Pickleball Ratings Part Two: Exciting Ways Computer Vision Can Refine Player Ratings

PB Vision Team
December 19, 2023

Finding flaws in existing pickleball skill rating systems doesn’t require much digging. Although they can be useful, DUPR, UTPR, WPR, and others rely on some level of subjectivity, leading to difficulty in accurately assessing player skills. This is where the innovative application of computer vision technology comes into play, heralding a shift towards a more objective and accurate evaluation of player abilities.

What is Computer Vision?

Computer vision is under the umbrella of artificial intelligence (AI), and it enables computers to interpret and make sense of the visual world. By processing digital images, videos, and other inputs, computer vision systems use deep learning algorithms to identify, classify, and respond to various elements within these visuals. The implications of this technology in sports analytics are significant–many sports have already benefited from the application of computer vision. Balltime provides AI-generated stats for volleyball, Trackman helps improve golf performance, SIQ helps train basketball players, Team Mustard analyzes skills for baseball and football, and SwingVision presents stats for tennis and pickleball. However, in the realm of pickleball, PB Vision stands out by offering the most detailed analysis of the game data. SwingVision’s pickleball technology focuses on real-time applications like line calling, acting more like a virtual referee, whereas PB Vision emphasizes post-processing game analytics to achieve a high degree of accuracy and game insight.

Consider a technology that can scrutinize every moment of a pickleball match, where every shot, player movement, and tactical decision is captured and evaluated from video footage. Computer vision analysis involves breaking the footage into individual frames and applying sophisticated algorithms to extract meaningful data from these images. The result is a rich, multi-dimensional understanding of the game, distilled into actionable insights. 

Capturing Data for Analysis

To effectively access PB Vision’s computer vision analysis, you first need a high-quality video recording of your match. Ensuring the right framing and resolution is crucial for accurate data extraction. For tips on capturing the best footage, refer to our best recording practices blog post

Analyzing the Data: Game Insights

The backbone of PB Vision is its sophisticated algorithm, which employs machine learning and computer vision to meticulously analyze videos of pickleball matches. It assesses various aspects of player performance, including shot accuracy, types of shots, player movement, and error rates (see below). This analysis is then transformed into precise data models, offering unprecedented insight into a player’s performance.

Once the video is processed through PB Vision's system, it yields a wealth of insights. Players can explore detailed visualizations of their performance, such as heatmaps (see below) showing shot placements, analysis of player movement patterns, and breakdowns of error rates. These insights go beyond mere statistics, offering a deeper understanding of a player's strengths, weaknesses, and areas for improvement.

For instance, a heatmap may reveal a player’s tendency to favor one side of the court, or a particular stroke analysis might highlight a recurring technical flaw. Such detailed feedback is invaluable for targeted skill development.

Translating Computer Vision Analysis to Player Ratings

While PB Vision does not currently assess player ratings, it lays the groundwork for a new kind of rating system based on objective game insights.

Imagine a rating system where your score directly reflects specific, measurable aspects of your game, analyzed by advanced AI, namely large language models (LLMs). These sophisticated AI systems, renowned for their ability to understand and generate human-like text, could take our objective analysis of player performance to new heights. By integrating LLMs, PB Vision could achieve a higher abstraction of insight, extracting nuanced interpretations from the raw data. This system could analyze physical gameplay and provide strategic advice, mental game coaching, and personalized feedback by "reading" more into the data than we currently comprehend. This integration could transform how we understand and improve in pickleball.

A more objective rating system based on these insights could go beyond a simple decimal like ‘4.0’ or ‘3.5’--players could have a more detailed breakdown of their ratings in categories such as speed, backhand skill, maximum forehand drive speed, spin ability, shot decision making, reset capability, and serve quality. Instead of relying on tournament outcomes or subjective evaluations, this system would consider the intricacies of a player’s performance. No longer would a player’s rating be dependent solely on wins and losses or a partner’s skill level– ratings would be as accurate a reflection of a player’s true ability as possible.

Assessing a Computer Vision-Based Rating System


  • Comprehensive Skill Analysis: Provides a layered and more nuanced view of player abilities, encompassing technical, tactical, and physical aspects.
  • Dynamic Performance Tracking: Allows players to monitor their development over time, adapting to different game situations and strategies.
  • Inclusivity: Offers valuable insights for players at all levels, regardless of their participation in competitive play.
  • User-Centric Adaptive Approach: Adapts to individual player needs and aspirations, such as strategies, playing tempos, and preferred shot types.
  • Objective Evaluation: Significantly reduces biases related to external factors, offering a more balanced and fair assessment of skills.
  • Privacy Protection: Does not store biometric data to best protect player privacy.


  • Potential Inaccuracies: Computer vision is still error-prone and will never be 100% accurate.
  • Reliance on Recording Quality: The error margin largely depends on recording quality.
  • Essential Need for Player Tagging: Players must be manually tagged across games.

Improving with an Objective System

To make the most of a computer vision-based rating system like PB Vision, players should:

  • Focus on Targeted Feedback: Use the detailed analysis to zero in on specific weaknesses or areas needing improvement.
  • Embrace Consistent Evaluation: Regularly upload gameplay videos to take advantage of performance tracking over time.
  • Engage Actively with Tools: Utilize the platform's various tools and training recommendations to systematically refine your skills and strategies.

The integration of computer vision technology in assessing pickleball player ratings opens up a new world of sports analytics. With the potential to offer objective, detailed, and actionable insights, this technology can revolutionize the way players understand and improve their game. As we continue to develop and refine these tools, the future of pickleball ratings looks more precise, inclusive, and data-driven, enabling players at all levels to elevate their game based on clear, objective metrics. Stay tuned to PB Vision as we lead the charge in this exciting new frontier of pickleball analytics.

Want to learn more about PB Vision? Join our pickleball community and subscribe to our newsletter. Don't miss out—sign up now for updates delivered straight to your inbox. 👇

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
☝ Sign up for our newsletter!

More from PB Vision