The G.A.T. engine is at a very good development stage. Probably at the end of this month (if everything runs as smoothly as it does now) there will be a version available to the public. Needless to say, the performance of the results is outstanding, ranging from 2x up to a few thousand times better compared to the HCD system of LA (and I assume anything else out there)!
So, here is a brief explanation of GAT and the concept involved. Draws, although seemingly look random, they do contain information that can be used to estimate how the next draw will be. If, for example, we concentrate on 20 subsequent draws, it is possible, up to an extend, to estimate how draw #21 might be like. So, GAT essentially tries to identify that relation among those 20 draws. I call this relation, a set of "attractors/repellers". Using this set, we can estimate (predict) draw #21. The engine formulates and evaluates many such sets of attractors/repellers, which are capable to identify very complex relations. At the end, we have sets that are capable to provide outstanding performance in hits! The process has nothing to do with the known and used Hot/Cold/Due systems, or any sort of analytical lowest-square or similar approaches, of course it is not a random number generator or really anything you have seen so far all these years regarding prediction. It directly inspects the random behavior inherent in the lottery draws and you get a direct feedback of performance over real tested draws.
A comparison between LA's HCD vs G.A.T. at the main numbers of the Greek lotto game Joker (5/45 field). This is what I got after a quick run (to match at least x hits) - might be possible to get even better results using other settings in GAT (especially the draws eval. parameter):
Code: Select all
requesting 15 numbers (tested over 100 past draws)
HCD [def.params - Combine (balanced bias)] : Match 0: 8, Match 1: 92, Match 2: 60, Match 3: 19, Match 4: 06, Match 5: 0
GAT [V.Deep - 300000 GAT, 50 draws eval.] : Match 0: 1, Match 1: 99, Match 2: 75, Match 3: 43, Match 4: 15, Match 5: 4
requesting 10 numbers (tested over 100 past draws)
HCD [def.params - Combine (balanced bias)] : Match 0: 27, Match 1: 73, Match 2: 32, Match 3: 07, Match 4: 0, Match 5: 0
GAT [Quick - 100000 GAT, 20 draws eval.] : Match 0: 09, Match 1: 91, Match 2: 51, Match 3: 20, Match 4: 6, Match 5: 2
GAT [Deep - 500000 GAT, 20 draws eval.] : Match 0: 08, Match 1: 92, Match 2: 53, Match 3: 20, Match 4: 7, Match 5: 2
requesting 6 numbers (tested over 500 past draws)
HCD [def.params - Combine (balanced bias)] : Match 0: 235, Match 1: 265, Match 2: 061, Match 3: 09, Match 4: 0, Match 5: 0
GAT [Quick - 100000 GAT, 20 draws eval.] : Match 0: 189, Match 1: 311, Match 2: 100, Match 3: 20, Match 4: 5, Match 5: 2
requesting 1 number (tested over 500 past draws)
GAT [Quick - 100000 GAT, 20 draws eval.] : Match 0: 411, Match 1: 89 (that is 17.8% success in picking 1 correct when requesting only 1 number!)
Code: Select all
requesting 1 number (tested over 50 past draws)
GAT [V.Deep - 1000000 GAT, 20 draws eval.] : Match 0: 42, Match 1: 8 (16%). For comparison, the probability to pick 1 correct number is 2.22%!
The logic in GAT is to identify which set to use (called a GAT table) and keep using it based on the hits it can produce over time. It is capable to provide any amount of requested numbers (1 to total balls - 1) and also can evaluate the bonus balls of the game (this gives further boost to the hits).
Not sure what more I can describe at this moment but all I can say is the waiting time for this brilliant program is almost over! In the next days I'll prepare the website for G.A.T. and I'll make an official announcement for the release!
cheers
lottoarchitect