for an esports idea and left them stranded. Producer Sarah James happened across a short clip I posted and got in touch. She said I had an unusual eye.
The project failed. But it taught me something I have applied to every analysis since: what does not get recorded is often what deserves recording most.
Arena Ghosts was not cancelled — it is only waiting for a season brave enough to tell it.
And the stat sheet from the 2026 Roland Garros final, hundreds of cells full, is also waiting for someone to read the empty cell beside it.
In the next few years there will be writers on tennis who no longer ask "who won", but "what went unmeasured in this match". When that question becomes standard, we will have a new generation of analysis — not the generation of prettier tables, but the generation of bolder hypotheses.
That is the generation I want to belong to. And if you are holding a stat sheet, look for the empty cell before the full ones.
