When the Billiards Database Returns an Empty Cell: Notes on Silence Before a Major Tournament

WHEN THE BILLIARDS DATABASE RETURNS AN EMPTY CELL: NOTES ON SILENCE BEFORE A MAJOR TOURNAMENT

1. Two forty in the morning, and one blank column

At 2:40 a.m. on 12 September 2026, my monitor in Hai Phong returned a blank column. I had queried a 4,312-row spreadsheet built over four years, filtered for a player entering the qualifying rounds of an international three-cushion event. No win rate. No average per inning. No head-to-head. Not even a match count. The software reported no error. It simply stayed silent, and that silence looked very much like a confirmation.

I nearly typed a sentence I swore four years earlier I would never type: "There are no risk indicators." My fingers were already on the keyboard. Then I remembered the lesson that shaped my entire professional method since I was seventeen: Data never lies, but I have misheard it before.

That night I opened a new page instead of a conclusion, and wrote a heading: "What is an empty cell?" This article is my answer.

2. What billiards records, and what it forgets

Snooker is obsessive. Every professional frame is logged shot by shot, ball by ball, safety by safety. Public databases track career centuries, maximum 147s, win rates by round, and average shot times across decades. Ronnie O'Sullivan and Stephen Hendry share the record of seven world titles, and every frame of those runs is retrievable. The world champion at the Crucible Theatre in Sheffield earns 500,000 pounds; the event runs 32 players across 17 days, with best-of-19 opening matches and a best-of-35 final.

Carom three-cushion is a different world. The world governing body publishes results for World Cups and world championships, but mostly at match-score level, without inning-by-inning data. A Vietnamese three-cushion player can have hundreds of domestic matches that were never recorded in any structured form. A great deal of knowledge about these players lives only in the heads of the men sitting around the table.

Chinese 8-ball has better internal records, but they rarely cross into English-language platforms. Pool has improved through World Pool-Billiard Association and Matchroom structures, though it still lags behind snooker.

The consequence is rarely noticed: a player in a heavily documented discipline looks stable, while a player in a thinly documented discipline looks like an unknown, regardless of actual ability. Crowds read silence as weakness. My own spreadsheet, if I am careless, does exactly the same.

3. Three kinds of blank

After that night I sorted every empty cell into three categories, and that sorting changed my work more than any model I have ever written.

Blanks from random absence: the recorder was not there, the camera was off, nobody logged the score. Harmless. The correct response is to widen the uncertainty interval, never to impute an average.

Blanks by design: nobody ever intended to measure it. In three-cushion, safety success rate essentially does not exist in any database, although it decides many elite matches. In snooker, safety is recorded but not standardised between providers, so cross-source comparison is unreliable.

The third category is dangerous. Blanks by avoidance. Someone chose not to record. An organisation chose not to publish. A market chose not to list. An empty cell of this kind does not mean nothing happened. It means something happened that the data pipeline does not want to touch.

4. The anonymous qualifier

The player I queried that night had won a national title once, beaten ranked players in regional events, and never been televised. My old model would have handed him the average of his cohort. That is the classic error: assigning an average to an empty cell turns not-knowing into a number that looks certain.

I now publish intervals instead of probabilities for anyone with fewer than ten structured matches. Readers want a single figure. I give them a range and a warning. Many leave. Those who stay check my ranges against real results after each tournament.

What I learned: anonymity in the data is usually a property of the recording system, not of the talent. Over four years I logged nine players with fewer than ten recorded matches who reached the semi-finals of an international event. Nine cases prove no rule. They were enough to make me stop writing the word "certainly".

5. June 2026: when the pipeline stayed quiet

On 6 June 2026, the World Professional Billiards and Snooker Association published the outcome of an investigation that began with a suspension in August 2026. Ten Chinese players were found to have breached match-fixing and betting rules. Liang Wenbo and Li Hang received lifetime bans; others received multi-year sanctions, some reduced on appeal.

