We turn pool into data
PerformStars uses computer vision and artificial intelligence to analyze pool matches, shot technique, and practice sessions.
About PerformStars
PerformStars is a technology platform that uses computer vision and artificial intelligence to automatically analyze pool, understand shot technique, and make training more effective.
The project began in 2020 with a simple question: could an ordinary video of a pool match be turned into accurate data that actually helps a player improve?
Today, PerformStars can automatically identify shots, locate the table and balls, reconstruct ball movements and trajectories, determine shot outcomes, and compile player statistics. Instead of spending hours reviewing footage manually, players and coaches receive a structured digital analysis of the game.
Our ambition goes beyond automating statistics. We want to measure skill, understand how it develops, and build technology for personalized training grounded in real performance data.
From video to a digital model of the game
At the heart of PerformStars is our own computer vision system.
We train specialized models based on modern neural network architectures to detect balls, segment the playing surface, and classify ball types and table layouts. A sequence of frames becomes a shot episode: the system identifies when the shot begins, tracks objects, reconstructs trajectories, and analyzes the outcome.
Additional machine learning models assess the quality of detected episodes, filter out false events, and help classify game situations. PerformStars analyzes what happens on the table over time, rather than relying on a single image or the final shot result.
Analysis can run on server infrastructure or locally on a dedicated computing device beside the table. This supports the development of a club system that processes live video in real time.
Explore the platformBuilding better training on real data
Over the course of its development, PerformStars has collected data on more than 100,000 shots recorded by users. We have also built a separate research dataset of approximately 300,000 shots played by leading international players in professional matches.
To us, this is more than a video archive. Each episode can provide an example of how an elite player solves a specific problem on the table.
This extensive record of professional play opens the way to the next stage: comparing a developing player’s decisions with patterns of expert performance and automatically selecting drills that address their individual weaknesses.
We want the system to answer more than “What happened?” We want it to answer “What should this player work on next?”
Looking beyond the shot result
Pocketing a ball does not, by itself, tell us how well someone played. One shot may be straightforward; another may demand exceptional accuracy, speed control, complex geometry, and positional judgment.
This led us to research mathematical methods for assessing shot difficulty. We developed an approach that accounts for the geometry of the position, distances, angles, ball locations, and other factors. It provides a foundation for a more objective assessment of skill: a player’s result can be considered alongside the difficulty of the task they faced.
The next mathematical challenge is to search efficiently for similar positions among hundreds of thousands of recorded shots. In the future, the system could draw on professional play as a store of experience, finding examples of how leading players approached comparable situations.
This research points toward elements of machine “pool reasoning”: a model that can recognize a position, compare it with previous experience, and evaluate possible actions.
Connecting video with player movement
A pool shot involves more than the movement of the balls. The player’s arm movement, acceleration, stability, and consistency of technique matter just as much.
That is why PerformStars is developing multimodal analysis: synchronizing arm movement telemetry from wearables and smartwatches with the shot detected by the computer vision system.
This allows us to study the relationship between movement technique and actual results on the table, with the longer-term aim of identifying each player’s individual movement patterns.
A personal digital coach
PerformStars uses language models to help interpret the statistics it collects. They translate large sets of metrics into explanations players can understand: identifying strengths and weaknesses, spotting recurring problems, and suggesting ways to improve training.
Our goal is to move gradually from conventional statistics to a personalized training system that knows a player’s history, understands the patterns behind their mistakes, and proposes a training program based on their actual play.
The technology is designed to support coaches. It adds an objective layer of evidence and helps them work more effectively with more players, including remotely.
Technology for a range of cue sports
Pool is at the heart of this platform: from recognizing the cue ball and object balls to analyzing pocketing, cue-ball control, and position play. The broader PerformStars technology also supports other cue sports. Their rules differ, but the underlying principle is the same: a camera turns play into structured data for training and analysis.
The app is already available in eight languages. We are building PerformStars as an international platform for the global cue sports community.
From research to a working product
PerformStars has moved beyond the laboratory prototype stage. Our iOS and Android apps and web platform are live. Users can upload practice recordings, receive automated shot analysis, and explore their statistics. We are also developing tools for coaches and clubs.
A separate area of development is club infrastructure built around fixed cameras and local computing devices. We assemble, configure, and adapt this equipment ourselves to capture and process play in real time.
The project is now entering commercialization and international expansion. We have made our first sales and are working with players, coaches, clubs, and cue sports industry representatives in different countries.
Mobile
Record practice sessions and analyze your game in the app.
Web
Explore statistics, review your training history, and track results.
Club / Real-time
Fixed cameras and local processing of live video.
Testing our technology openly
For sports analytics, knowing the system’s real accuracy is essential. We publish the results of PerformStars validation tests on complete practice recordings and provide the source footage so others can examine the results.
In the latest published test, the system detected 93.5% of the shots actually played. Among the episodes for which it made an outcome decision, accuracy was 97.2%.
See the validation resultsResearch and intellectual property
PerformStars has grown from a practical product into a research project in its own right, spanning computer vision, machine learning, and mathematical modeling of sporting performance.
One of our key developments, a method for determining and visualizing ball trajectories on a billiards table, is protected by a patent. Our proprietary technology provides a foundation for continued product development.
Our research continues across mathematical assessment of shot difficulty, analysis of movement patterns, and methods for quickly finding similar game situations in large sports datasets.
PerformStars in the global cue sports industry
In 2025, we presented PerformStars at the Guangzhou Billiard Exhibition in China.
We demonstrated the project to professional players, cue sports industry representatives, and equipment manufacturers. It was a valuable opportunity to hear directly from the people shaping the sport worldwide.
This international focus is now becoming one of the main directions of PerformStars’ development.
Sharing the science behind pool analysis
Alongside developing the technology, we share practical knowledge about computer vision, artificial intelligence, and their application to pool.
Our Research & Engineering articles explain how ball detection, geometric calibration, tracking, and physical modeling work together to reconstruct a shot.
We see education as an essential part of the project. Players and coaches should be able to understand what the system measures, where its limitations lie, and how to use the results to improve.
Explore Research & Engineering
The PerformStars community
An international pool and cue sports community has gradually grown around PerformStars.
Our combined audience across social media and video platforms is approaching 50,000 subscribers. We regularly share shots, professional matches, visualizations, and analysis, while experimenting with new ways to present sports data.
Individual videos created using PerformStars footage and technology have attracted more than a million views.
This community does more than help us reach new people. It shows us, day by day, which game situations interest players, gives us direct feedback, and helps us understand whether new ways of presenting data make sense to the people using them.
Where we are heading
The next stage for PerformStars is to bring our accumulated technologies together into a comprehensive, intelligent training system.
We are developing personalized recommendations, automated drill selection, comparisons with examples of professional performance, movement telemetry analysis, and real-time club infrastructure.
Over time, a common set of digital metrics could bring a new level of consistency to pool: making player progress measurable, results across training sessions comparable, and the experience of professional players available to the next generation.
Our longer-term research goal is a system that can understand a table position, draw on accumulated experience, and evaluate possible solutions. Beyond a digital coach, this technology could eventually support autonomous robotic systems.
Let’s move pool forward together
PerformStars welcomes collaboration with players, coaches, clubs, academies, federations, tournament organizers, equipment manufacturers, research organizations, investors, and technology partners.