[quote=“bebediktus, post:32, topic:1171”]Forester no matter what we here talk - essence is found some solution . For now is - that we have one speed what we clocked but ball ends diferent. Here is starting point. Not look to tilt or to overlap clocked time is the same ball ended 0.88 rounds other. That 0.88 can be 0.25-0.5 what do prediction - wrong. And that i see too often in mine play, from that i think that what we use now for prediction is not enough , or we must use something other.
I always think abuot such variant - say we are predicted that ball will go 7.0 rounds , but it can do 6.75 or 7.25 the same as 7.0.
And that +0.25 can be longer in time or equal so from longer by say 600 ms till equal because ball simply moved faster, but falled after longer distance. The same with -0.25 can be shorter or equal.
And that can happend not looking to our statistick, because never statictick will be so much in front it will be always something 30-40-30 between 3 DD. And what we have in such situattions- ball fall earlier say prediction was 0 and that must be at 12 DD ball fall at 9DD earlier and here was number 1 at 12 DD was number 12 - so 0 even not come to 12 DD. If ball travel longer and hited to 3DD our prediction 0 pass 12 DD and in moment of hit over 12 DD was number 19 and ball hitted at 13. So from three theese possibilities we have possible hit to 1-0-13.
But that is hits - not where ball stops. But we must place bets where ball will stops not where it hits . So if ball had energy to hit to 0 but by some reasons it hited earlier to 1 - probably it will scater longer say +20 pockets when it will hit normally as expected he will scater +10 pockets in averidge when it wwill hit later say in averidge it will not jumps at all and fall where it hited . This way instead disperssion 26 pockets (1-0-13) we have disperssion only 6 pockets.
Now imagine that we colect as data not only hits but also scatering from diferent diamonds but not separate from diamonds but in such way - we are predicted as hit to 12 DD - ball hited to 9 DD we colect scater - we are predicted 12 DD ball hit right 12 DD we colect scater - we are predicted 12 Dd ball hited 3 DD - we colect scater.
So not only scater from Dd but in relationship what was prediction. This way easy can be situattion that diferent scater can compensate diferences in hitting place. This way we can have 4 diferent variants , where every variant have three subvariants. Program can compare them between and find best prediction for every variant.
Maybe possible analyse more deep when hit is to oposite DD or even after 0.75 or even 1.0.
Now in corections we have only speed - prediction and outcome, but finally in data is speed and outcome without relationship to what was predicted.
So mine mind is how to make that will be hiting prediction - real hit - outcome. Maybe even something more. And that program will look for relationship between data.
So mine conclussion is that algorythm which we use for prediction is too simple and it not able to solve tasks which are given for it.[/quote]Finaly some reasonable tolk. Jump depending on energy ball has to spend. Less decceleration- more trevel, or more jump and vise- versa. Even so would be nice to avoid back spins in prediction, these jump different.