The testing of the latest version has shown the timings are accurate to within 5ms. This is for just one revolution, not multiple. So if the actual ball revolution is 600ms, the Hybrid will detect it to within 595-605ms. The earlier version was accurate to within 15ms which was more than enough anyway. The 5ms accuracy is a big difference to the 50ms for any manual clocking computer. In simpler terms, the Hybrid v3 produces timings TEN TIMES more accurate than any roulette computer where you click a button to take timings, and it's all done completely automatically. For a manual clocking computer to achieve the same accuracy, multiple revolutions must be clocked, which results in predictions being too late on many wheels. Although the real benefit is that no player needs to even look at the wheel, so more can be won without drawing too much attention.
It is doing this with 30fps video. Anyone inexperienced in these matters may suggest such accuracy is not possible with 30 frames per second video. But the hybrid takes the location of detection and angle into account, then compensates to refine the timing. Distortion of ball image is not at all a problem (what a human and camera sees are quite different).
The actual timings are being tested with a 500 FPS video camera, so there is a negligible 2ms discrepancy.
Considering this is being achieved while the camera is being waved around to try and challenge the equipment, the overall results are better than I expected.
Coding for other hardware the Hybrid uses is about half complete. A full working kit ready for demonstration and/or use may be ready in roughly 2 weeks, but it is more conservative to estimate 4 weeks considering a lot of bug testing may be required. But either way soon I'll send a video to players waiting for it so you can see progress.
