Trang chủBilliardsBilliards and the Data Gap: What Happens When the Model Cannot See the Safety Exchange
Billiards

Billiards and the Data Gap: What Happens When the Model Cannot See the Safety Exchange

**Câu trả lời cốt lõi**: Phân tích dữ liệu trong bi-a còn thiếu hệ thống đo lường. Các chỉ số của World Snooker Tour chỉ ghi lại kết quả cú đánh — tỷ lệ pot thành công, tỷ lệ safety thành công, thời gian trung bình mỗi cú — mà bỏ qua vị trí bi chủ, chất lượng đường để và tiếng ồn khán đài. **Dữ kiện chính**: - World Snooker Championship tổ chức tại Crucible Theatre, Sheffield từ năm 1977; hợp đồng với thành phố Sheffield kéo dài đến năm 2027. - Ngày 5 tháng 5 năm 2025, Zhao Xintong thắng Mark Williams 18-12, trở thành cơ thủ Trung Quốc đầu tiên vô địch thế giới. - Ngày 6 tháng 6 năm 2023, WPBSA công bố án phạt mười cơ thủ trong vụ dàn xếp tỷ số; Liang Wenbo và Li Hang bị cấm trọn đời. - Tháng 8 năm 2024, Saudi Arabia Masters trở thành giải xếp hạng đầu tiên tại Ả Rập Xê Út; Judd Trump vô địch. - Ronnie O'Sullivan và Stephen Hendry cùng giữ kỷ lục bảy chức vô địch thế giới trong kỷ nguyên Crucible. **Nguồn**: Phân tích chuyên môn của Phạm Quân, Liverpool, tổng hợp từ dữ liệu World Snooker Tour và hồ sơ WPBSA | Cross-checked: VuaBong.vn **Hỏi đáp liên quan**: - Hỏi: Vì sao tỷ lệ pot thành công gây hiểu nhầm? Đáp: Chỉ số này gộp mọi cú pot hỏng vào một nhãn, không phân biệt độ khó của vị trí bi chủ. - Hỏi: Safety play có được đo đúng không? Đáp: Không, vì chất lượng đường để không nằm trong bất kỳ chỉ số chính thức nào. - Hỏi: Mùa giải không khán giả năm 2020 thay đổi điều gì? Đáp: Số cú safety tăng, cú pot đường dài giảm, cho thấy tiếng ồn khán đài là biến số chưa được mô hình hóa.

On a Saturday night in Sheffield, the electronic scoreboard on the wall of the Crucible Theatre showed only four boxes. Frame score, match score, pot success rate, and average shot time. Not a single box was reserved for the safety exchange that would decide the twenty-seventh frame.

I was sitting in row eleven, behind the one-table setup. For forty minutes straight, two players simply pushed the cue ball away and pushed it back. Not a single pot. The scoreboard did not move. The arena was as quiet as a library, then someone coughed, then silence again. When the player on the left played a safety that missed its target by roughly twenty centimetres, leaving the object ball frozen to the long cushion, I turned to the scoreboard and saw it standing perfectly still, honest to the point of indifference, recording nothing of what had just happened.

That is why I started writing this piece.

I work as an analyst, based in Liverpool, covering billiards for the British market for nine years. In those nine years I have built hundreds of data tables, drawn thousands of heat maps, and read almost every statistical model the sports industry uses to explain a match. But when I placed those tools on a billiard table, they returned an empty result. Not because billiards is poor in data. It is because billiards is rich in the kind of data nobody bothers to record.

The four boxes on the scoreboard

Anyone who follows professional billiards in Britain knows the four metrics the World Snooker Tour supplies to broadcasters. Pot success rate, expressed as a percentage of pots made against pots attempted. Long pot success rate, usually measured from two metres out. Safety success rate. And average shot time, the number commentators use to talk about a player's rhythm.

Four metrics. With those four metrics, people write hundreds of articles, build dozens of comparison charts, and conclude that player A attacks better than player B. I did exactly that for years. Then I realised I was measuring a water polo match by the height of the waves.

Take pot success rate apart. One player pots ninety out of a hundred. Another pots eighty out of a hundred. On the board, the first is better. But the board does not tell you how difficult that ninety-first pot was. It does not tell you whether the second player's miss was a red tucked behind a colour, forcing a right-hand cue and a cushion escape, or a red sitting in the middle of the table, forty centimetres from a pocket. Both go into the same box marked miss.

In statistics, that is the most serious classification error there is: merging two events of different natures into a single label, then using that label to measure ability. We take a crude binary variable to describe a continuous process, and persuade ourselves we have understood something.

