Does anyone know deep neural networks well enough to turn it loose on roulette? It would be interesting to see what method it came up with as being the most profitable over time, even if it loses. I tried to create a neural network on kaggle, but still learning how to set it up correctly.
I think it’s not different than any statistical approach, in other words and in the long term it won’t improve predictions accuracy
However the guys behind this website claim that they developed a NN machine learning that can beat the game, I tried to contact them without success, I think they discovered later that the approach loses
link:s://ntm-corp.com/blog/2020/1/26/ai-wins-3638-in-roulette
Quote from: Roulettebeater on Jan 17, 04:45 AM 2021
I think it’s not different than any statistical approach, in other words and in the long term it won’t improve predictions accuracy
However the guys behind this website claim that they developed a NN machine learning that can beat the game, I tried to contact them without success, I think they discovered later that the approach loses
link:s://ntm-corp.com/blog/2020/1/26/ai-wins-3638-in-roulette
That is cool, but instead of predicting the outcome, I was hoping more for an Alpha Go approach. Where it starts out not knowing how to play, then learns, where the goal is to make the most profit, even if it is negative. I would be interested to see what strategy it uses. It might ultimately decide the most profitable play is to not make a bet. lol