The Empty Data Room: Why the Best Tennis Analysts Sometimes Say Nothing

On a Tuesday night at Indian Wells, I stayed behind in the data room after the sound of the rackets had died. The screen showed the stat sheet of a quarterfinal: first-serve points won at 78%, fourteen net approaches, return points won at 31%. With those three numbers, anyone could build a tidy conclusion in two lines of a tweet — that this player won on serve, or lost on the return.

The context column was empty: surface type, wind direction, fitness after five sets, opponent quality, and whether the man across the net was nursing a wrist. I closed the file, shut the machine, and wrote nothing. That silence was a calculated decision. In tennis, what is more dangerous than missing data is acting as if you already have enough. When the market laughed at Salah, the data nodded quietly; and in the other direction, when a pretty stat sheet is pushed onto the front page, the data also has the right to stay silent and refuse to sign it.

Over roughly fifteen years, professional tennis entered an era in which almost every ball strike leaves a trace. The ATP and WTA publish serve data, return depth, rally length, the receiver's position. The Grand Slams added electronic tracking that measures spin rate and landing point to the centimetre. The volume of raw data grows exponentially, but the ability to interpret it does not grow at the same pace. That gap is where distorted conclusions are born.

Based on my experience watching matches, the heaviest pressure on a tennis analyst comes not from a shortage of numbers but from the expectation of always reaching a conclusion. Newsrooms need headlines. Fans need answers. Algorithms need fresh content every day. Inside that churn, the sentence “there is not enough information to conclude” becomes a hard product to sell, yet it is usually the most honest product in the room.

The chain of evidence and the trap of a single metric

In 2026, I wrote a 3,000-word analysis of a winger moving from Serie A to the Premier League for 42 million euros. His shooting and box-entry numbers sat in Europe's top 5%. I concluded he would score more than 30 goals. He scored 32. But in that same piece, I predicted a midfielder worth 45 million pounds would dominate his new team's midfield, and he faded all season. The data did not lie. I was the one who ignored the most important variable: the role the manager handed him.

That role variable is harsher still in tennis. A player can keep every serve metric identical and still flip the result because of one small adjustment: standing half a metre deeper to receive the second serve, or shifting the serve position from the middle of the box to the wide corner. On the stat sheet, the points-won rate rises. Look closer, and you understand why.

I once erred that way with a sequence of data that looked harmless. The sample ran to three matches. Three matches are far too few to say anything with weight about a player. But three matches are enough to manufacture a headline. A player wins three straight on hard courts and the media declares he is “back.” In reality, to separate a genuine resurgence from a friendly run, I need at least twelve to fifteen matches, cross-checked against opponent quality, and a look at whether the surface suits his game.

Here I apply a rule: before quoting any number, write one short sentence about the context in which it was collected. A 78% first-serve points-won rate on the grass of Halle means something entirely different from 78% on the clay of Monte Carlo. The altitude of Mexico City makes the ball fly faster, turning an ordinary serve into a weapon. Indoor courts kill the wind, giving big servers an edge they lack outdoors. The same number, three contexts, three opposite conclusions.

Another example sits at break point. A player can post an impressive break-point conversion rate at one event, then collapse at the next. The cause rarely lies in the number itself. It lies in the opponent serving better at the decisive moments, or in the player grinding through several long matches and losing his legs. A single metric cannot tell the whole story. That is why I reject any conclusion built on one line of data.

Surface adaptation is the clearest example of context having to come before the number. A drop shot from Carlos Alcaraz can be a devastating weapon in dry conditions, but it loses much of its value in strong wind. Iga Swiatek's clay-court dominance has been proven across many seasons, yet every time she switches to grass, her numbers have to be read from scratch. Novak Djokovic's famously deep return position works only while his fitness and reflexes can cover the extra ground. A common formula for all three is hard to find.

Multi-layer verification: three sources, two systems

After the 2026 World Cup episode, I built myself a multi-layer verification process. I had once published that one team generated only 0.8 xG while its opponent generated 2.1, then used that number to say the winner did not deserve to advance. The community pushed back hard, and they were right on one point: football operates differently from a computer simulation. I had to withdraw, rewatch every penalty shootout, and discover that the winning goalkeeper dived to one side 2.3 times more often than the other. Only then did I understand that my data had never recorded the thing that decided the match.

From then on, I stopped using phrases like “deserving” or “undeserving.” I replaced them with probability descriptions: that team won inside a sequence of events carrying a probability of roughly 18%, and the rest is what my data could not yet explain. This phrasing is less attractive on a headline, but it holds up when tested.

For tennis, my process stacks three layers. The foundation layer is raw data from the official tracking system. Laid over it is a second independent system, used for cross-checking when the two diverge by more than 5%. On top sits my own direct observation, through video or through matches I watch myself. When all three layers agree, I allow myself to write one conclusion carrying a probability level. When the first two conflict, I stop and record the conflict rather than picking the side with prettier numbers.

My sufficiency threshold is set before I write, not after. Three independent sources, or two data systems plus one visual confirmation. If the threshold is not met, the piece carries a section called “data limits.” That section is not for self-defence. It is a map showing the reader where my conclusion might collapse.

The counter-intuitive angle: correlation is not causation

What keeps me awake is not the lost matches but the correlations so pretty they look suspicious. A player changes coach and wins repeatedly. The media instantly credits the new coach. Yet in many cases that winning run coincides with a swing onto the surface that best suits the player, or with a lucky draw after the strong seeds fell early. Correlation arrives first; causation is only a hypothesis that comes later.

I have learned to doubt my own beliefs. When I want to believe a coach produced the turning point, I force myself to find at least two other explanations that do not involve him. If the data still points to the coach as the main factor, only then do I write, attaching a probability level instead of a declaration. The truth lies deep beneath the stat sheet, where a headline never reaches. The market forgets nothing; it merely disguises itself as a new summer.

The biggest blind spot in tennis analysis sits here: we have plenty of data about the shot but too little about the person. No metric measures a player sleeping four hours from anxiety, or going through a heartbreak, or a court so hot it changes the whole feel of the ball. Those variables live outside the spreadsheet, yet they often decide the result.

The signal for the next round

In the coming weeks, I will track a signal few people notice: the share of analyses that have to admit their own data limits. If that number rises, the industry is maturing. If it stays at zero, we are still producing headlines prettier than the truth. Fans look with their eyes; I look through a probability distribution. Both of us deserve to know when an answer still lacks the data to exist.