The First Three Shots: The Biggest Data Gap in Modern Table Tennis

The First Three Shots: The Biggest Data Gap in Modern Table Tennis

In September 2026, in Verviers, Belgium, I sat in a national broadcaster's booth as the data analyst for the Table Tennis World Cup. The data table they handed me had seven columns: the score, the number of games played, the number of timeouts taken by each side, who serves next, the count of service faults, game duration, and the referee's name. Seven columns. Three of them were purely administrative and carried not one ounce of tactical information.

That same week, back in Munich, I was running a different table for a Bundesliga fixture. It had 340 columns. Among them: xG, xGOT, PPDA, progressive passes, field tilt, pressure regains, expected threat broken down by each eighth of the pitch. Seventy-two of those were derived variables I had written scripts to generate myself.

That 333-column gap is the entire subject of this piece. Table tennis — the sport I have played since I was eight, the sport that shaped my reflexes before I became a data analyst — is running on an information architecture roughly forty times thinner than football's. Every argument about modern table tennis, from who is the best player to who deserves an Olympic place, takes place on top of a dataset that is almost empty.

That is a structural problem, and it can be measured.

Context: two different data civilisations

When I say "seven columns", I do not mean it as a metaphor. I mean it as a CSV file. Between 2026 and 2026 I worked with both streams: the football feed from European providers, and the table tennis feed from ITTF event organisers. The difference was not technological. Both could register an event within 200 milliseconds. The difference lay in the question each system asked itself.

Football, from around 2026 onward, asked: how much scoring probability did this team create, and how did it create it? From that question came xG, then PPDA, then expected threat, then hundreds of derived variables. Each new variable was another way of re-reading the same sequence of events.

Table tennis asked: who won this game? And once answered, that question was treated as the end of all information needs.

The consequence is a paradox: table tennis has the densest event structure of any combat sport. A competitive best-of-seven can contain 400 to 600 individual rallies, each lasting on average 3.5 to 5 seconds, and each rally can be decomposed into at least six independent technical decisions. Multiplied out, that is roughly 3,000 technical decisions in a single match. Football, over 90 minutes, produces about 1,000 passes and 30 shots.

In other words: table tennis has a higher potential data density than football, but an extraction rate roughly two orders of magnitude lower.

I know this because I have sat on both sides of the wall. In 2026 I ran the data segment for both the Table Tennis World Cup and the Sudirman Cup in badminton. For badminton we had shuttle speed, trajectory height and placement distribution. For table tennis we had seven columns.

Core: five metrics table tennis is missing

1. The first three shots — and why they are everything

In table tennis, the first three shots (the serve, the receive, and the third ball) decide most points. Every national-level coach knows this intuitively. But intuition does not produce an index.

Consider a metric I call the FTI — First Three Shot Index. The method: for each player, determine the win rate on points ending within the first three shots, categorised by (a) serve spin type, (b) serve placement along the vertical and lateral axes, (c) the receiver's choice (short push, flick, backhand loop, forehand loop), and (d) conversion rate from the third-ball attack into a direct point.

Four variables. Those four variables split a dataset that already exists at every international event into at least 32 distinct cells. No tournament publishes them.

I tested manual FTI coding across two ITTF World Tour events in the 2026 season, working from video captured by two camera angles. The result showed something I believe matters: among the top eight players in the world at the time, the spread in win rate over the first two shots was only 4.1 percentage points. But the spread in win rate on the third ball after a short receive reached 19.3 percentage points.

In other words: serving skill has been almost flattened at the elite level. The gap is created by the third ball — the decision to attack a ball that is short, low, and usually heavy with backspin.

An index like FTI would turn this from a commentator's observation into a testable fact. And it would change how academies train: fewer hours of rhythm hitting, more hours of attacking off the short receive.

2. The rally termination index — table tennis's PPDA equivalent

In football, PPDA measures how many passes you allow the opponent before you engage. The lower the number, the higher the pressing intent. Japan's PPDA of 6.2 in 2026 was not accidental; it was a manifesto written as a number.

Table tennis needs an equivalent. I propose the RTI — Rally Termination Index: the average rally length a player accepts, measured in ball-to-table contacts, split by serving and receiving situations.

The logic is simple. A player with an RTI of 3.4 on serve is declaring that he wants the point finished in four contacts or fewer. A player with an RTI of 6.8 is declaring the opposite: he trusts endurance and waits for errors.

This is the kind of data that can be extracted automatically. It requires no new technology. It requires only a decision: that long or short rallies are information, not dead time.

While tracking matches at WTT Champions Frankfurt and rounds of the German national championship, I recorded a fairly stable pattern: European-born players carry an average RTI about 1.2 contacts longer than Asian players of the same ranking tier. But when they face each other, the player with the lower RTI wins 61% of the points in the first and second rally of each game.

I say "fairly stable" because my sample is only about 40 matches. That is a limitation I will return to at the end.

3. Points defence pressure and the 52-week rolling mechanism

The WTT ranking system operates on a 52-week rolling mechanism: a tournament's points expire after exactly one year, and a player must replace them with new results or lose them.

This mechanism creates a pressure that can be quantified, yet is almost never quantified. I call it PDP — Points Defense Pressure, and the formula I use internally is:

PDP = (total points expiring in the next 90 days) ÷ (maximum points obtainable in the next 90 days, based on the registered tournament calendar)

A PDP above 1 means a player is forced to compete more than they would like, at events they do not need, simply to hold position. A PDP below 0.4 means a player has the luxury of a selective schedule.

Each season, roughly ten to fifteen players inside the world's top 30 fall into a PDP above 1 for at least one stretch. That group shows higher injury rates and more early exits at majors — but nobody publishes this number, so every debate about a "decline in form" proceeds without its most important variable.

