Introduction

A judgement framework allowing two agents to compete with each other, and make sure they’re following the game rules.

Usage

Implement agent

Implement your own agent following the interface IAgent. Please refer to agents for examples.

#include "judge/IAgent.h"

class MyAgent : public judge::IAgent
{
  // Implement your own agent
};

Feed start board

There are hidden information like cards held in opponent’s hand, but a game state is a state that all hidden information is already determined. So, typically there are multiple possible start boards. A callback function will be invoked by the judger to get a start board. Typically one can do a sampling on all possible hidden information over a predefined distribution.

#include "state/State.h"

state::State GetStartBoard()
{
  // Return the starting board
}

Invoke judger

#include "judge/judger.h"

int main()
{
  judge::Judger<> judger;
  MyAgent agent1;
  MyAgent agent2;
  judger.SetFirstAgent(agent1);
  judger.SetSecondAgent(agent2);
  judger.Start(GetStartBoard, agent1, agent2);
  return 0;
}

The neural network also uses this judge framework to generate training dataset.