Three Cushions, Data, and the Gap Nobody Has Counted

There was one evening when I sat in a billiards club near Lach Tray Street in Hai Phong, watching Tran Quyet Chien work through the final cushion sequence of a three-cushion frame. My phone captured exactly 34 seconds. On screen, a commentator blurted out: "He plays on instinct." I rewound that clip seven times. Seven times, I found no instinct. I found an ordered chain of decisions: choosing the second contact line, shifting the dead point on the cushion, keeping the spin inside a limit the wrist has to feel. Three seconds before releasing the cue, his eyes swept across the two remaining balls and returned to the cue ball. It was a model compressed so tightly that no number remained visible. Compressed long enough, people call it intuition.

I write this not to belittle intuition. I write it because in four years of sports-betting analysis, I have often heard myself say "instinct" when what I was actually missing was data. Data never lies, but I have misheard it. And in billiards, I mishear it more than in any other sport.

Football has Opta, has StatsBomb, has thousands of matches labelled pass by pass. Three-cushion has no equivalent. No public database counts the success rate of a thick hit versus a thin hit. Nowhere records that in frame twelve, with the score at 38-40, a player chose the safety route instead of attack, and whether that choice was correct.

What is worth noting is that billiards is not short of numbers. Every frame is a continuous stream of figures: score, innings, time per inning, average scoring rate, miss count. Professional three-cushion tournaments publish each player's average after every match. But average is a crude aggregate. It is like looking at a football team's possession rate and concluding the whole match from it. It tells you the result, not the process.

I once tried to build a manual tracking sheet for a regional three-cushion event. I logged 41 frames across eight players. For every inning, I marked three things: cue-ball position, the option chosen, and the outcome. After two weeks, I had a dataset so small it was pitiful. But it was enough to show me something no scoreboard showed.

What I saw was this. In three-cushion, top players do not win by scoring more points per inning. They win by cutting down dead innings — innings where, after the shot, the cue ball lands in a position that cannot be attacked and forces a safety. Across the 41 frames I logged, the players with the highest average were not necessarily the ones with the most long scoring runs. They were the ones with the fewest stuck innings.

That number never appears on the scoreboard. It never appears in federation statistics. It only appears when you are willing to sit down and count.

I remember a semifinal in which Nguyen Duc Anh Chien faced a Korean player. On the board, the two players' averages were almost identical. But when I rewound the footage, the difference lay elsewhere. The Korean player had beautiful three- and four-point runs. But after each one, his cue ball tended to drift into the dead zone near the long cushion. Duc Anh Chien was the reverse. His scoring shots were less flashy, but the cue ball always came to rest in the central zone, where at least two further attacking options existed. Anyone watching for fun will remember the Korean's four-point run. Anyone counting sees that Duc Anh Chien controlled the match better.

This is where I want to pause for a moment, because it concerns how we read billiards.

In football, people have learned to separate possession from control thanks to metrics like PPDA or progressive passes. In billiards, we still use average as our only measure. The problem is that average lumps everything into one number. It does not distinguish a scoring shot in an easy position from a scoring shot in a hard one. It does not distinguish a proactive safety from a forced safety. A player who is good at safety and a player who is bad at safety can end up with the same average.

I am not writing to convince anyone. I am writing so that the data has a witness.

Let me tell one more story. Last year, I followed an indoor three-day tournament in northern Vietnam. A young player I had noticed had a lower average than his opponent in all three group matches, yet still won two of the three. Looking only at the scoreboard, people would call it luck. I sat down with the footage and saw a different pattern. This player chose safety very early in the frame, usually on the third or fourth inning. That style kept his average low, because each inning scored only one or two points. But it also dragged opponents into situations where their success rate dropped sharply.

In other words, he was trading points for position. And he was winning that trade.

No metric on earth records that trade. Average records only the final result, not the price the player paid to reach it.

The strange thing is that the billiards industry has had the technology to do this for a long time. A current smartphone shoots 240 frames per second. Computer vision is already capable of recognising ball positions on the table with sub-millimetre error. The problem is not technical. The problem is that nobody pays for the labelling. In football, a data company earns millions from bookmakers and broadcasters. In three-cushion, that market is so small that nobody wants to build the infrastructure.

I know that feeling. I once spent a month building a dataset for a small tournament, and in the end I sold exactly two reports. Two. That same effort, poured into football, might have earned ten times as much. But I did it anyway, for the same reason I write this piece: if nobody counts, then nobody will ever know.

There is one more variable that football-style data cannot capture: the table surface. In three-cushion, the table is not a constant. Humidity changes with the weather, with the number of people sitting in the room, with whether the air conditioning is on. A table in Hai Phong in April differs from that same table in December. Cushion speed changes, ball bounce changes, and every spin calculation a player makes has to be recalibrated.

This is why I never use a single metric to draw a conclusion about a player. A handsome average achieved on a fast table does not mean the same as a similar average achieved on a slow one. If you merge the two, you are comparing apples to oranges. And I have made that mistake. I once praised a player based on a run of matches on fast tables, then felt let down when he moved to slow ones. The fault was not his. The fault was in my model, which forgot that a cue does not strike an abstract table.

But I have to be honest with myself here. There is a reason billiards lacks a data system like football's, and that reason is not only money.

Football needs data because it is a sport of collective probability. A team can play well and lose, play badly and win. Fans need a tool to separate luck from quality. Billiards is different. In a three-cushion frame, everything happens before your eyes. You see where the cue ball goes, you see what the player chooses. There is no tactical smoke screen like a four-man defensive line. The transparency of billiards makes the need for data feel less urgent.

And there is one more thing, harder to say: the billiards community may not actually want data. Mystery is part of this sport. When you say "he plays on instinct", you are granting the player a halo that an Excel sheet cannot grant. Data would peel that halo off, turning an artist into a decision-maker. Not everyone wants that.

I used to think data was neutral. It is not neutral. It is a choice about what we want to see in this sport.

If tomorrow someone builds a decent three-cushion tracking system, I believe the first thing it changes is not how we watch billiards, but how we name it. "Instinct" will become a hypothesis to be tested, not a compliment. Three thousand matches taught me that a single match can teach more than all of them. I am not sure I have the courage to believe that right now. But I will write it down.