Been lurking a while and finally curious enough to ask. With all the machine-learning stuff everywhere now, I've been tinkering with running models against historical draw data just to see what shakes out — Monte Carlo, Markov chains, a couple of neural-net experiments.
Upfront: I'm not claiming any of it beats the odds. Draws are independent and I know how that math works. It's more that I find it fun to compare what different methods "favor" and watch how wildly they disagree with each other on the exact same data.
A few things I'm wondering:
- Is anyone here running their own models, or is everyone sticking with wheels, frequency charts, and hand-built systems?
- For those who've tried the ML route — did any method turn out more interesting than the rest, or did they all basically converge on noise?
- Does anyone bother running this stuff on the newer games like Millionaire for Life, where there's barely any history yet, or is it pointless until there's a real sample size?
Not trying to sell anybody on anything — genuinely curious what this crowd thinks since you all actually know the math. Roast me if I'm overthinking it.