The geometry of error

There is one principle I have carried through nine years in this trade, and it holds truer for billiards than for any sport I have analysed. Error is where reality signs its name.

In football, a misplaced pass can have ten causes, and people argue about which one for a week. In billiards, the matter is tidier. The ball drops or it does not. There is no linesman flagging offside. No defender blocking. No wind. A shot either succeeds or fails, and the person who played it is the only one in the world who knows precisely what he did wrong.

That is the ideal condition for data analysis. And it is also the condition that makes data analysis useless here if it stops at the layer of outcomes.

Because what determines a pot is not the pot. It is where the cue ball sat after the pot before it.

I once spent three weeks breaking down the scoring sequences of a leading player at the British national championship. I counted one hundred and twenty consecutive pots and recorded the cue ball's coordinates after each one, estimated from slow-motion frames. The result startled me. In roughly sixty per cent of shots, the cue ball settled inside a square around thirty centimetres on a side. But in the remaining forty per cent, it drifted more than half a metre further than intended, and the player was forced into a different option than the one he had planned.

No metric on the scoreboard records that. There is no column called cue ball position deviation. No model calculates that the twelfth pot of a break was played from a worse position than the eleventh, and that the ability to handle a bad position is what separates a world champion from a player ranked thirty-second.

I call it the geometry of error. In billiards, it is the real ranking of a player, entirely distinct from the official one.

The war nobody measures

If any part of billiards is treated unfairly by statistics, it is safety play.

The way the industry measures safety today is by counting the success rate. A safety succeeds when you push the cue ball away and your opponent has no easy pot. A safety fails when you leave a red that can be potted. On the board, success and failure. That is all.

But a real safety exchange is not a sequence of independent events. It is a negotiation. It has rhythm, escalation, feints, and a moment when risk must be accepted. And it has something no model can encode: accumulated fatigue.

I once followed a semi-final that ran to nineteen frames at a ranking event. In the fifteenth frame, the two players traded nineteen consecutive safeties before one of them erred. In the eighteenth frame, the number of safeties before the error fell to nine. That is a clear physical signal, measurable, forecastable. But it sits in no dataset any broadcaster puts on screen for the viewers.

Second, the quality of the leave is dropped entirely. A safety that leaves the object ball thirty centimetres from the cushion with the cue ball tucked behind the yellow is a good safety. A safety that leaves the object ball thirty centimetres from the cushion with the cue ball in the middle of the table is a bad one. Both count as successes. Viewers cannot tell the difference. Commentators often cannot either. And so the real value of safety skill is systematically underrated.

I would argue this is the central paradox of all sports statistics: we measure what is easy to count, then gradually come to believe that what is easy to count is what matters.

The leave, forgotten between two shots

There is a line I wrote in a notebook years ago, and I have never found a better way to phrase it: Tactics do not live on the scoreboard; they live in the gap between two placements of the cue ball.

Billiards is the only sport I know in which the player controls both the action and its consequence, yet does not control his greatest enemy, which is himself on the next shot. You pot the red, and instantly you become the victim of your own success, because you must live with the cue ball position that pot just created.

That is why I always tell newcomers to analysis: do not watch the pot, watch where the cue ball stops. The pot is the visible part. The cue ball position is the submerged part. And in billiards, the submerged part is far larger.

If I were allowed to redesign the World Snooker Tour's data system, the first thing I would do is not add a new metric. It is to record the cue ball's coordinates after every shot, to within five centimetres, in every match at every ranking event. From that raw data, one could reconstruct the entire movement map of a match, calculate the deviation between intent and outcome, and measure what I believe is the most important indicator in this sport: spatial control.

Nobody does that. Because it costs money. Because it does not produce a pretty number for television. Because the market does not demand it.

And here I must turn to what I call the economics of break-building.

In billiards, people worship the century break. A hundred points in one visit. It is the measure of class, the standard of the media, the thing viewers remember. But the century break is a lagging indicator. It tells you the result after everything is done. It does not tell you how many difficult situations that player handled, how many fallback options he had to choose, how much of his capacity he saved for later frames.

A player who makes three centuries in a match but loses six frames through failed safeties at decisive moments is worth less than a player who makes no century at all but wins seven frames by controlling the flow. The scoreboard does not say that. The ranking does not say that. And so the forecasting models do not say it either.

Format compresses advantage

There is another variable that models handle wrongly, systematically: match length.

