Professional snooker players are required to excel at a range of different skills, the most important being potting.

A snooker game in progress, the blue ball is centred, and the cue ball is to the top right.

Take a look at the image above, a game between Ken Doherty and Mark King. An interesting question here is “How likely is it Mark King will pot the blue ball”?

In Football, a similar question would be “How likely is this player to score from here?” A statistic to answer this is Expected Goals (xG), which takes a shot and assigns it a score between 0 and 1. This score being the probability of the scoring a goal. Sam Green talks about this in the article “Assessing the Performance of Premier League Goalscorers” (https://www.statsperform.com/resource/assessing-the-performance-of-premier-league-goalscorers/)

In this article, I present the new snooker statistic Expected Pot (xP), inspired by xG. Expected Pot attempts to answer the question: “How likely is it Mark King will pot the blue ball?”

What is xP?

Expected Pot (xP) is a number between 0 and 1 that represents the probability of the target ball being potted successfully into the target pocket. The target ball being the red or coloured ball that the player is aiming to pot. The pocket being the pocket the player is aiming to pot the target ball into.

Look at the image below, in this scenario the target ball is the blue ball in the centre of the table. There are two separate pockets which the player would attempt to pot the blue ball into, the middle and bottom pockets on the right. The middle pocket is the easier of the two to pot the ball into, and therefore would have the higher xP.

Possible xP ratings for these shots could be:

  • 0.9 xP into the middle right pocket, meaning the player would pot the ball 90% of the time.
  • 0.6 xP into the bottom right pocket, meaning the player would pot the ball 60% of the time.

A snooker table. The cue ball is in the middle on the left, and the blue ball is right in the centre. A white line represents an easy pot in the middle right pocket. A purple line represents a harder pot into the bottom right pocket.

Why use xP?

For newcomers watching snooker, understanding the tactics being played can be challenging. These viewers will be relying on the context provided by commentators, with a number of visual overlays at their disposal. These visual overlays provide an excellent usecase for xP.

Where can xP be used?

A great use of xP could be to show an on screen popup displaying the difficulty of the shot the player is attempting. This communicates to the viewer the risk and reward involved in the shot.

A snooker table, the player is attempting to pot the red ball. The expected pot statistic hovers over the ball. It shows an xP of 0.81.

Another opportunity is to generate an xP heat map. This would show the the ideal location for the cue ball to be in. Look at this shot below, Mark Williams is lining up an attempt on one of the red balls. Following this shot, Mark will probably want the cue ball to be in a good position to pot the black. The broadcasters could use a visualisation to display a heat map showing the “good” places for the cue ball to be.

A snooker table, a heat map is layered on top show the ideal place for the cue ball to be to pot the black ball into the bottom right pocket.

There is even further potential if you consider how players could use xP to enhance their skills behind the scenes.

How is xP made?

This first iteration of xP is powered by a machine learning model. This model is trained using hundreds of historical snooker shots from a number of matches. For each discrete shot, a number of features are extracted:

The coordinates of the cue ball. The coordinates of the target ball. The pocket the player is attempting to pot the ball into. Whether the target ball is potted successfully. Below is an image showing the entirety of the dataset powering this model. The arrows plotted show the cue ball to the target ball, and then the target ball to the target pocket.

A graph showing all the data. Arrows from the cue ball to the target ball, and target ball to the target pocket.

Blue Arrow = Successful Pot, Red Arrow = Missed Pot

Where did the data come from?

I was initally unable to find data that could be used to power this model, which meant that I had to generate my own. I created a tool which could be fed screenshots of snooker matches, like those in this article. By using computer vision, and neural networks, this tool would turn those images in to data that would train the machine learning model.

I am hoping to write about some more details about the technical aspects of this project at a later date - so please keep an eye out!

The future of xP and Snooker Statistics.

This first iteration of xP is a proof of concept. A model like this requires a large amount of data, which I was unable to generate alone. Given further data and an improved model, there is loads of opportunity to take xP and turn it into something really amazing. Baseball, Basketball, and Football are all sports which have been revolutionised by using data to its full potential, there is still plenty of room for people to make a name for themselves in the snooker data world.

Thanks for reading!