Table Tennis, Fourteen Stat Columns and Nine Empty Cells: The Data Doesn't Lie, Only the Reader Hasn't Been Honest Enough

On a late-June evening, I opened the data table for a men's singles quarterfinal on the WTT circuit and counted fourteen stat columns. Nine of them were empty. Not empty because the match had not been played, but empty because nobody had recorded them. Win rate on rallies longer than seven strokes: absent. Service-point win rate broken down by landing zone: absent. Third-ball attack efficiency: absent. Three days later I tried to rebuild that same match from video and spent four hours and twenty minutes to produce a table that was merely usable. An elite table tennis match lasts about forty-five minutes. I needed six times the length of the match just to learn how the match had actually unfolded.

That is the entire story of this sport, compressed into one sentence.

Table tennis is the fastest information-processing sport in the combat category. The ball can leave the racket above one hundred kilometres per hour, the distance between the two rackets is under three metres, and an elite rally ends in less than a second. Human eyes are therefore not fast enough to record it, and every stat table a spectator sees is the output of a manual or semi-automated coding process behind the scenes.

The problem is that this process is not standardized. In football, the major data providers have built a column system in which two people on two different continents, coding the same match, arrive at the same number. Table tennis has not. WTT runs its own data system for the top-tier events, but it stops at overall point-win rate, service points won, and a few aggregate figures. Below the top tier, where hundreds of young players are trying to break into the top one hundred, almost no data is recorded in a reusable way.

I have followed WTT-circuit matches across four consecutive seasons and logged them by hand in my own spreadsheet. First-hand tracking gave me three observations that the official tables never show.

First, a player's true value in table tennis lies in the zone where his service-point win rate is lowest, not highest. A player serving sidespin into the opponent's forehand-side zone may hold a win rate above fifty per cent. But when he is pushed into the zone where he must serve backspin just to stay safe, that win rate drops below thirty-five per cent. That weak zone is what decides the next round, once opponents have watched the footage. An aggregate table cannot distinguish a player who serves evenly from a player with a single weapon he manages to hide all tournament.

Second, average rally length is the most misread metric in the sport, because it is a tactical signal, not a physiological one. People habitually read it as a fitness gauge. But when I split my data by set, average rally length rose in the fourth and fifth sets not because both men were exhausted, but because one had switched to short serves and blocking. Reading this signal wrongly leads to wrong conclusions about a player's conditioning, and to wrong rotation decisions in team events.

Third, and this is what consumed most of my time: roughly fifteen per cent of points in an elite table tennis match end without any genuine attack. Rallies finish on service faults, receive errors, or edge balls. A table that records only "point won" folds those points into the same column as points built through three consecutive attacking strokes. The result is that a patient counter-attacking player looks identical to a player who simply pounces. Two people, one column.

There is one more variable no public dataset tracks: the racket surface. In table tennis, a player can change his blade or rubber mid-season, and his technical profile shifts almost instantly. Serve spin drops, block control rises, and every accumulated statistic before that point becomes the data of a different person. I once recoded three matches for a player before and after he changed his left rubber. His service-point win rate fell eleven per cent, while his win rate in long rallies rose. Nobody logs the date of a rubber change. So when I read a season-long cumulative table, I always ask myself how many people I am reading on a single row.

I rebuilt this chain of evidence for a group of young players aged nineteen to twenty-two. Of the eighteen players for whom I had enough video to code, nine had a higher head-to-head match-win rate while their overall point-win rate was lower. That means they won by picking the right moment, not by dominating. The overall point-win column cannot see it. Nor can the ranking. And a scout holding only the ranking will skip precisely this group.

This is where a paradox appears that I have not fully resolved. The table tennis analysis community is running in two opposite directions. One camp demands more data: more metrics, more sensors, more models. The other says data ruins the feel of the sport, that table tennis is a matter of instinct and reflex, not spreadsheets.

Both are right in what they say and wrong in what they omit. Correlation is not causation, and in table tennis this trap is more dangerous than in any other sport, because the clean sample is so small. A player can win six straight matches with a high service-point win rate and people immediately conclude that serving is his weapon. But if across those six matches he never met a good receiver, then what was measured was not ability but luck in the draw.

When I cross-checked matches between players of the same level band, the correlation between service-point win rate and match-win rate weakened markedly. Against the best receivers, the metric barely separates winners from losers. What separates them is the ability to hold rhythm through the first three strokes of the receive game, a metric nobody records.

The data doesn't lie; only the reader hasn't been honest enough. And honesty here does not mean publishing more numbers, but publishing the empty cells too.

I do not think table tennis needs to become a sport of prediction models. I think it is being mispriced, and the cost lands on those with the least voice: young players in low-tier events with no data to prove they are good, no metric for a foreign club to see. An open, standardized data system at the base of the pyramid would change who gets discovered, not just how we watch a match.

The next round of the season starts in late August. I will sit down again with fourteen stat columns and again see nine empty cells. The question is not when someone fills them. The question is who will be the first to understand that those empty cells are where the real player can be read.

There are evenings when I sit with numbers longer than with people, and I have never felt lonely. When the stadium is empty, player behaviour finally tells the truth. In table tennis, when nobody records the metrics, player behaviour tells the truth in exactly the same way.

People ask me whether girls watch football. I answer with ninety-two pages of data. For table tennis, the number will be fourteen columns and nine empty cells.