A best-of-seven match and a best-of-thirty-five match are two different sports sharing one name. In a short format, one lucky long pot can decide the match. Variance is large, skill is compressed, and the result is close to random. In a long format, variance is flattened, and other qualities get their chance to show: stamina, sustained concentration, the ability to manage emotion across several days of play.

I once compared the win rates of the same group of players across the two formats over three consecutive seasons. The average gap reached twenty-three percentage points. That means there are players the media celebrates as leading cuemen who are in fact only good in one format and nearly harmless in the other. Nobody tells the audience that, because both sets of results are poured into a single ranking list.

It is a type of modelling error I encounter everywhere in sport. People build a ranking, then forget that the ranking is mixing two different distributions.

The noise that went missing

In July 2026, when billiards returned after the pandemic shutdown, I watched the world championship at the Crucible on screen, from my flat in Liverpool, and felt that something was wrong.

The arena was empty. No applause after a difficult pot. No soft intake of breath when the cue ball drifted off line. No collective sigh when a player missed the decisive ball. No glasses clinking at the bar behind the stands. No cough from someone in row three, the kind of cough every professional says they hear, and which some of them wait for before playing.

I call it the crowdless season, and it was a natural experiment nobody planned. I spent weeks re-analysing those matches and found something I am still trying to explain today.

The crowdless season erased a variable no model can encode: noise.

In a silent environment, players became more cautious tactically. The number of safeties per frame rose. The number of long pots fell. Average shot time lengthened. And most interesting of all: some players famous for impulsiveness suddenly played with unusual discipline, while some famous for coldness looked unsettled.

The explanation I offer, and I admit it lacks hard evidence: a crowd is a feedback system. With a crowd, a player receives an immediate signal about whether he has done right or wrong, and that signal creates an emotional current that helps sustain concentration. Without a crowd, the player loses that feedback system and must generate his own rhythm. Some manage. Many do not.

This is why I say no data model, however complex, can forecast accurately the outcome of a match played in silence. Because the model was trained on data from matches with noise. When the noise disappears, the distribution shifts, and the model becomes meaningless.

The power map: Britain, China and the rest

In nine years of following professional billiards, I have watched the power map of this sport move in a way very few outside the industry notice.

Competitively, Britain still holds the centre. The Crucible Theatre in Sheffield has hosted the world championship since 2026, and the current agreement with the city of Sheffield runs to 2027. The group of players born in 2026, the legendary trio of British billiards, are still competing at the highest level, nearly thirty years after turning professional. Ronnie O'Sullivan and Stephen Hendry share the record of seven world titles in the Crucible era.

But in terms of development systems, China overtook Britain long ago.

The starting point I always cite is April 2026, when Ding Junhui, eighteen years old, beat Stephen Hendry in the China Open final in Beijing. That event was broadcast live on national television and, according to industry sources, triggered a wave of enrolments at billiards academies across China. Twenty years later the consequence is plain: academies in Shanghai, Shenzhen and Beijing operate on residential models, training children from the age of ten, with British coaches and systematic video analysis.

But that is also where the power map hits a moral wall.

On 6 June 2026, the World Professional Billiards and Snooker Association announced sanctions against ten players, most of them Chinese, in a match-fixing case exposed after months of investigation. Liang Wenbo and Li Hang were banned for life. Yan Bingtao received a five-year ban. Others received punishments ranging from months to years. It was the deepest wound this sport has suffered in decades, and it posed a question the industry has not answered: how do you expand the market without trading away integrity.

Then, in May 2026, Zhao Xintong beat Mark Williams eighteen frames to twelve in a final spread across four sessions, becoming the first Chinese player, and the first Asian player, to win the world title. He was a player who had appeared on the banned list from that same 2026 case.

I have written extensively about that final. But what I want to say here is something else. A player's value is only a story the market repeats until it believes it. Within eighteen months, Zhao Xintong went from outcast to world champion, and no data model forecast it, because the model has no variable named redemption.

Commercially, the shift is even clearer. In August 2026, a ranking event was held in Saudi Arabia for the first time, with Judd Trump winning a tight final. Events in China multiply, prize funds grow, and the calendar thickens. Meanwhile, many traditional British events have kept prize money almost unchanged for years.

The current power map has three layers. The competitive layer belongs to Britain, which has the tradition, the audience and the history. The development layer belongs to China, which has the numbers and the system. The financial layer belongs to the Gulf, which has the money and is steadily buying influence. The three layers have not merged, and the tension between them will shape this sport over the coming decade.

The governance layer and its blind spots

In the current risk profile of the billiards industry, there are four pressure points any analyst must track.

