When the Dataset Returns Zero: Nine Empty Columns of Vietnamese Billiards

Two in the morning at a billiards club on Lạch Tray Street, Hải Phòng. Three windows open on the screen: a spreadsheet holding 4,218 hand-logged rows from four seasons of three-cushion matches, a tab replaying a World Cup final, and a short script I wrote to extract data from my own notes. I hit run. The script returned an empty file.

Not a syntax error. Not a corrupted drive. My filter was too strict — it kept only matches with both turn-timing data and cue-ball path data — and the spreadsheet returned exactly what it held: zero.

I stared at that file for ten minutes. Four years in this trade, I am used to data saying less than I want. I was not used to it saying nothing at all.

This piece starts there. It is not a report on any tournament. It is a record of the night my analytical toolkit returned a skeleton made entirely of blank cells — and of what Vietnamese billiards has taught me about reading those blanks.

A loud sport with a thin data field

Vietnam's billiards record is thick enough to make outsiders look up. In carom three-cushion, the Union Mondiale de Billard (UMB) runs an annual World Cup circuit, and the stop held in Bình Thuận is one of its fixed Asian legs, drawing the world's leading cueists to Phan Thiết. In pool, Matchroom brought the Hanoi Open Pool Championship to Hanoi, placing Vietnam on the professional ranking map. Domestically, names such as Trần Quyết Chiến, Ngô Đình Nại and Nguyễn Đức Anh Chiến in carom, and Dương Quốc Hoàng in pool, have become technical benchmarks for a generation of students.

The data infrastructure trails the results by a wide margin. Kozoom supplies live data for international carom legs, but it is largely score and turn sequence. You know how many points a player scored in a turn. You do not know how many centimetres separated the cue ball from the object ball, nor which pattern of positions he failed on. Domestic events, from club level to national qualifiers, mostly leave behind a final score and a few lines of commentary, occasionally a photograph of the scoreboard.

Dense competition, thin data field.

I once built a model for carom matches by merging UMB data with my own handwritten logs. For three seasons straight, the gap between prediction and outcome ran wider than I could accept. The reason sat in plain view: I was predicting a discipline in which I could only measure outcomes, while the causes lay out of reach. Data never lies, but I have misheard it.

Discipline identity and playing style

If the data existed, the first column I would fill is turn structure. For carom, that means average per turn, mean length of a scoring run, turns needed to close a match, and the miss rate of position patterns by table region. For pool, it means pot success by distance, break quality in terms of table layout left behind, and the rate at which a player forces his opponent to sit after a safety exchange.

What I actually hold, after four seasons, is half of that picture. I log successful turns with care: where the player stood, how he solved the position, how many points he took. I barely log failed turns. This is a survivorship bias baked into note-taking, and it poisons every model built on top, because the model learns that any layout can be solved with enough touch. You measure only what succeeded, and every model built on half the data carries manufactured optimism.

The fix is cheap: make logging failed turns mandatory. But to do it, the person logging must accept that a broken turn deserves the same attention as a brilliant one. In a billiards culture where errors are still treated as things to hide, that is a cultural requirement, not a technical one.

Players and competitive form

For each player, the minimum viable table needs: matches per season, points per turn, wins against opponents inside the top twenty, head-to-head records, and month-by-month form distribution.

Sample size comes first. A Vietnamese carom player competing internationally might have only fifteen to twenty fully logged matches in a year. Split that across opponent tiers and you are left with three or four matches per group. No model is trustworthy at that threshold. To say anything meaningful about form, you must pool data across borders and across seasons, and accept that the most recent season does not weigh as much more than a season three years back as it feels.

This collides with a professional habit of mine. After every round, messages arrive asking who will win. The more useful question is where a player's data series is leaning, and how long it has been leaning that way. One match proves nothing. Three thousand matches taught me that a single match can teach more than all of them, but only once you have three thousand as a floor.

One match stays in my notebook because it broke a prejudice. A young player entered an event with a clearly lower points-per-turn average than his opponents, then won four straight qualifiers. When I pulled the logs apart, the only difference sat in the opening turn of each match: he chose safety over attack, pushing opponents into taking the first risk. No aggregate metric told me that. Only turn-by-turn logging did.

Tournament format and structure

International carom legs combine group qualification with knockout rounds, a few dozen entries, and prize money distributed by finishing stage. Domestically, most events publish only the total purse, with detailed breakdowns appearing occasionally. This is an underrated indicator: an event that funnels money to the champion produces different competitive behaviour from one that pays evenly down to the semi-finals.

What is missing: maximum turns per match, break lengths, table conditions per round, and most importantly the split between qualifying places and invited places. That is the data that decides who truly has a path, not the ranking printed on paper.

The power map

In carom, the world divides rather cleanly. The title-contending group holds players whose average per turn stays high across consecutive seasons. The middle tier can beat anyone in a single session but cannot hold form across three. The rest is the pipeline, needing two or three more seasons of accumulation.

Vietnam has representation at the top, a handful in the middle tier, and a visible gap in the pipeline. That gap is not about talent. It is about competitive rhythm. A young cueist who wants to reach the top group must play enough, and against hard enough opposition. Domestically, few events carry high-quality fields. Internationally, cost is a barrier. The result is a silent filter, and that filter runs on exactly the data we do not have.