Before August 2026, their statistical profiles looked entirely ordinary. Win rates stable. Century counts unremarkable. No indicator strayed outside its normal band. The single missing column was called "off-table conduct". Nobody measured it. Nobody sold it.

Throughout that period, betting indicators on the affected matches showed normal volatility. The market did not scream. The data pipeline did not scream. The only signal came from human investigation.

In my notebook I wrote: an empty cell in a compliance column is not a clean bill of health. It is an unmeasured cell. The lesson is not to suspect everyone, which is its own intellectual laziness. The lesson is to state openly that a column never existed, rather than letting readers assume the absence of a signal means the absence of risk.

6. Cues, cloth, and Hai Phong humidity

Billiards is more exposed to physical environment than most sports. Humidity changes how the cloth accepts spin. Room temperature changes cushion rebound. Different cloth produces noticeably different ball speed. A player who practises in an air-conditioned club meets a completely different table in a large, crowded arena with doors opening all day.

For my first three years I had no column for any of this. That was a blank by design, and it made me misread many matches.

In 2026, while working with European football data, I tracked 81 matches played without spectators in the final nine rounds of the German top division of 2026/20. Home win rate fell from 44.7 per cent to 33.3 per cent. I proposed cutting the home coefficient in my betting model to 0.18 goals per match. A forum moderator attacked the sample size. I ran a chi-square test, obtained p = 0.045, published the result with limitations stated, and won 62 per cent of Asian handicap bets in that window.

One line from that period still applies to billiards: When home is no longer a fortress, I learned to listen to empty stands.

In billiards, empty stands are literal. In June 2026, snooker returned with the Championship League at Milton Keynes behind closed doors, and later events that year ran with few or no spectators. Reviewing footage, I noted that players famed for reading a room lost part of their edge, while young players who normally froze in a theatre played with unusual freedom.

My sample was 34 fully recorded matches. I stated that number in the article itself, with a warning that 34 matches cannot establish a long-term trend. A reader called this self-contradiction. I disagree. It is the only way an early observation can be useful without becoming a cheap prophecy.

For Vietnamese billiards this is concrete. A northern three-cushion player trains year-round in high humidity. In Europe, where air is dry and tables are heated, ball speed changes. No international database has a column for "training humidity". When such a player underperforms abroad, analysts blame psychology. Psychology is part of it. So is the cloth, and the cloth never enters the data.

7. One hundred and forty-seven, and the law of small numbers

A 147 is a tail event: fifteen red-black pairs, then all six colours in order, without a single miss across roughly thirty-six shots. It depends on the position after the break, on table quality, on whether the reds sit kindly.

Comparing two players by maximum breaks is therefore a sample-size error. A player contesting 400 frames in a season cannot be compared with one contesting 120. I made this mistake in a 2026 article, ranking players by career 147s and drawing conclusions about "best scoring form". A reader in Da Nang pointed out that I had ignored frames played. He was right. I published a correction.

I rebuilt the reading. Instead of counting 147s, I now count centuries per hundred frames, always with the sample attached. The same applies to carom: average per inning is the most quoted figure and the most opponent-dependent. A player averaging 1.8 against weak fields can fall to 1.1 against the top group. Without an opponent-quality column, the average tells a false story.

My rule: never present a billiards metric without three things, the sample size, the playing conditions, and the opponent quality.

8. Ankara, September 2026

On 10 September 2026 in Ankara, Bao Phuong Vinh of Vietnam won the UMB World Three-Cushion Championship, beating Kim Haeng-jik of South Korea in the final. It was the first world title for a Vietnamese player in the discipline.

What matters here is not the trophy. It is the state of my spreadsheet before the event. Bao sat in the thin-data group: no deep run at a world championship, few structured matches, most results in regional and domestic events where records are not standardised. My old model would have assigned him a mid-tier average and a very low title probability.

That model was not mathematically wrong. It was epistemically wrong. It was measuring a player's data coverage and calling the result ability.

