In September 2026, in Ankara, the final of the world three-cushion carom championship was an all-Vietnamese affair. Bao Phuong Vinh faced Tran Quyet Chien. The hall was packed with Vietnamese fans, applause rolled down from the stands, and on the table were two cueists I had watched shot by shot for years.

When the match ended, my phone buzzed. My editor sent one line: “Got any numbers? Send them in thirty minutes.”

I opened my spreadsheet. It was the one I had built for football, with every column in place: passes allowed per defensive action, duel win rate, expected goals. Every cell was blank. Not blank because I was lazy. Blank because nobody measures those things in carom.

Thirty minutes later, I sent back three characters: N/A.

It was the first time in four years of working that I filed a piece with no numbers at all. It was also the first time I felt lighter after submitting.

World champions, but no data layer

Vietnamese billiards is in what people in the trade call a boiling phase. The number of clubs in Hai Phong, Hanoi, Da Nang and Ho Chi Minh City has risen steadily for seven years. Domestic three-cushion tournaments have bigger prize pools, television contracts, livestreams that reach hundreds of thousands of views. On the pool side, Duong Quoc Hoang is a name anyone following world billiards has to mention. On the carom side, Tran Quyet Chien, Nguyen Duc Anh Chien and Bao Phuong Vinh form a generation that Belgians and Dutchmen have to calculate around when the draw is made.

One thing has not grown with them: the data layer.

I followed football before I followed billiards, so I was used to everything having a number. A single English Premier League match produces more than two hundred metrics, from touches inside the box to the average distance between two centre-backs. Billiards has no equivalent ecosystem. The world carom federation publishes results, publishes the average per inning, publishes the highest run. That is roughly the whole of it. No defensive metric. No break-shot quality metric. No rhythm metric.

There is a second complication, and it is the one most articles fall into: vocabulary across disciplines looks the same but means different things. Snooker has the century break, one hundred points in a single visit. Three-cushion carom uses the “serie”, and a run of twenty already brings a hall to its feet. American pool uses the rack. Chinese 8-ball mixes a snooker table with pool rules. Russian pyramid has its own table, balls and rules. An analyst who writes about a carom player's “century rate” is committing a category error, not a wording error.

So the first step in every analysis I write is identifying the discipline. If the discipline cannot be identified, every comparison is methodologically invalid. That sounds obvious, but the number of Vietnamese sports articles that break this rule is not small.

Four questions before using any number

Every piece I write starts with a list I call required verification conditions. Four questions, always in the same order.

First: who measured it. A carom result published by a federation is not the same as a result written down by a spectator in row ten. Both can be correct within their own scope, but they do not carry the same level of reliability.

Second: how it was measured. Does the organiser record points from the electronic board, or does someone use a stopwatch on each inning? Does anyone review the video to count how often a player left the opponent a scoring position?

Third: under what conditions. Three-cushion carom depends enormously on the cloth. A freshly covered table, fast cloth, new balls, produces a different scoring average from a table three days into a tournament with slack cloth and worn balls. The same cueist, the same opponent, the same day, two different tables can produce two different matches.

Fourth, and my favourite: what this metric is hiding. Scoring average per inning is the perfect example. It says how many points a cueist scored per visit. It does not say whether those points came from aggression or from patient position play. Two cueists averaging 1.8 can have opposite risk profiles. One grinds the opponent into difficult positions and takes small scores. The other gambles for long runs. Same cell, two different stories, and looking only at that cell cannot tell them apart.

In football, the fourth question once cost me a year to answer.

The lesson from a goalkeeper who kept missing

In 2026 I was seventeen, applying expected goals to Vietnamese football for the first time. I pulled the numbers for Hai Phong against Sanna Khanh Hoa in round 18 of the V.League. Hai Phong generated 2.8 expected goals. The opponent generated 1.0. I confidently predicted a 3-1 Hai Phong win and posted it on a forum with the tone of a man who had found the holy grail.

The match ended 0-1.

