Scoring Assumptions

The algorithm assumes standard rules: games played to 11 or 15 points, with a 2-point margin for victory, and normal (non-rally) scoring. We’re working on relaxing these assumptions, but for now, deviations may result in errors or the algorithm not displaying results.

Ensuring Optimal Recognition

• The primary input is the server and serve order.

• Ensure all players are clearly visible in the video. If a player is cropped or obscured, the algorithm might guess incorrectly. (See video/framing tips)

• Additional data like line calls is used, but with less confidence (after all, if it’s hard for you to judge an “out” call, it’s challenging for us too).

How the Algorithm Handles Scoring

• It tries to reconstruct a plausible scoring sequence with minimal corrections. Sometimes this means adjusting what might seem like clear and obvious rallies if errors occur earlier in the sequence.

• The final score is prioritized, but the running score may occasionally be off by a rally or two due to ambiguous situations.

• Unfortunately, there’s no way to manually correct the running score in the current interface at this time.

When the Algorithm Makes Mistakes

• Errors by players (e.g., serving out of turn) can confuse the algorithm. Without manual correction tools, it can only make educated guesses.

• If the algorithm struggles to generate the correct final score, you can help by entering it manually in the Overview tab. These corrections are incredibly valuable for improving the system.

Additional Tips for Best Results

• Remove warm-up footage from the start of the video. The algorithm doesn’t process warmups, but you can send them separately for insights on your practice shots.

• If you use score-tracking software or manually track scores, feel free to share to enter the score manually on the Overview tab. Rally-level details help us refine the algorithm further.

The Bigger Picture

We’re excited about the potential of this scoring algorithm. Beyond scorekeeping, it paves the way for deeper statistical analysis of shot outcomes and decision-making patterns. Stay tuned—we’ll soon share more about these developments!