Billiards Does Not Speak One Data Language: Four Tables and the Trap of the Number Reader

I remember that March night last year, sitting alone after work, reopening the recording of a three-cushion final to cross-check the numbers. The player I was tracking had an average of 1.842 — meaning he scored 1.842 points per inning on average. Good range for the international stage, but not exceptional. I wrote that number down, and right beside it, I added a line from an old snooker analysis: that player had three century breaks across four frames. And I stopped, pencil hovering over the page, because I realized I was about to do something meaningless. Those two numbers did not sit on the same coordinate system. They shared no unit, no rhythm of play, no definition of victory. I had nearly priced a three-cushion player with a snooker ruler.

That was the moment I call "misreading the language." And in my line of work, it is the most expensive kind of mistake, because it is not caused by carelessness. It is caused by confidence. You believe you understand billiards, so you believe every billiards number speaks the same tongue. But billiards is not one sport. Beneath that single name lie at least four competitive systems so different they are almost four separate sports, sharing a cue and a cloth-covered table. Each system speaks a different data language. And if you are an analyst, a bettor, or simply a fan who wants to understand what is happening on the table, the first thing you must do is not ask "who is better," but ask "which kind of billiards are we talking about." Data never lies, but I have misheard it — and I misheard it not because the number was wrong, but because I forgot I needed to identify the language before translating.

Context — A sport with four heads, and an industry only now realizing it

Before going into detail, I need to redraw the map. Billiards, in the popular Vietnamese understanding, is usually collapsed into one block. People say "going to play billiards" without distinguishing. But step into a real practice hall, and you will see the separation is sharper than you think. There are pocket tables and carom tables. There is traditional green baize and there are pockets cut to very different tolerances. There are thin, hard-tipped cues and thicker cues meant for carom. Each difference drags along its own rulebook, its own technique, and of course, its own dataset.

I classify it into four major branches. The first is American pocket billiards — 8-ball and 9-ball (with 10-ball as cousin) — where the goal is to pocket balls in a prescribed order. The second is snooker, played on a large table with small pockets, with the most complex scoring system and the longest competitive tradition. The third is carom — pocketless billiards — whose peak is three-cushion, where the cue ball must contact at least three cushions before touching the other two balls. The fourth is Russian pyramid and variants like Chinese 8-ball, where pockets are cut extremely tight and success rates are low enough to become a data signature of their own.

What is notable is that about a decade ago, almost no one in Vietnam talked seriously about billiards data analysis. Football has xG, basketball has PER and true shooting, tennis has serve statistics. Billiards had... feeling. People judged players by "form," by "cue feel," by nights when the cue moved as if it had eyes. And I do not deny those things — they are real. But the problem is that feeling cannot be verified, cannot be compared over time, and cannot be used to price a bet.

The change came slowly but steadily. When international carom tournaments began publishing detailed statistics after each match — average, high run, innings count, scoring rate on decisive innings — the door opened. The problem is that when the door opened, many rushed in without noticing that each room has its own house rules. They carried snooker reading onto pool, pool reading onto carom, and then wondered why their models kept failing.

I was once in that group, so I say this not to judge.

I began my analytical career by applying tools I had learned from football to billiards. I used the concept of "control" — in football, the team with more possession is usually considered the proactive side. In a 9-ball match, I once assumed that the player pocketing first more often was the one in control. Wrong. In 9-ball, "pocketing first" can be a trap: you break the 1-ball but leave a bad layout, and the opponent needs only one clean run to win. My numbers then counted how often someone potted, not the quality of the leave. I measured the wrong thing. That was the first lesson: in a sport where an inning can end the entire match, the number of innings does not indicate control. The quality of the inning does.

Core — The four data languages and how they fool you

To make clear why identifying the "kind of billiards" is a prerequisite, I will walk through each data language in turn. For each, I will point out the core metric, how it is measured, and the trap that appears when you read it with another sport's ruler.

The snooker language: telling stories through sustained scoring

Snooker has the most mature data system, simply because it has the longest television history and the largest body of public data. When you analyze snooker, you do not count shots. You count breaks — continuous scoring within one visit. The basic unit of analysis is the "frame," and within a frame, a player may score 50+ or 100+. A century break is the 100-point mark in one visit. The number 147 is the maximum possible in a clean frame.

In this language, "control" does not exist. What exists is "chance conversion." A good snooker player is one who, when given an opportunity, turns it into a long break and keeps the opponent in the chair. The safety metric — defensive shot quality — is the flip side of the coin: if you cannot clear the table, you must leave the opponent in trouble severe enough that they are forced to foul.