Goalkeeper Tran Buu Ngoc made seven saves, three of which I still believe were unsaveable when I watch them back. My whole model collapsed inside ninety minutes, and I learned something that became the foundation of everything I write: expected goals does not account for goalkeeper form, and in matches with a low defensive block, goalkeeper form is the decisive variable.

I did not patch the model. I did something slower. I sat down and hand-logged twenty consecutive matches, one page each, cross-checking expected goals against actual results, and writing down everything the metric left out: goalkeeper form, set pieces, cards, weather.

Those twenty matches taught me more than the two thousand automated matches sitting on my hard drive.

One goalkeeper missing a hand is a mistake. Three goalkeepers missing a hand is a signal. The problem is that to see three goalkeepers missing a hand, you have to accept counting every single one yourself instead of trusting a pre-aggregated cell.

The summer without crowds and the small-sample trap

In 2026, mid-pandemic, I was twenty. The Bundesliga returned and played its final nine rounds of the 2026/20 season as eighty-one matches without spectators. I collected all of them.

Home win rate fell from 44.7 percent to 33.3 percent. Average away expected goals rose from 1.15 to 1.32. From that I proposed cutting the home coefficient in my betting model to 0.18 goals per match.

A forum administrator attacked me for a small sample. He was right about the sample. He was wrong about the conclusion, because he assumed a small sample means you are not allowed to conclude anything.

I ran a chi-square test, got a p-value of 0.045, and published the results with an explicit limitations note: eighty-one matches, nine rounds, one league, one country, one specific pandemic period. That model won me 62 percent of Asian handicap bets in that window. But 62 percent is not the number I want to recall. What I want to recall is that limitations note sitting directly beneath the results table.

Empty stands did not kill football. They only stripped away my layer of adjustment.

What I learned that summer was not that home advantage had lost value. It was a different question: had the home coefficient I used for years actually measured home advantage, or had it measured the presence of a crowd? Until those two can be separated, any home coefficient remains a black box with a very confident label.

When three-cushion carom has no PPDA

Back to Ankara. After sending that N/A, it took me nearly a year to build a manual logging system for three-cushion carom.

Four variables. First, the highest run of the match. Second, total innings. Third, how often a player left the opponent a scoring visit of three points or more. Fourth, average time per inning, measured with a stopwatch on video.

I logged forty-six matches. That is a number small enough that I do not dare publish a single coefficient from it. Forty-six matches is enough to see shapes, not enough to assert anything.

Shape one: matches in which a cueist allowed the opponent a three-point-plus visit four times or more usually finished one way. Shape two: average time per inning rose noticeably in the middle phase of matches. Shape three: highest run and match result showed almost no correlation.

Those three shapes live in my notebook, not in any article. I hold them back because I know exactly what happens if I publish early: someone will cite them as a law, and three months later, with more data, I will have to write a correction, and readers' trust in data will wear down another layer.

This is where I have to be clear about a very common mistake in Vietnamese sports analysis: importing metrics from another discipline.

PPDA in football measures how many passes an opponent is allowed before the ball is recovered. It is a beautiful metric, and I used it in my 2026 piece about Mexico beating Germany 2-1 at the World Cup. Germany had 66 percent possession and made 613 passes, but Mexico's PPDA was 8.4. I concluded Germany would exit early. The blog was mocked. Two weeks later Germany lost 0-2 to South Korea and went out. I received twelve emails from readers admitting I had been right.

But that story did not teach me that PPDA is a universal truth. It taught me that PPDA only means something in a sport that has passes. Three-cushion carom has no passes. Nobody passes to anybody. Putting PPDA into carom is using a thermometer to weigh something.

The same happens with snooker's century metric. A century is a meaningful threshold in snooker because the table has fifteen reds and a maximum of one hundred and forty-seven points. In three-cushion carom there is no ceiling on a single visit. A cueist can keep scoring until they miss. So talking about a carom player's “century rate” is using the wrong word, and using the wrong word leads to the wrong analysis.

