Table Tennis' Data Gap: Nine Layers of Analysis, Eight Empty Cells

On Saturday night I opened a table tennis dataset on my machine. It was designed with nine analytical layers: technique and equipment, player data, event system, competitive landscape, rules and governance, coaching staff and talent pipeline, risk surface, public narrative, and industry transmission. I opened them one by one. Eight were empty. The only field with content was a two-word label: table tennis.

In the business of reading numbers, a result like that is usually treated as an input error, and most of the time it is. This time I could not call it an error, because I knew exactly what I was looking for, and I knew the thing I was looking for does not exist. I have spent more than twenty years reading sports data tables, ten of them tied to betting models in Chengdu. I am used to datasets that are full. I have never seen an elite sport this empty.

An empty cell is also a data point, provided we are willing to read it.

In 2026 I spent three months measuring PPDA — the passes an opponent is allowed before your team tries to win the ball back — across all 16 Chinese Super League clubs. Chongqing Lifan had the lowest PPDA in the league, 8.2, yet covered the handicap in 12 of 15 matches. I presented the conclusion that this side was better at ceding territory and counter-attacking than at controlling the ball. My boss dismissed it as a Western fad. I placed a small stake on the model, won 8 of 10 rounds, and from then on had my own database and one belief: a sport with data infrastructure lets you bet on the truth.

Football has that infrastructure. A single Premier League match generates millions of positional and event data points, plus xG, PPDA, possession sequences, pressure counts, sprint distances broken down by speed band. Now the uncomfortable part. Table tennis has no equivalent.

So I tried counting. An elite match lasts about forty minutes, split into five to seven games of eleven points each. Re-watching a WTT semifinal and tallying phase by phase, I logged roughly three hundred ball contacts across the whole match. Three hundred events. A football match can pass a thousand passes alone, before anything else is counted.

Three hundred events in forty minutes is why that empty cell is diagnostic. Table tennis produces far too few recordable data points, while each data point contains far too many variables to isolate.

The technique and equipment layer shows this most clearly. A topspin loop consists of spin rate, contact point, swing speed, apex height, racket-face angle. No sensor in an official match measures any of that systematically. Most of the information exists only in the eyes of the person standing across the table and in the memory of whoever re-watches the tape. Long pimples can turn an opponent's heavy loop into a nearly spinless ball, and no index records the conversion.

The legislative history of the ITTF shows a sport repeatedly changing its own equipment to cool itself down. In 2026 the ball grew from 38 to 40 millimetres. In 2026 the service rule forced the free arm away from hiding the ball. In September 2026 speed glue was banned. From the mid-2010s the 40+ plastic ball gradually replaced celluloid. Each time, the entire body of cross-era comparative data was silently invalidated, and no conversion table was ever published.

The player-data layer has the same problem. The ITTF changed its ranking system in 2026, moving to a best-results-over-twelve-months model. WTT launched in 2026 and brought a new points structure with it. For anyone building a model, the time series is cut into pieces, and every cross-era comparison needs a risk note attached. Head-to-head records from the last three years still carry reference value; head-to-head records across ten years carry almost none.

I still remember how a ranking actually operates. An Olympic berth is not decided by one match but by a sequence, with different weights at different tiers of event, and those weights shift by cycle. A player must both farm points and defend old ones before they expire. At this layer the data exists, but it is administrative data, not competitive data.

The competitive landscape is clear enough that no model is needed. At the Paris 2026 Olympics, China took all five table tennis golds: men's singles, women's singles, mixed doubles, men's team, women's team. Fan Zhendong completed a career Grand Slam in men's singles. Chen Meng defended the women's singles title. Wang Chuqin and Sun Yingsha won the mixed doubles. Ma Long closed his Olympic career as the most decorated table tennis player in Games history. On the other side, Sweden's Truls Möregårdh reached the final after eliminating Wang Chuqin, Félix Lebrun took bronze at home, Brazil's Hugo Calderano reached the semifinals, and Japan's Tomokazu Harimoto remains a permanent threat on the WTT circuit.

