The Empty Column in Billiards Data: The Trap of Reading 'Not Measured' as 'No Risk'

The Empty Column in Billiards Data: The Trap of Reading 'Not Measured' as 'No Risk'

In March 2026, I received a 4,128-row tracking sheet from a young colleague. Three columns were blank: discipline identification, cross-reference chain, compliance risk. The email attached to it contained a single sentence: "I left them empty because I couldn't find any problem."

I sat still in front of the screen for two minutes. The first feeling was not irritation. It was recognising myself ten years earlier.

In June 2026, I submitted a professional snooker tracking report with one line of note: "No compliance anomalies detected." Four years later, the World Professional Billiards and Snooker Association (WPBSA) announced disciplinary sanctions against ten Chinese players, two of whom received lifetime bans. My table was not wrong in a single cell. It had simply never contained that cell.

Data never lies, but I have misheard it before.

Four gates before trusting a data table

My job is to re-price probabilities. I do not predict who wins. I calculate what the market is paying for a scenario, then set it beside my model. If the gap is wide enough and the data dense enough, I write. If not, I stay quiet.

But there is another kind of silence, far more dangerous: silence because nothing was ever measured.

After 2026, every tracking sheet of mine has to pass four gates. First, identify the discipline before identifying the player, because the same word "break" exists in three disciplines with three entirely different meanings. Second, every metric must answer three questions: who measured it, how, and on what table and what balls. Third, every conclusion needs at least two independent sources. Fourth, every blank cell must be explicitly marked "not measured" with a reason, never left empty and unexplained.

The fourth gate took me the longest to accept, because it forces me to write sentences that sound deeply unimpressive: "I don't know."

In billiards analysis, most errors do not come from misreading a metric. They come from comparing two metrics that never shared a scale, or from drawing conclusions about a region of data that was never collected.

"Break" is not one concept. It is three.

In snooker, a "break" is a scoring run within a single visit to the table. Breaks are measured in points: 50+, 70+, 100+ (a century), and 147 (a maximum). Ronnie O'Sullivan has 15 career maximums, the most in history; his fastest, at the 2026 World Championship against Mick Price, took 5 minutes 8 seconds. There is also a separate metric called "points per visit" — the average points earned each time a player gets to the table. The two sound similar and are constantly mixed up in amateur statistics.

In American pool (9-ball, 10-ball), the "break" is the opening shot. It is measured by the rate at which an object ball is pocketed directly from the break, the rate at which a dry break leaves the opponent safe, and the rate of winning the entire rack after a successful break — a "break and run."

In Chinese 8-ball, the break happens on tables with very tight pockets, so break-and-run rates are considerably lower than in American pool on tables of comparable size.

Three disciplines. Three definitions. Three scales. If someone hands me a table with a line reading "break average: 0.62" and no discipline noted, I cannot say anything of value. Not because I lack data, but because that data has no meaning yet.

This is why the first gate exists. Before asking "how strong is this player," you must answer "which discipline are we discussing."

In three-cushion carom, the word "break" is barely used in that sense at all. People talk about "average" — the mean points scored per inning. A player averaging 1.5 at world level is already in the very strong group. Drop that same value into a snooker table and it means absolutely nothing.

In Vietnam, I follow three-cushion more than anything else. Tran Quyet Chien won the 2026 world title in Binh Duong, beating Bao Phuong Vinh in the first all-Vietnamese final in the tournament's history. What interests me is not the medal. It is that if someone downloaded that event's statistics and compared them directly with an American pool event, they would produce an entirely false conclusion — one that looks highly scientific, because it has numbers, tables, and charts.

I almost did exactly that in 2026. At seventeen, I applied a football expected-goals metric to a Vietnamese league match, confidently predicted a 3-1 outcome, and got a 0-1 result plus seven saves from the opposing goalkeeper. I thought the lesson belonged to football. It belonged to every discipline: a metric only means something inside the right frame of reference.

The compliance void: when silence is read as cleanliness

Back to the 4,128-row sheet. The compliance risk column was empty. My young colleague read an empty cell as "no problem." This is the most common logical error in the entire sports analysis industry, and also the most expensive.

In June 2026, the WPBSA announced sanctions against ten Chinese players following an investigation into match-fixing and betting. Two lifetime bans were issued. Names that had sat in snooker's most anticipated group vanished from the professional system after a single announcement.

Before that, in my tracking sheets, that region of data was completely blank. I had never built a single proxy indicator for it. I did not measure abnormal odds movement in low-tier events. I did not compare prize-money income against the real cost of living for players ranked 60 to 100. I did not record when a player suddenly went silent on social media for three straight weeks.

After 2026, I added those four proxy indicators. They are imperfect. They are only proxies, and I always state that clearly in the footnotes. But a poor proxy still beats an unexplained blank cell, because at least it forces me to admit I am speculating.