The trap here is this: someone used to reading snooker will look at pool and search for "sustained scoring." In 9-ball, long continuous scoring does not exist in an equivalent sense. You may clear the entire table in one visit — a break and run — but that is a binary achievement: either you clear it, or you do not. There is no such thing as "clearing 70% of the balls and leaving a decent layout." In 9-ball, any inning that does not end in victory is close to a failure if it leaves a good layout for the opponent.

The pool language (8-ball, 9-ball, 10-ball): telling stories through run-out probability

This is where billiards data becomes the most brutal, and also where the analyst is most prone to overconfidence. In pool, the unit of analysis is the rack, and the life-or-death metric is the break-and-run percentage, along with the pot success rate on the break.

The break shot in 9-ball is one of the most studied shots in the sport. It decides almost the entire position. A good break does not merely pocket a ball — it controls the cue ball position, the target ball positions, and most importantly, it does not leave the opponent an "easy" layout. The metric of "percentage of breaks that leave a bad layout for the opponent" is one of the most valuable hidden metrics I have ever used, and it is almost never published in amateur events.

The trap of the pool language is false symmetry. People think that because 8-ball and 9-ball are both pool, they speak the same tongue. Not true. In 8-ball, you are assigned a group of balls (solids or stripes), and the match has a clear phase structure. In 9-ball, balls must be potted in numerical order from 1 to 9, and the 9-ball is the only meaningful target. These two structures generate two different sets of tactical data. A great 8-ball player may play 9-ball very differently. And in Chinese 8-ball, where pockets are cut ruthlessly tight, the success rate on each shot drops sharply, making safety play far more important than in American pool. Both are called "8-ball," but the data in each place tells a different story.

The carom language (three-cushion): telling stories through average and high run

This is the language closest to Vietnamese fans, because Vietnam is one of the top carom powers in the world. In three-cushion, you do not count breaks or racks. You count points and innings. The core metric is average — points per inning. Alongside it are high run — the longest scoring streak in one inning — and best average, the rate achieved in a specific match.

The data structure of three-cushion has a feature the other two lack: it is dominated by a psychological effect I call "inning lag." At high levels, a single inning can last many minutes and score many consecutive points. That means a player can sit out for a long time, and when they return to the table, they must immediately rediscover their feel. The average metric therefore depends heavily on breaking the data down by match phase. An average of 1.8 across a whole match may conceal that the player scored only 0.9 in the first half and 2.7 in the second — or conversely, a blazing start that collapsed as pressure mounted.

The trap of the carom language is the illusion of linearity. Beginners reading numbers tend to think "higher average means better." But in three-cushion, how innings are distributed matters no less than the mean. A player with an average of 1.5 who maintains steady rhythm may beat a player with an average of 1.8 who depends on a few explosive runs. Because in three-cushion, your opponent is not only the person across from you, but also "the innings remaining" — if you lose your rhythm, you lose control of time.

I learned this through a specific mistake. I once built a small model predicting carom results based on the historical averages of two players. It predicted correctly about 58% of matches. Not bad for a naive model, but I knew I was missing something. When I added the variable "inning variance" — the volatility of the scoring streak — the accuracy rose to nearly 66%. The lesson: in three-cushion, consistency matters more than peak. Peak wins you a match. Consistency wins you a tournament.

The Russian pyramid and Chinese billiards language: telling stories through failure rates

This is the least discussed language, but it contains the most overlooked signals. In Russian pyramid, pockets are extremely small and balls are very heavy, making pot success rates far lower than in other sports. In Chinese 8-ball, pockets are also cut tight. When the success rate per shot drops, the entire weighting of the match shifts from "attack" to "no mistakes." The data language of these sports therefore revolves around negative metrics: unforced errors, innings lost by leaving too good a layout, and the rate of safety shots that force opponents into fouls.

The biggest trap here is when an analyst accustomed to an "attacking" environment carries over into a "no-mistake" environment. They judge a player by spectacular finishes, while the match is actually decided by who misses less. In these sports, the winner is usually not the best player, but the one who commits the fewest errors. If you price by spectacular shots, you will misprice.

The core contrast: four questions, four rulers

To sum up this core section, I want you to remember one thing: each sport answers a different question, and therefore each sport needs a different ruler.

Snooker asks: "How well do you convert opportunities into long breaks?" The ruler is centuries and break-building.

Pool asks: "How well do you finish racks and control the break?" The ruler is break-and-run and leave quality.

Three-cushion carom asks: "How steadily do you score across innings?" The ruler is average and inning variance.

Russian pyramid and Chinese 8-ball ask: "How well do you control risk and limit errors?" The ruler is error rate and defensive quality.

Four questions, four rulers, four languages. When you mix them, your model is not wrong in the number. It is wrong in the grammar.