4. The equipment adaptation window

In 2026, the ITTF moved from the 40mm celluloid ball to the plastic 40+ ball. It was a physical shock: bounce, spin and flight speed all changed. Many players needed more than a year to rebuild their technique.

Yet no public dataset measures this adaptation window. No index answers the question: how many matches does a player need to return to their pre-change performance level?

From internal data I could access between 2026 and 2026, the window ranged from 8 to 47 matches depending on playing style. Heavy-spin players were affected for longer than speed-based, close-to-the-table players. That makes physical sense: the plastic ball carries a lower spin coefficient, so a player whose game is built on spin needs more time to restore order.

An equipment rule change without accompanying adaptation data is a change that cannot be evaluated. That is why debates about the 40+ ball still rest mainly on personal memory.

5. The age curve and the China-versus-the-rest map

At Paris 2026, China won all five gold medals. Fan Zhendong beat Truls Moregard in the men's singles final. Felix Lebrun took bronze in the men's singles on home soil. Wang Chuqin was eliminated early by Moregard.

But reading results tells us nothing about the future. To read the future, you need an age curve per technical metric, not per biological age.

This is where I believe table tennis is missing the most. In football, we know sprint speed peaks at 24–26 while decision-making peaks at 28–31. For table tennis, we know nothing equivalent.

At what age does reaction time peak? Spin reading? The ability to hold up under pressure at a deciding point? There are no answers grounded in public data, because there is no public time-series data long enough to answer them.

Meanwhile, global competition is shifting faster than our ability to describe it. Japan has produced a generation capable of beating anyone on a given day. Sweden has a player who has reached a world final. France has a pair of brothers playing penhold with unusually high attack indices. Brazil has a player regularly inside the top 10. That picture needs data to be described, and it is currently being described by commentary.

The contrarian angle: when data is empty, stories fill the void

This is the section I want to spend the most time on, because it is the least discussed.

When a system does not produce data, it does not leave a void. It leaves a void that something else will fill. In table tennis, what fills it is the language of character, of heart, of the moment of brilliance.

I do not deny the existence of those things. I deny our ability to test them without data. And once a quality cannot be tested, it gets distributed by sentiment — usually according to a player's fame rather than their actual contribution.

Fate was written in advance — we simply need enough data to read it.

But here a second paradox appears, and I want to state it clearly because it argues against my own case. More data does not automatically produce more understanding. Table tennis has a severe sample-size problem that football does not have to the same degree.

Take a best-of-seven that ends 4–3. That match contains roughly 70 to 90 points. Across those 90 points, the difference in point-win rate required to produce the result is only about 3 to 4 percentage points. At a 95% confidence level, a sample of 80 points cannot distinguish a player winning 52% of points from one winning 48% at statistical significance.

Put plainly: the majority of elite table tennis matches have results that sit inside the noise band.

That does not make results meaningless. It means every conclusion drawn from one match needs an error label attached. And it means that a table tennis analytics discipline, if built, would have to operate differently from football's: on rolling windows of 20 to 30 matches rather than on single games.

I have made this mistake myself. In 2026 I published an analysis of Japan's PPDA at the football World Cup. It was directionally correct, but I presented it with more certainty than the data allowed. That piece reached 1.2 million views. Nobody read its error bars.

Since then I have applied a rule: every conclusion I publish must come with an explicit statement of the model's limits. It makes the writing less appealing. It also makes it more correct.

The summer of 2026 emptied stadiums but filled data tables — it turned out football had been missing something. When the Bundesliga restarted on 16 May 2026, home win rates fell from 42.4% to 24.7% across the remaining 81 matches of the season. I sent a recommendation to a client club fighting relegation: push your pressing line higher away from home. They won four of six away matches and stayed up.

But the real lesson of that summer was not about pressing. The lesson was this: a variable everyone treated as a constant turned out to be a variable. The crowd. The noise. The social pressure inside the stadium.

Table tennis has never had a summer of 2026 of its own. Tournaments continued, but under no-spectator conditions, and nobody — as far as I know — recorded the shift in server win rates, rally lengths or service fault counts during that period.

That was a natural experiment wasted. And it will not come back.

My own blind spots, stated publicly

I must be explicit about three limitations in everything above.

First, my sample is small. The RTI observations rest on about 40 matches I coded by hand. Their standard errors are far larger than I could comfortably present.

Second, manual coding contains subjectivity. When I classify a receive as a "flick" or a "short push", I am making a judgement. Someone else might classify it differently. The FTI is only trustworthy once classification is automated and cross-checked.

Third, and most importantly: I am imposing a football framework on table tennis. That may be wrong. Table tennis may not need xG. It may need an entirely different metric set that I have not thought of, precisely because eighteen years inside the football ecosystem have shaped me. Read from the top of the stats sheet, the 2026 World Cup turns out to be a poem written in PPDA — but table tennis may need a different poetic form.

I publish these blind spots for professional reasons: a model that does not state its limits is a model that cannot be falsified. And a model that cannot be falsified has no scientific value.

Takeaway

Japan proved that pressing is not instinct, it is an arithmetic exercise. Table tennis is waiting for someone to prove the same about the third ball.

The first federation to publish an open dataset on first-three-shot outcomes, split by spin type and placement, will restructure how this sport is coached within five years. Not because that data contains truth. Because that data will end the situation in which every argument finishes as a story.

The summer transfer window is nothing more than a slower version of the stock market: numbers decide, not rumours. Table tennis does not yet have that market. Table tennis is trading on belief.

The question I leave is not for the federation but for the viewer: if tomorrow you were handed the FTI of two players before a final, would you still want to hear the commentary about character?