Best Neural Network architecture for game prediction


For Forecasting the outcome of a game of League of Legends, I Want to Make a Neural Network. What is the ideal structure for the community?

As far as inputs proceed for each player (10 total):

  • Champion Mastery (Number symbolizing period performed with a champ)
  • Their positional rank
  • The champion they are playing
  • Their total ranked win-rate
  • Their win-rate on their winner

I’m outputting of winning the game, their portion

  • Is getting 50 inputs a lot of?
  • The number of samples should I get?
  • The number of hidden layers should there be?