Behind the Scenes of VBR: How We Measure True Pickleball Skill with Vision-Based Ratings

At Dinkmate, we’ve always believed that ratings should reflect more than just wins and losses. Ratings should capture how you really play — your decision-making, your shot selection, your teamwork. That belief is what drove us to create VBR: Vision-Based Rating.

If you’ve played pickleball before, you’re already familiar with the 2.0–8.0 scale. VBR lives in that same world, but with a twist: it’s a floating number to two decimal places. Think 3.47 or 4.82 — small shifts that reflect meaningful differences in skill, just like the ones coaches and players notice in real life.

But what does it take to calculate a number like that from just a video? Let’s open the curtain.

Why Camera Setup Matters

Every VBR journey begins with video footage of a match. From there, we lean heavily on computer vision. Using techniques like: object detection & segmentation to find the ball and players. Court detection to anchor everything in space. That’s why framing matters so much. The better your video captures the court and players, the more accurate your rating will be. We’ve put together a camera setup guide to help players get this right — because a clear view means better data.

From Pixels to Performance

Raw video only tells part of the story. The real work comes after. From each match, we extract over 500 distinct features. These aren’t just basic stats like “winners” or “errors.” They capture subtleties like: shot variety and pace, serve consistency, rally effectiveness by its type, team positioning and many more. This is the layer where your VBR starts to take shape.

Expert Knowledge, Modeled in Data

Numbers alone can’t define a player — not without context. That’s why we leaned on thousands of hours of footage rated by expert coaches. Their expertise provides the foundation, and our statistical modeling maps that expertise into a repeatable system.

The result isn’t just a number. It’s a reflection of how expert coaches would see your game, distilled into a data-driven, objective rating. Even better, our models highlight why. They can point to correlations — serving strength, drinking proficiency, team chemistry — showing what drives your rating up or holds it back. That's what we called personal performance insights.

The Future of VBR

We’re not stopping here. The next generation of VBR is already in the works. We’re exploring how to factor in age and sex, since physical attributes can play a role in how performance translates across levels.

Our goal is simple: make VBR the most accurate, fair, and transparent skill rating in pickleball. Not just a number, but a mirror of your true game with the help of machine learning and data of course.