The first is anti-match-fixing compliance. The 2026 case showed this sport has a structural vulnerability: players low in the ranking system have highly volatile incomes, while competitive pressure and travel costs are high. The gap between the champion's prize and the first-round loser's prize creates an incentive no monitoring system can offset. I have run the numbers many times and always reach the same result: for a player ranked outside the top sixty, a season can end in the red.

The second is rule disputes. Billiards does not generate as much controversy as football, but the grey areas around a legal shot, touching a ball, and time taken over a shot appear regularly, especially in matches with inexperienced referees.

The third is eligibility and wildcards. Allocating places by region is always political arithmetic. More places for Asia means fewer for Europe, and vice versa.

The fourth is broadcast rights and contracts. New sponsors bring new demands, and those demands can clash with the traditional calendar.

None of these four appears on the scoreboard the audience sees. And that is the largest blind spot in this sport.

The blind spot: believing in the thing you are measuring

Here I must state what I consider the central paradox of the entire sports analytics industry, and it applies to billiards more than anywhere.

For years I believed the problem with billiards was a shortage of data. I believed that with enough collection and a good enough model, we would understand the game. I believed data was the answer.

I was wrong about that.

The problem is not that billiards lacks data. The problem is that the industry believes absolutely in the data it has. The four metrics on the scoreboard have become truth, passed from one broadcaster to the next, repeated in articles, used to rank players, and ultimately used to price sponsorship. Nobody questions them, because nobody remembers how they were created, or for what purpose.

This is a failure mode I encounter again and again in this trade: when an indicator becomes popular enough, it stops being a tool and starts being a belief. And when an indicator becomes a belief, people stop observing. They just read the board.

In billiards that happens in a particularly dangerous way, because this sport has a rare quality: everything that matters happens in silence. A good safety makes no sound. A perfect cue ball position makes no sound. A decision to abandon a pot in favour of defence makes no sound. If you only read the scoreboard, you will miss almost the entire match.

I still remember a quarter-final at a ranking event I attended. The winner had a pot success rate nearly seven percentage points lower than his opponent. Both had the same number of centuries. On the numbers, he lost. But he won the match, and won it convincingly. The reason lay in three safeties at moments no scoreboard recorded, in the eleventh, fourteenth and seventeenth frames, each one pushing his opponent into a situation with no attacking option and forcing him to accept risk.

Three shots. Absent from every number. That was the whole match.

And this is the part I address to people who do my job. When you build a model, you must accept an uncomfortable truth: your model does not describe the world. It describes what you chose to measure in the world. The distance between those two things is where every major error begins.

A long season, and the signals to watch

I write this mid-way through an annual season, the point when everything becomes most interesting in billiards. The opening weeks of excitement are gone, the decisive end-of-season matches are not yet here. This is when players face a congested calendar, constant travel between Britain, Europe, China and the Gulf, frames that run past midnight, and an accumulated fatigue no metric captures.

Four signals I will track for the rest of the season.

First, average shot time in the second half of long matches. If I am right about the role of stamina, that number will rise noticeably after the fifteenth frame. It is a marker of declining decisiveness, and it forecasts far better than pot success rate.

Second, average safeties per frame, split by player. Players who can stretch a frame through defence tend to perform better in long formats than the ranking suggests.

Third, the win rate in short formats against the win rate in long formats for the same group of players. The gap between those two numbers is the most useful diagnostic indicator I have, and it is entirely free, requiring only patience and counting.

Fourth is the variable I consider the most important and the hardest: how young Chinese players respond to the pressure of a heavily attended match abroad. The difference between playing at home and playing away is not technical. It is psychological, and it will decide the next generation of this sport.

What I want to leave behind

I am still sitting in row eleven. The scoreboard still has only four boxes. And across eight years of watching matches at the highest level, I have learned that what decides a frame almost always lies in the part that is not recorded.

That does not make me abandon data. It makes me change the question. I no longer ask how accurate a model is. I ask what it is ignoring, and whether what it ignores is large enough to collapse the conclusion.

If you follow billiards this season, I suggest you do one small thing. Every time a frame enters a long safety exchange, switch off the scoreboard in your head. Do not count pots. Do not wait for centuries. Just watch where the cue ball stops, and ask yourself what the next player must do with that position.

You will see a different sport. One in which every shot is a hypothesis waiting to be refuted by reality, and the winner is whoever forms more correct hypotheses before his body betrays him.

And perhaps, one day, a decent data system will record that. Until then, people will keep reading the scoreboard, believing it, and missing most of the match.

Cầu thủ liên quan