After 12 September I had placed him in the category of severe random absence, with a note that his uncertainty interval must be wide. In my notebook I wrote: "Insufficient basis to say he will win. Insufficient basis to say he will exit early. This is a case to record as unknown." A week later he won.

I do not tell this story to claim credit. I tell it because it illustrates the point exactly: when data is empty, the right move is not to guess, but to say clearly that you do not know, and to hold the interval open. A friend in Hai Phong messaged after the final: "What does your model say?" I answered: "My model says I lack data." He laughed. I laughed. I still keep the message, because it reminds me that honesty about data is routinely mistaken for weakness of expertise.

9. What the spreadsheet cannot measure

Analysts are moving deeper into the dressing room. We measure heart rate, sleep, reaction time. We build indices for anything that accepts a sensor. And we begin to believe that what cannot be measured does not matter.

In billiards this produces a specific blindness: my spreadsheet knows how many innings a player takes per hour; it does not know how long he pauses between innings, or what he thinks during the pause.

Reviewing a three-cushion semi-final, I timed the players. The winner averaged roughly four seconds more preparation per inning in the second half. Four seconds is not a large metric. Read only as speed data, it suggests he slowed down, and slow reads as bad. He slowed because he was working through a difficult position, and the slowing was control, not hesitation.

This is the blank that data cannot fill. I can only log qualitative observation: rhythm between shots, facial expression after a missed inning, how a player circles the table before addressing the cue ball. These notes cannot be added or multiplied. They let me recalibrate quantitative models when the two sources conflict.

I once wrote about football: One goalkeeper misjudging a catch is an error. Three goalkeepers misjudging the same catch is a signal. In billiards the equivalent is: one missed inning is an error; three missed innings of the same type at three different points in a match is a signal about table conditions or psychological state. My spreadsheet counts the misses. It does not spontaneously notice that they are identical.

So: quantitative data raises the alarm, qualitative observation explains it. When they conflict, I do not rush to choose. I record the conflict in a footnote and wait for another match.

10. Signals for the next cycle

Before every major event I now run a short protocol. First, I check data coverage before data quality, recording how many fully structured matches each player has in the last twelve months. Under ten, and my judgement stays in interval form. Second, I sort blanks into the three categories, which takes about forty minutes for a 32-player field and prevents more premature conclusions than any tool I own. Third, I check environmental variables: venue, cloth, climate, travel schedule. Fourth, I write a "data required" list before writing any conclusion, so I remember I am working with an incomplete picture.

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

And if one line had to close what I learned that night, it is the line at the front of my notebook: Three thousand matches taught me that one match can teach more than all of them. That night the empty cell taught me nothing about the player. It taught me something about myself: that I had spent four years learning to read what is recorded, and not nearly enough learning to read what is left blank.

11. Technical footnote: how I build my indices

Data coverage index: the number of matches in the last twelve months with complete inning-by-inning scoring and opponent names. Thresholds: ten matches for a point judgement, thirty for a comparison between two players.

Adjusted scoring index: for snooker, centuries per hundred frames with the frame sample attached. For three-cushion, average per inning split by three tiers of opponent quality.

Environmental index: venue, estimated room temperature, relative humidity, and cloth type where available. Most events publish none of this, so the column is frequently empty, an accepted blank by design rather than a gap to be papered over.

Uncertainty index: every judgement carries a low and high bound plus a confidence rating of low, medium, or high.

None of this is advanced. It is expensive in one place only: it forces me to admit my limits in every paragraph.

One year later

I keep a habit. About a year after any contested analysis, I reopen it and check myself. If I was right, I record why. If I was wrong, I record why, and I publish the error.

For this piece, I will return in September 2026, open the spreadsheet, and count how many cells in the "data required" column have been filled, and how many remain blank. If many remain blank, that is not my failure alone. It is the unfinished map of a sport I chose to follow to the end.