So what would a real data layer for three-cushion carom need? I think three things, and none of them currently exists in standardised form.

First, defensive efficiency: what position you leave the opponent in when you leave the table. Can it be measured? It can, if someone reviews video and scores it on a scale. But there has to be one scale for an entire tournament, and there has to be a reliable scorer.

Second, break-shot quality: how many points you score after the break shot. This matters because in carom a good break opens a run, and runs open matches.

Third, rhythm: time per inning, and time between innings. Rhythm is something the eye feels but almost nobody records. I once watched a match where a cueist spent nearly twenty seconds per inning in the opening phase and dropped to twelve seconds in the closing phase. Looking only at the scoring average, you see nothing. Looking at rhythm, you see a man losing control of his own clock.

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

Blank-cell propagation: the difference between zero and unknown

There is a systemic error I have witnessed many times, and it is more dangerous than a wrong number.

I call it blank-cell propagation.

A report with an empty cell is passed from one person to another. The recipient does not read closely and assumes the empty cell means zero. A third person reads the summary, sees a tidy string of numbers, and concludes an analysis exists. By the fourth person, the empty cell has become a claim.

In statistics, zero and unknown are entirely different things. An average of zero means the cueist scored nothing. An undefined average means nobody measured. Merging the two is the fastest way to produce a headline that has to be corrected three months later.

In Vietnam this error appears constantly in sentences like “how many world champions has Vietnam had”. Nobody supplies a denominator. Such a question can be answered with five, eight or ten depending on which tournaments you count, whether you include team events, whether you include invitational events outside the ranking system. Without a denominator, the number is just a manner of speaking.

Data never lies, but I have misheard it before.

My everyday handling of blank-cell propagation is simple. Every blank cell in my spreadsheet carries its own marker and a note: not measured, measured with the wrong method, or measured but unpublished. When I file a piece, I keep those markers in. An editor may delete them, but I will not delete them myself.

This is why I regularly write “insufficient data, cannot assess” in places where readers want a decisive answer. A piece with three such moments looks weaker than a piece with three strong claims. But the first piece is still true next month, and the second one is not.

The other side of transparency

Here I have to argue against myself, because otherwise this piece becomes a hymn to caution, and caution does not deserve hymns.

Transparency about limits has another side: it becomes a hiding place.

I have read three-thousand-word analyses in which two thousand words were a list of limitations and the conclusion said nothing at all. The author was not wrong methodologically. But they also did not do their job. Facing a match, a reader needs a judgement, not a catalogue of unknowns.

So I set myself a rule: the collection deadline must be fixed before collection begins. If the deadline arrives and the data is still insufficient, I must still make a call, timestamp it, and archive it so I can grade myself later.

That rule came from a time I spent four months waiting for more data on a cueist, and by the time it arrived the tournament was over. I was methodologically right and professionally useless.

Two years after Ankara, I went back and reread my piece on Mexico and Germany, reread the limitations note in the empty-stadium piece, reread even the lines I got wrong. The crowd laughed. The numbers did not. A year later, I rewrote that piece.

Rewriting matters more than being right. An analyst with no archive to check themselves against is not doing analysis; they are selling predictions.

Signals for the next cycle

Three things I will be watching in the coming tournament cycle.

One: any Vietnamese broadcaster that starts publishing per-inning data, even raw data. The emergence of an original, verifiable data source matters more than any elegant analysis built on sand.

Two: any change in how the world carom federation publishes data, especially if it adds a defensive metric. The day that metric appears, everything I hand-logged across forty-six matches becomes a cross-check.

Three: the young generation of cueists. Not to find the next champion. To see whether any of them will sit down and take notes.

Three thousand matches taught me that one match can teach more than all of them. But only if someone bothers to write it down.

---

Method note: this article draws on the author's personal logs from 2026-2026, official tournament results, and public data from the 2026/20 Bundesliga season (81 matches without spectators). Sample sizes and analytical limits are stated at each section. The content is for sports information reference only and does not constitute betting advice. Sports results are highly uncertain.