I made an experimental program.
It is designed for tilted wheel.
The program can predict at any time during the spin.
For example clock the ball at 15 sec. to the end or 5 sec. to the end and get reasonably same result. It is very simple program.
Result would match prediction as someone with extraordinary visual skill can spot on when the ball is about 15 sec to the end and make appropriate adjustment for faster/slower rotor.
Program is not accurate in prediction as TILT 2 program but it will produce advantage. Also the program doesn’t have “point set” to avoid spins where the ball will not hit dominant diamond.
Main advantage of this program is more flexibility so it can be applied on wheels where the ball decelerates very fast. We can clock ball with speed of 0.3 sec per rotation.
Even if we are wrong by full rotation we do not miss by much since rotations are short and errors are only how much rotor will move for that short time.
In addition some additional inaccuracy comes because rotor is clocked very early in time and of course it needs to be calculated all the way to the end.
I build the program on FFA hardware. FFZ wouldn’t be suitable since main objective of program is to get early prediction as much as it is possible. Zap wouldn’t be very efficient with fast ball.
Calculation for this program is very simple and it looks as it works fine.
I did some tests so if I clock the ball at 15 sec to the end, then next time at 12, then 10 ,8,6,…etc. and result deviates only by few pockets.
14 rotations to end predictions = 31, 28,9,12
12 rotations to end predictions = 7,29
10 rotations to end predictions = 7,28,35
8rotations to end predictions = 28,12
6rotations to end predictions = 35,28
Snowman made me think about this aproach. 
If TILT 2 doesn’t use “point set” I would probably work more on this program and adopt it as my favorite. Performance would be equal perfect VB at any time during the spin.