Implement a gladiator ring that "pits" each method against each other.

Methods include:

• Random
• Policy Gradient
• Tabular
• MCTS

• Implement scripts that train REINFORCE and "play" policies against each other.
• Implement trajectory generation in env similar to how it's implemented in REINFORCE right now.
• Build some notion of statistical confidence in comparison

Score non-optimal policies for hex

Solve for optimal state values for stochastic tic tac toe

Changes to evaluation and policy gradient:

• Implement nxn evaluation
• Update policy gradient to be epsilon greedy

Wrap any pettingzoo environment for a turn based game into a richman-compliant environment.

Implement policy gradient and random policies on Hex

1. Implement policy base class: RichmanRL/policy.py
2. Implement NN policy for subgame: RichmanRL/policy.py
3. Implement reinforce algorithm for subgame learning: RichmanRL/reinforce.py and RichmanRL/main.py