Full Scoreboard, Empty Data Column: Lessons from a Billiards Night in Hai Phong

Late on a Saturday night, table four for three-cushion at a billiards club on Lach Tray Street, Hai Phong. Two amateur players went into the thirty-fourth visit tied at 38-38. I sat on a chair beside the table with a notebook already ruled into four columns: the visit number, the score after each visit, the time each visit took in seconds, and the opening ball of that visit along with the first cushion it touched.

The match finished 40-38 after forty visits. When I closed the notebook, the last two columns held seventeen rows. Twenty-three visits were blank, because I had spent more time watching the balls run than watching the clock.

What I did next is the part worth reporting: I published nothing about that match. No comparison table, no prediction for the rematch. I was holding a file that looked a great deal like data, and inside it was a hole. For anyone who works in betting analysis, staying silent after a night like that is far harder than putting out a wrong conclusion.

Vietnamese billiards lacks data at the source

Vietnam is strongest in carom three-cushion. In 2026, at the UMB World Three-Cushion Championship in Cairo, the final was an all-Vietnamese affair between Tran Quyet Chien and Nguyen Quoc Nguyen; Tran Quyet Chien took the title. That event put Vietnamese carom on the international map, and since then players from the country have regularly been seeded at major events.

The data that travels with the results has not kept pace. At most amateur and semi-professional events in Vietnam, organisers publish the final score, sometimes the score of each game, but almost never visit-level data. There is no official table showing a player's average per visit, their longest run, or the position an opponent inherits after a safety exchange. Spectators know who won. They do not know how.

Meanwhile each billiards discipline uses its own vocabulary, and the words do not translate cleanly. In snooker, a break is an entire scoring visit, one that can last dozens of minutes. In pool 9-ball, the break shot is the opening shot, and a break-and-run means winning the rack straight from it. In carom three-cushion the unit is the visit, and the two root metrics are the average per visit and the longest run.

When two people use the same word to measure two different things, the data is not wrong — the reader is. That is the first reason I wrote nothing about that night: my opening-ball column had no agreed definition, so even with all forty rows filled it could not sit beside any other statistical table.

One more technical difficulty: in carom three-cushion, the number of visits in a match is not fixed the way the number of racks is in pool, because it depends on who holds the table. A match that ends 40-38 can take barely more than thirty visits if both players sustain their runs, or run past sixty if both keep playing safe. That means the average per visit, the root metric of the discipline, is shaped directly by the rhythm of the match, not purely by skill.

A correct metric can still be useless

In 2026, at seventeen, I applied expected goals to a V.League match between Hai Phong and Sanna Khanh Hoa. The numbers gave Hai Phong 2.8 goals worth of chance quality and their opponents 1.0. I predicted 3-1. The match ended 0-1, and goalkeeper Tran Buu Ngoc made seven saves.

My metric was not calculated wrongly. It simply did not measure goalkeeper form, which was the decisive variable that night. Data never lies, but I have misheard it before.

I carried that lesson to the billiards table. My column for time per visit measured the pace of the stroke, not the cloth. Humidity in the room, the speed of the cloth, whether the ball set was new or worn, the light on the table — those four variables shift hour by hour in a session, and not one column in my notebook recorded them. An average of 1.8 per visit on fast cloth in the evening cannot be placed beside 1.8 on slow cloth in the afternoon, even though the division is identical.

In 2026, when European football returned during the pandemic, I collected data from 81 matches played without crowds across the final nine rounds of the Bundesliga. The home win rate fell from 44.7% to 33.3%; the away side's expected goals rose from 1.15 to 1.32. I proposed cutting the home-advantage coefficient in my betting model to 0.18 goals per match, published the sample size and the significance level of p = 0.045, and warned that the sample was small. Over that stretch the model won 62% of Asian handicaps.

The value of that calculation lay in the measurement conditions attached to it: no crowd, nine rounds, one league. Empty stands do not kill football. They only strip away my layer of adjustment.

Three other examples show that a number only means something when its measurement conditions travel with it. In 2026, at the World Snooker Championship in Sheffield, Ronnie O'Sullivan completed a 147 maximum in 5 minutes 20 seconds; that record has stood for nearly three decades because the conditions were tightly defined — fifteen reds, fifteen blacks, an official match table, a stopwatch running. If someone tells you O'Sullivan once made the fastest 147, the sentence means nothing until you know fastest under what circumstances.

In 2026, Kyren Wilson won the World Snooker Championship for the first time, beating Jak Jones 18-14 in a final spread over four sessions. In 2026, Zhao Xintong beat Mark Williams 18-12 and became the first world champion from Asia. Those two facts go beyond results reporting. They reset the baseline of the whole data system behind them: when the champion group expands beyond Britain, every historical comparison table before it has to be read again.

Here is the point. An analytical table can be complete in form and hollow in substance, like a score sheet with the right result, the right names and the right date that still cannot explain why the score was what it was. In a file like that, every cell has a valid format, every field has a name, and the actual value does not exist. A reader skimming it will assume it is analysis.

In billiards this risk is more serious than in football. Football has dozens of independent data providers that can be cross-checked. Carom three-cushion does not. A table of averages per visit typed out by one person and checked by nobody becomes a reference figure after just a few citations.

An empty column is a result

Analysts are usually judged by how many conclusions they deliver, not by how many they withhold. I used to think I had failed when a data file came back with too many empty cells. I now think the opposite.

An empty column is a valid result, provided the analyst states why it is empty: because nobody measured, because two people measured differently, or because the sample is too small to separate from noise. Those three causes call for three different actions. If nobody measured, organise the measurement. If the standard was wrong, rewrite the definition. If the sample is small, wait — but wait with a deadline.

Three thousand matches taught me that one match can teach more than all of them. But that match has to be recorded properly.

Correlation is not causation; a full spreadsheet always looks like causation. A player averaging 1.9 per visit across a tournament may simply have met three opponents who kept leaving easy positions. Without recording the positions left behind, you cannot separate the good player from the lucky one.

There is one variable I still cannot measure, and I am honest about it: breathing rhythm. On a decisive visit, when one target ball is left and the whole club goes quiet, the arm changes. The seventeen rows I lost on that night in Lach Tray fell mostly on the tense visits — a pattern, not a coincidence. The person keeping the record is also a person being pulled into the match.

Signals to watch

For Vietnamese billiards, the signal worth tracking is not who wins. It is whether domestic tournament organisers publish visit-level data: the score after each visit, the number of visits per match, the longest run. With just those three fields, recorded to one published definition, two seasons would produce something nobody has had for ten years.

The file from that night in Lach Tray is still in its own folder, named by date, with seventeen valid rows and twenty-three empty ones. I have not deleted it. I do not write to persuade anyone. I write so that the data has a witness — even when the only thing it can bear witness to is its own silence.