Contrarian — The trap of the number standing alone

In this section I want to go against a popular belief in analytical circles: that more data is better. In billiards, that is not quite right. The problem is not the quantity of data, but which language that data belongs to, and whether we are inadvertently fusing two languages.

I once witnessed a forum debate about which player was "greater" — a three-cushion world champion or a snooker world champion. The debate ran for hundreds of comments, and each side offered impressive numbers: one side world titles, the other century breaks. But both sides made the same mistake: they were comparing two things that cannot be compared. It is like comparing a marathon runner with a swimmer based on medal count. The numbers are real, but the question is wrong.

In the work of a betting analyst, this mistake can cost real money. The billiards betting market is increasingly popular, especially for carom, and I noticed an interesting phenomenon: odds usually reflect the bettors' understanding of the specific billiards type, not of "billiards in general." Those who understand three-cushion read three-cushion odds far better than those who understand snooker yet try to read three-cushion odds. This sounds obvious, but it says something profound: knowledge of one billiards variant does not transfer automatically to another. Expertise is specific, not general.

And here is the most counterintuitive point I want to stress: the best metrics for analyzing one kind of billiards are sometimes the worst metrics for predicting it. For example, in snooker, the number of century breaks is an excellent measure of a player's past class. But in the short term, it predicts poorly, because it depends on how many chances the opponent leaves — a variable outside the player's control. A player may play better in one match yet have fewer centuries than in the previous, simply because the opponent defended better and gave no chances. If you use century breaks to predict the next match, you are measuring chances, not class.

The majority laughed when an analyst said attacking metrics cannot predict results. The numbers did not laugh. Years later, when I went back to cross-check, I found that models focused on "chance quality" rather than "quantity of achievements" produced far more stable results. I had to rewrite that lesson.

Another counterintuitive angle concerns the human factor — which purely quantitative models tend to neglect. In billiards, psychological rhythm has a far greater influence than raw data suggests. A three-cushion player trailing behind may play worse or better depending on whether they feel the "rhythm" of the match. There are matches where the data says one thing but the player's body says another, and the body is often more right in the short term. I once ignored this signal because it could not be measured by numbers, and I paid with a string of wrong predictions.

I am not saying feeling matters more than data. I am saying data does not cover everything, and a good analyst is one who knows where data stops and where qualitative observation is needed. The rhythm of entering the table, the expression after a miss, the pauses between innings — those are not in the stat sheet, but they are on the table.

And there is one more trap, the one I call "the number standing alone." A number without context is a meaningless number. An average of 1.8 in a low-competition tournament is not the same value as an average of 1.8 in a world final against top players. A 60% break rate on a wide-pocket table is not the same value as 60% on a tight-pocket table. Every number needs an asterisk, and the analyst's job is to write that asterisk, not erase it to make the table look pretty.

I always remind myself: one goalkeeper missing a handshake is an error, but three goalkeepers missing handshakes is a signal. In billiards, one miss is an accident, but a pattern of misses repeated across many matches is data. The problem is that you need enough sample to see the denominator. One match says nothing. Three thousand matches taught me that one match can teach more than all of them — but only when placed beside thousands of others.

Takeaway — A thought moving forward: relearning the grammar of each sport

What I want to leave behind is not a formula, but a habit. Before reading any billiards number, ask three questions. First: which kind of billiards is this. Second: how is this metric measured, and under what conditions. Third: what is this metric hiding.

Those three questions are not complicated, but they completely change how you view a stat sheet. They turn you from a number reader into someone who understands the grammar. And in a sport with four heads like billiards, understanding grammar matters more than knowing vocabulary.

When the home ground is no longer a fortress, I learn to listen to the empty stands. With billiards, when a number no longer stands alone, I learn to listen to the context. A three-cushion player with an average of 1.8 is not a number. He is a story about rhythm, about pressure, about long innings and silences. The stat sheet is only the table of contents of that story.

The future of billiards analysis, in my view, is not in creating more metrics. It is in classifying metrics by sport, and in admitting that some things cannot be measured by numbers. The best models I have ever used were not the ones with the most variables. They were the ones that knew how to stay silent when the data was not enough.

Perhaps the most telling thing about this work is this: the more I analyze, the less certain I am about big claims. I know more about how numbers are produced, and therefore I doubt them more, not less. That sounds like a paradox, but it is the truth of anyone who has worked with data for years. Beginners trust the number. Veterans trust how the number is made.

And there is one thing I still ask myself, every time I reopen a match recording: if each kind of billiards speaks a different tongue, is anyone listening closely enough to translate correctly? I do not write to persuade anyone. I write so that data has a witness. And the best witness is the one who, before recording testimony, has asked what language the witness is speaking.