What I learned was not that "the system has holes." What I learned was this: a null result is not a clean result. It is an unknown result, and in risk analysis, "unknown" must be ranked higher than "checked and found nothing."

The case of Zhao Xintong is the example I use to re-educate myself. He returned to competition after his ban in September 2026, and by May 2026 he had won the World Championship as a qualifier, beating Mark Williams 18-12 in the final. Look only at his ranking when he was suspended and no model predicts that trajectory. But look at the cohort of players returning from bans in general — a cohort I had never tracked — and this becomes a pattern worth recording, not a supernatural event.

The career-age curve and a prophecy delayed twenty years

There is another data region I left empty for years: the career curve of older players.

Ronnie O'Sullivan, John Higgins and Mark Williams were all born in 2026 and turned professional in 2026. This group is known as the Class of '75. The prophecy that "the new generation will take over" has been repeated every year at least since the mid-2000s. The data keeps refusing it. By May 2026, Mark Williams was still reaching a World Championship final at the age of 50.

If someone had asked me in 2026 to predict that a player born in 2026 would still be contesting world finals a decade later, I would have said "low probability," and I would have been wrong.

The crowd laughed. The numbers did not. A year later, I marked my own homework. More precisely: I went back to my old forecast sheets, recorded exactly where I had gone wrong, and published the correction. It is not a comfortable exercise, but it is the only way a model stays usable.

Here is the interesting paradox: the player who actually won the world title in 2026 was born in 2026, and the 2026 champion was born in 2026. The "new generation" that took over was not the post-2026 cohort the public imagined. It was the early-1990s cohort — the group that a decade earlier had been written off as "out of waiting time."

When I redraw the career-age curve of champions over twenty years, the average age at victory does not decline. It moves sideways, with fluctuation, and no clear trend. A sideways line across twenty years is a far stronger signal than a single-season jump.

This is also why I never conclude from a single tournament. Three thousand matches taught me that one match can teach more than all of them — but only once I have those three thousand matches to place it in the right position.

Correlation is not causation, and an empty cell is not evidence

There are two traps I fall into most often, and both involve blank data regions.

The first is reading correlation as causation. Example: players who change their cue tip mid-season tend to perform better over the following three months. It sounds entirely reasonable. But look closer, and the tip-changers are largely players who just secured new sponsorship, or changed coaches, or came back from a break. Changing a tip may simply be a consequence of a larger shift, not its cause. If I write "changing tips improves performance," I have sold readers a conclusion that cannot stand.

The second trap is publication lag. In 2026, I collected data on matches played without spectators in a European national championship and found that home advantage had declined markedly. I hesitated for three weeks because the sample was small. By the time I published, most of the informational value had already been absorbed by the market. I ran a statistical test, reached an acceptable significance level, but I was late.

Since then I apply one rule: if I do not yet have enough data to publish, I still publish the part I do know, with an explicit statement of limitations. An article saying "I have 42 observations, this is what I see, and this is what I cannot yet conclude" is more useful than a perfect article published after the event is over.

In billiards, this lag is even more dangerous, because tournament cycles are short and a player's number of matches in a season is usually too small to form a large sample. Thirty matches in a season is normal. Thirty observations cannot support any claim about long-term form. They only support a question for next season.

And here is the point I want to make clearly: my biggest mistake over the past ten years was not a wrong prediction. It was three occasions when I published a tracking sheet that looked complete, structurally sound, visually polished — but inside were blank cells that were never flagged. Readers had no way of knowing. They saw a sealed table. They read it as safe.

A report that is complete in form but empty in substance is more dangerous than a report that is entirely blank. A blank report sends people looking for data. A sealed report makes them stop looking.

Signals for the next cycle

Three signals I am tracking in the coming cycle.

First, compliance proxy indicators in low-tier events — where prize money is not enough to live on and financial pressure creates grey zones. I measure odds movement 48 hours before play, not the closing price. Early movement is a signal; late movement is mostly rational money flow.

Second, the cohort of players returning from bans. I track their ranking-climb velocity over the first twelve months, because this is a group with unusual motivation and little modelling coverage. Zhao Xintong is the outlier case, and outliers are always worth recording rather than discarding.

Third, the retirement curve of the older cohort. When a generation leaves at once, tournament structure changes: the number of qualifying places, ranking standards, and how sponsors allocate budgets. This is a long-lag signal, which is precisely why it tends to be ignored until it is too late.

I do not write to persuade anyone. I write so that the data has a witness.

As for that 4,128-row sheet, I returned it to my young colleague with one requirement: fill in the three empty columns, and for every cell that cannot be filled, write down why. That table will look far worse. It will have unfinished rows, messy notes, sentences reading "unable to measure." But next time someone opens it, they will know exactly what they are reading — and what they are not.