Rules, governance and compliance

This is the darkest column. Not because something bad happened, but because nobody publishes enough for an outsider to verify. For a discipline tied to betting at a low but real level, the necessary items include: a process for reporting unusual matches, monitoring of betting activity at major events, criteria for wildcard entries, and contract terms between players and sponsors or clubs.

Without those facts, an analyst has two choices: stay silent, or assert what cannot be verified. Serious professionals stay silent, which is why the blank in this column is wider than any other. I do not write to persuade anyone. I write so that the data has a witness.

Career ecosystem and psychology

How does a professional cueist in Vietnam live? Income typically comes from four sources: prize money, sponsorship of cues and accessories, unofficial competitive play, and income from coaching or streaming. Those four sources have very different stability, and their mix shifts with career age.

On the psychological side, the metric worth tracking is win rate on decisive turns — the final turn of a tight match, or the first turn after a break. Hand logging can capture this, if the logger sits long enough. It sits closer to the human part of the discipline than any average. A player can hold a 1.4 average all season yet lose six of ten decisive turns. The ranking table will not tell you. The notebook will.

Risk and the missing-data loop

Competitive risk: dependence on a small group at the top while the pipeline thins. Income risk: most players lack the cash-flow stability to sustain a professional training schedule. Systemic risk: decisions about entries and scheduling rest with very few people, with no data to test their reasonableness.

These three are not independent. They stack into a familiar loop: missing data produces imprecise decisions, imprecise decisions misallocate opportunity, misallocation thins results, and thin results make fewer people willing to publish numbers. The loop feeds itself, and it needs no one to actively maintain it.

Public narrative and expectations

Every time a Vietnamese player goes deep in an international event, domestic opinion builds a story larger than the event. That is understandable and partly necessary, because it brings sponsorship and new students. But there is a gap between expectation and structure: expectation is built on one match, structure on many seasons. The gap only narrows when long-horizon data exists for comparison.

I have been on the wrong side of that gap. In 2026, after a favoured team lost, I wrote that the metrics had signalled the outcome in advance. The crowd laughed. The numbers did not. A year later, I filed that piece again. The lesson was not that I was right. The lesson was that crowds and data can be wrong together, differing only in when the error surfaces.

How the value chain transmits

Data flows downstream along the value chain. Clubs and practice halls are the first touchpoint, where information about recreational players forms: age, frequency, table type, style. Equipment is the second link, with cue, tip and table sales acting as an early indicator of market health. Broadcast and sponsorship is the third, where audience size sets contract value. Coaching and the talent pipeline is the fourth. The derivative market — digital content, analysis, sports data — sits at the end and only survives when the earlier links supply raw material.

In Vietnam, the first two links are lively. The third is growing fast on the back of international events hosted at home. The last two are close to empty, not for lack of interest, but because no data stream reaches them.

A blank is a data point, not a defect

The default reaction to missing data is to go and find more. I did that for years, and half of that time was wasted. Because blanks are usually not a by-product of underdevelopment. They are a by-product of incentives.

Ask: who benefits if a player's failed turns are fully logged? The honest answer is almost nobody, at least in the short term. Players do not want weaknesses stored as files. Organisers do not want entry allocation scrutinised. Sponsors do not want audience numbers independently audited. When every party has a reason not to publish, the blank becomes an equilibrium, and every call for transparency runs into that wall.

Put differently: the emptiness of a dataset is an indicator of the sport's incentive structure, not of its technology level.

I learned this the uncomfortable way. In 2026, I used an xG-based model to predict a V.League match and got it badly wrong, because the model had no goalkeeper variable. In 2026, when European leagues played without crowds, I rebuilt the home-advantage coefficient on a small sample and drew criticism. The significance test came back at p equals 0.045 — enough to trust, not enough to be smug about. The model knew in October. I only had the courage to believe it in May.

What I carried away was not a better model but a habit: whenever the dataset goes blank, stop and ask who benefits from the blank. When the home ground stops being a fortress, you learn to listen to an empty stand. Now I am learning to listen to empty cells as well. They talk too, in a language the analytics trade rarely wants to translate.

What data still needs to be added

The minimum for the next cycle is four sources. Complete turn-level data, including failed turns, for at least sixty national-level matches per season. Detailed entry allocation and prize breakdowns for domestic events. Month-by-month form data for players inside the world's top fifty, pooled across at least three seasons. And a published process for reporting unusual matches, with a named accountable party. All four are affordable. What is missing is the decision to do them.

Looking to the next round

The empty file from that night still sits in my working folder. I have not deleted it. Every time I open it, it reminds me that however powerful the toolkit, it can only draw the boundary of what has been recorded. Everything beyond that boundary is where this sport actually happens, and also where hasty conclusions are generated most.

Next season, when a Vietnamese player again goes deep at a World Cup leg, someone will ask me for a prediction. I will answer with a list of conditions to verify first. If that list is longer than in previous years, it is probably growing for the right reason.

And if someone asks why I keep writing about an empty dataset, the answer is this: once you know what you are missing, you stop guessing. That is the entire value of four years of note-taking, packed into a file that contains nothing.