The Home Table Is No Longer a Fortress: Re-reading Home Advantage in Vietnamese Carom Billiards

Last April, in a carom 3-cushion semi-final held in Binh Duong, the home player led 38-32 going into the 24th inning and needed just two points to close it out. He missed a cushion-first escape, his opponent ran seven, and the match ended 40-39 for the visitor. I stayed in the arena for nearly two hours afterwards, reopened my notebook, and found a number that forced me to rewrite an assumption I had carried for years.

Across 41 semi-finals and finals I have watched in person at domestic tournaments since the start of 2026, players who qualify as home — those who train regularly in that very arena, not those whose household registration lists the city — won only 21. A 51.2 percent rate. That is close to a coin toss. For an advantage that anyone in the stands assumes is self-evident, the result deserves a pause.

Data never lies, but I have misheard it before. This time I want to retrace how I misheard it, from the beginning.

Context: an ecosystem measured in cloth and humidity

Vietnamese billiards has two large currents, and they run on almost separate logic.

The first is carom — mostly three-cushion and one-cushion. This is where Vietnam holds genuine international standing. Players such as Tran Quyet Chien, Bao Phuong Vinh and Nguyen Duc Anh Chien have lifted Vietnamese carom from regional invitee to a group contending for world titles over roughly the past six years. The domestic circuit has thickened quickly: open tournaments in Binh Duong, Hanoi, Vung Tau and Da Nang, plus qualifying rounds feeding the World Cup stops.

The second is pool — 9-ball, 10-ball, 8-ball. It is more widespread at grassroots level, tied to the network of halls and amateur events, but it has also produced figures who have reached continental competition, Duong Quoc Hoang among them.

What the two currents share matters more: most Vietnamese indoor arenas are not purpose-built billiards venues. They are multi-purpose training halls, rebuilt each time a tournament is staged, with tables, cloth, cushions and air conditioning brought in on a schedule. This is the single most important detail in this whole story. At major international events, the playing table is fixed and highly standardised. At most domestic events, tables are assembled from whatever the organisers can source.

Which means: when we say home advantage in Vietnamese billiards, we are talking about two different things and usually merging them into one. The first is crowd advantage — applause, breathing, familiar faces, the feeling of playing for a home audience. The second is table advantage — familiarity with a specific set of tables, a specific cloth speed, a specific habit of striking the ball.

I used to believe the second was stronger than the first. My notebook disagrees.

Core: the evidence chain

Data layer one — average per inning

The most important metric in carom is not match wins. It is the average per inning. A player who wins 40-30 in 22 innings has an average of 1.82; a player who wins 40-35 in 34 innings has an average of 1.18. The same win on the scoreboard, but two entirely different technical states.

I split those 41 matches into two groups by playing conditions, using average rather than final result as the measure. Here is what came out.

In the group that trains in this arena, the mean average over the first six innings was 1.64. The full-match average was 1.57.

In the other group, the first-six average was 1.41. The full-match average was 1.58.

Read the last two numbers. Full-match average is nearly identical: 1.57 and 1.58. A gap of 0.01 carries no statistical meaning at this sample size.

But the first-six-inning average diverges by 0.23 — roughly 16 percent. That is a large gap.

Interpretation: the home player warms up markedly better, but that advantage disappears before the match ends. The table advantage does not vanish; it is absorbed by something else. And what absorbs it, according to my data, is crowd pressure in the decisive phase.

A cross-check: of those 41 matches, isolating the 20 with the largest crowds, the home win rate was 45 percent (9 of 20). Isolating the remaining 21, it was 57 percent (12 of 21). Small sample, I know, and I will return to that point below. But the direction of the signal is consistent with the hypothesis: the fuller the stands, the thinner the home advantage.

Data layer two — the opening inning

In carom, the opening inning of a match is the hardest from an information standpoint. The table is untouched, no line has been read, and the player must choose between two options: a safe cushion-first escape to hold position, or an early attack.

I recorded the opening-inning choice across all 41 matches, and again the result split cleanly.

Home players chose the early attack in 61 percent of opening innings. Visiting players chose it in 44 percent.

The early attack in an opening inning has a lower success rate — by my notebook, around 38 percent against 52 percent for the safe escape. But when it lands, it opens long scoring runs.

This is where I misheard. My first reading was: home players are more confident, so they attack earlier. That reading is logically wrong. Home players do not attack early out of confidence. They attack early because of the crowd. A safe escape takes time. Nobody applauds a safe escape. The tactical choice in the opening inning of Vietnamese carom is decided not by probability but by the need to produce an event the stands can respond to. And that need is stronger in the home player, because the home player has a stand.

The result is a lower success rate precisely in the innings that lay the foundation for the whole game.

