ABC can explain what were his problems at the start and how he solved it.
I can only highlight why FF would predict less spins then what is should do.
Theoretically FF on standard wheel should predict every spin.
If spin is not predicing, there is a reason for that.
FF is designed to work within specific limits.
That limit is based on time differences in between rotations and it is 120ms to 255ms.
Sure it can be done wider but there must be a limit somewhere. Wider limits will make system less accurate when doing different tasks during calculations.
Standard Huxley wheel which you can find in most casinos is about 180-200 ms.
So there is still plenty space for different wheel or errors in clocking and for difference if time is taken at different moment of spin. There are different controls for prediction and for what system will learn and save from spins. Limits are also changing based on ball speed. If we look one rotation at 1.00 sec per revolution the difference to previous rotation will not be same as if we had time of, for example, 1.200 sec. We can’t even estimate it, we can’t use one spin to define it, and we can’t even say will it increase or decrease. Because it changes. My first attempt, E5, used earlier prediction and depending on a wheel I was using, system was increasing. So all of the system was based on that. Nothing wrong except if you come to a wheel where that is not the case. Ball deceleration changes and we do not know when and by how much, but IQE6 does, because it analyses all entered data and selects what is important then uses it to find even better results.
After some time it narrows limits to get the best from your clocking and decides by where in time of spin you are and how much limit is allowed for prediction.
I found out why I started getting less predicted spins. But I am not sure was that part used in that format on old chip and when did I redesign those parts of program. Basically limit was set to tight allowing only +16 or -16 ms for error in clocking. It actually should be 32. It was 32 but I changed it to different way of calculation and due to that it actually becomes only half. Dependently how error is created the system should reduce it to zero or close to about 10 ms. If it is 10ms it may result in 2-3 numbers of errors or system will not predict. With some tests I disabled wheel calculation to test only ball clocking and test showed more then 50% spins where within single number. So from that we have errors in rotor clocking and small error in revolution when system positioning ball over predicted number. With all together system most of the time stays within 3 numbers of accuracy. Can’t wish better then that, remaining is only to find wheel where deformations on track (or slight tilt) wouldn’t affect ball so much and where randomness of scatter is lower.
Another reason.
If wheel has much slower / faster deceleration system may still predict, some spins we catch ball at 1 sec and it may be out of limits but maybe when ball is slower deceleration comes within FF’s limits. So we get only some percentage of predictions.
And last reason is of course really bad clocking.
I consider myself average clocker; I could run 100 spins (hand clocking) and get prediction on all of them.
That is why i was suprised with recent 70% predictions. :-\
We really do not need such accuracy because final result is not so much dependant on that.
If we have 100% accurate system predicts as this. Where zero is prediction to rotor hit
-4
-3X
-2X
-1X
0 Xxxxxxxxxxxxxxxxx
1X
2X
3X
4
Or the other one not so accurate and predicts as this
-4
-3X
-2Xxx
-1Xxxxx
0 Xxxx
1Xxxxx
2Xxxxxx
3Xx
4
5
After scatter we will probably have same result on the end.
On top of that we have deformations and diamond deflection which makes everything harder. But if you followed few recent posts with graph you will understand how much build up at same distance from prediction is needed that after ball bouncing we still have advantage. On some wheels it can be better then on my graph but on some it could be worst or close to nothing.
We can only predict based on detected ball speed (of course and rotor), if IQE6 can’t predict on particular wheel that most ball hits to rotor have constant distance to prediction, nothing will predict that wheel. There are no miracles, voodoo, or whatever someone may claim.
It is very simple, if ball with particular speed every times ends up differently, prediction is impossible.
But also if ball with same speed hits rotor with not enough consistency, that advantage after scatter is applied will be lost. Even if still exists it would be so small and it will take us thousands of spins to find where it is. Until then parameters on the wheel will probably change and we can’t take advantage of that.
