Jordan Moore Data · Quant · AI

Watch · Snooker research · 2 minutes 13 seconds

A comeback, frame by frame.

Follow a Robertson–Bingham comeback through the table state and a frame-win research model, with a short explanation of how the picture becomes a probability.

Silent research replay, 2 minutes 13 seconds. Out-of-fold predictions come from a model that did not train on this match group. Some state smoothing uses later stills within the frame, so the sequence is an offline explanation rather than a live forecasting result. Broadcast footage © WST; research overlays by Jordan Moore.

What to look for

  1. The setup. At the 2025 Players Championship, Bingham leads the match 4–3. Robertson falls 29–66 behind in frame eight and needs snookers.
  2. The changing position. The panel tracks Robertson’s estimated frame-win chance through the break, safety exchanges and fouls. It is a model estimate, not the chance of potting the next ball.
  3. The explanation. A pause shows the captured still, ball detections, score and turn reading, and a top-down table. The cue ball is behind the blue, blocking a straight path to the brown.
  4. The finish. Robertson clears to win 67–66 on the black. The final curve gathers the frame into one visual story.

The comeback was selected to explain the model, not to represent its typical performance. The October evaluation compares models across a frozen population of held-out positions, rather than judging quality from this one result.