Data layer three — physical variables

I do not have enough data to separate the table factor from the crowd factor by pure statistics. So I did it another way: I logged the physical conditions of each arena.

The three conditions that vary most by venue are air humidity, room temperature and cloth age. In some northern arenas during the seasonal transition, evening humidity can differ noticeably from the same day's afternoon. Damp cloth slows the ball, shortens the roll, and most importantly changes the rebound coefficient of the cushions.

A player who trains in that arena three evenings a week automatically calibrates stroke speed to the average humidity of his training hours. If the match is played in a different window, that calibration drifts away from reality. This is table advantage in reverse — an advantage that turns into a disadvantage when conditions shift.

Of the 41 matches, I could isolate nine played at a start time at least three hours away from the home player's usual training window. Win rate in that group: 3 of 9.

I must be explicit: a sample of nine matches permits no conclusion whatsoever. I list it because it is the kind of signal I want to keep tracking, not the kind I want to price.

Data layer four — the no-crowd window

In the summer of 2026, when domestic sport had to run under crowd restrictions, I got a short window to observe billiards under something close to a control condition.

Events in that period were staged with very few or no spectators. I logged 17 semi-finals and finals under those conditions.

Win rate for the group that trains in this arena: 9 of 17. That is 52.9 percent. Against 51.2 percent for the full crowd-era sample, the gap is 1.7 percentage points. The sample is so small that I can say nothing beyond this: the signal does not contradict my hypothesis, and it does not confirm it either.

When the home table stops being a fortress, I learn to listen to an empty stand. And the empty stand, in this 17-match sample, appears to have taken from the home player exactly the share of advantage I had assumed sat in the table.

The contrarian turn: correlation is not causation

Here I have to brake, because this is where people who work with data, myself included, slip.

The entire chain of reasoning above can still be explained by an entirely different mechanism, and that mechanism has nothing to do with the crowd.

The first alternative mechanism is scheduling. At most domestic events, seeds and home players tend to be bracketed favourably. If home players meet weaker opponents in the early rounds and accumulate fitness better, then the better warm-up I measured may simply be the result of facing a lower-ranked opponent, not of familiarity with the table.

The second alternative is sample selection. I only log matches I watch in person. Those matches are not a random sample; they are the ones organisers judged attractive, meaning they feature a home player or a famous one. This is the bias every personal notebook carries, mine included.

The third alternative, and the one that worries me most: data cannot measure rhythm. A player who walks to the table with hot hands after three straight days of practice and a player who walks to the table after a four-hour flight may share the same season average on paper while differing completely in the first inning. My notebook records the first inning. It does not record the three days before it.

Those three mechanisms together are enough to substantially lower the confidence of the claim that the crowd eats the home advantage. I am not withdrawing the observation. I am lowering my confidence in it to worth continuing to track, not established.

There is one more thing I have to say, and say plainly. Analysts like me are pushing deeper into rooms where we are not present: the practice hall, the dressing room, the silence between two innings. We measure outcomes, sometimes we measure process, but we do not measure rhythm. A player told me after the Binh Duong semi-final that he never felt the crowd. He felt the table was faster than yesterday. The table was not faster. The arena's air conditioning was running slightly differently because the match started later. The feeling and the data are both true, and neither is enough.

A small blind spot: the exceptions

Two of those 41 matches break my description entirely, and I want to record them rather than sweep them aside.

The first: a home player won 40-19 in 16 innings, an average of 2.5 — nearly double his own season average. The second: a visiting player won 40-38 at an average of 1.05, in a match where both men shot below 1.2.

These two exceptions are not the same type. The second is the product of an arena with abnormal physical conditions — a table placed near the entrance, directly exposed to outside wind every time the door opened. The first is a player at a short-term peak, in the best week of his season.

These two types need two different handling rules. The second is a signal about playing conditions and belongs in the baseline data. The first is noise and should be removed from the sample. That I have bundled both into the same anomalous-case column for four years is a methodological error, and I am correcting the notebook.

What needs adding

To reach a firm conclusion I need three things I currently lack.

First, stroke-speed data for opening innings by time window. I need an objective instrument, not eyeball observation.

Second, a complete tournament calendar with each player's travel times. Without it, I cannot separate the fitness factor from the table factor.

Third, and most important: sample size. Forty-one matches is far too few. I need at least 150, spread across at least three arenas with clearly different physical conditions, before I dare publish any figure that could be used for pricing.

Takeaway

The signal I am tracking in the next tournament cycle is not the home win rate. It is the gap in first-six-inning average between the two groups, and how that gap evolves from the seventh inning onward. If the warm-up gap stays above 0.15 points while the full-match gap stays near zero, then the mechanism I described above is repeating. If both gaps widen together, I am wrong, and I will sit down and rewrite from the start.

Three thousand matches taught me that one match can teach more than all of them.