TABLE VISION
HONEST DEMO MODE uses a synthetic camera source. The same normalized observations feed the production YOLO + Picamera2 adapter.
HONEST DEMO MODE uses a synthetic camera source. The same normalized observations feed the production YOLO + Picamera2 adapter.
Make the play. Track the shoe. One score grades both skills.
Basic strategy picks the move. The count measures when the composition of the remaining shoe changes the training scenario.
Vision updates the count, the policy explains the decision, and two distinct pulse patterns keep the messages separable.
Your accuracy comes only from submitted answers. System coverage is shown separately and never blended into your score.
See exactly where count drift changes otherwise-correct play.
Combined strategy and exact-count accuracy from your submitted answers.
Every supported decision and count update maps to an explainable recommendation and haptic code.
Coverage is a software capability measure—not a claim of win rate or guaranteed outcomes.
| TRUE COUNT | ≤ −2 | −1 | 0 | +1 | +2 | +3 | ≥ +4 |
|---|---|---|---|---|---|---|---|
| ATTEMPTS | |||||||
| STRATEGY | |||||||
| COUNT |
Empty cells mean no hands have landed in that true-count bucket yet.
| HAND | PLAYER / DEALER | YOUR PLAY | POLICY | COUNT | RESULT |
|---|---|---|---|---|---|
| Complete a Trainer hand to populate this trace. | |||||