A Neural network Agent optimized by Evolutionary algorithms to play BlackJack or 21 (using just python/numpy/pandas)
This is a neural network agent evolved by genetic algorithm using only python and numpy library to play blackjack.
This implementation is slow and take some time to converge for good results.
I used a minimalist version of blackjack (A game known as 21 or vinte-um in Brazil).
The neural network use as inputs itself score, the player with higher score less than 21 and the cummulative sum of all cards played until that particular turn.