Wow! My new white ball predictor got another 2W out of 5 prediction. Two 2W hits out of two draw tests. 0.14% chance of that happening by uniform random selection chance. The predictor is actually five separate one white ball predictors, and the same one white ball predictor got a hit two draws in a row. This is just an interum predictor since I haven't completed the parameter search. so the overall 5 white ball predictor could improve.
Very encouraged by this class of predictor. It's been a long slog, but I might be onto something. It's about a PhD dissertation amount of work so far.
I have yet to apply this predictor class to the powerball, so I'm hopeful the same class of predictor will improve powerball prediction.
It's not all roses, though. The training predicts that this predictor only improves 1W, 2W, and 3W hits. Training did not have any 4W and 5W hits, which is where the money is. Since the predictor comes in five parts, it makes it easy to mix two high performing 5 ball predictors together, so hopefully I'll discover another 5 white ball predictor that performs as well.
I had to use my trusty computer for something else the past few days, but it's working away now on improving this class of predictor.
I have not applied this new class of predictor to mega millions yet. I will in the near future. Parameter search for this class of predictor takes over a week to complete in an interpreted language. I could speed things up many fold by rewriting in Cuda. This parameter search easily parallelizes across multiple cores.
PB and MM are nearing $80 million take home after taxes (in non-inflated, non-future dollar, non-annuity) jackpot, which is when I start to play these low odds lottos.