That dominance is not a data story. It is a pipeline story. China's system runs like a closed selection machine: thousands of children train inside a provincial network, the national squad keeps at least two successor layers for every slot, and internal pressure is so heavy that a world champion can still lose an Olympic place. No algorithm produces that depth. Correlation is not causation: Chinese table tennis does not win because of data, and the rest of the world will not close the gap simply by buying more software.

The risk layer also differs from football. Wrist, shoulder and lower-back injuries are the common cluster, and a minor injury is enough to destroy ball feel for weeks. The adaptation period after a rule or equipment change lasts years. The risk of being decoded does not sit in opponent data, since opponents hold no data on you; it sits in tape being analysed by hand by a handful of experts behind closed doors.

The public narrative layer is the most dangerous of all. In recent years Chinese table tennis has seen an unprecedented wave of fandom around a few young players. That heat brings views, sponsors, tickets. It also brings something else: demand for statistics. When real data is empty, the market manufactures fake data — unsourced comparison tables, percentage figures with no sample, charts built from a handful of clipped moments.

The industry transmission layer mirrors that gap. The equipment market sells blades and rubbers to tens of millions of recreational players but sells data to no one. WTT broadcast coverage carries a thin information layer, stopping at game scores and a few slow-motion replays. A sport with hundreds of millions of players is making most of its professional decisions by eye.

Here I have to turn against what I just wrote.

The comfortable explanation is that table tennis lacks data, and that more data would fix everything. I do not believe it. Table tennis lacks stable variables; it does not lack data in the technical sense. With roughly three hundred ball contacts in a match, every point carries too much weight. A ball clipping the edge, an umpire's service call at a decisive moment, a missed flick at 9-9 can decide an entire match. Football has ninety minutes for signal to overwhelm noise. Table tennis has forty minutes split into eleven-point blocks, where noise beats signal most of the time.

Copying football models into table tennis is wrong in kind, not wrong in technique.

The bigger trap sits on the opposite side. When a data layer is empty, what fills it is almost always story: character, form, competitive mentality, momentum. Those things are real inside the arena, but they have no measurable definition, and once they are attached to a spreadsheet they become the most dangerous object in my trade: a number with no source.

Data does not lie; we simply have not learned how to ask. A gap in the technique layer is genuinely a gap in the technique layer. Colouring it in does not fill it; it only removes the reader's ability to tell what we know from what we want to believe.

I stand on the side of the number, even when the number stands alone. I have also learned to stand on the side of the empty cell, even when the empty cell makes me look like I have nothing to say.

For the Vietnamese market this gap means something else. Holding up the Vietnam–China mirror, I see a paradox: Vietnamese table tennis has a deep grassroots base, a national tournament system, and players around the SEA Games medal bracket, yet almost no data layer is published regularly. Based on my experience following matches across both markets, most domestic player assessments still rest on collective memory and a handful of memorable games. A table tennis nation that wants to move up must build what football built: a searchable body of facts, with sources and dates, so anyone can verify instead of argue.

In the betting market, the consequence is concrete. Table tennis is a sport where odds swing hard inside the playing window, because bookmakers price mainly off live scores and money flow rather than a sufficiently deep technical model. For someone in my trade that is both opportunity and trap: opportunity because the market prices more slowly than the information, trap because the information needed to reprice is too thin.

The signals to track next cycle sit in three places: whether WTT publishes point-level event data covering spin type, placement and service sequences, or keeps stopping at game scores; whether any rule change on the ball or rubber is big enough to force a full revaluation of historical data; and whether smaller table tennis nations, Vietnam included, dare to invest in a data pipeline before investing in medals.

For the next forty minutes, when the next WTT semifinal starts, I will still be counting by hand. But I count to see which of those nine layers gets filled first. The answer to that will say more about table tennis than any ranking list.