Hi, I am implementing a new card game. I achieved good learning results with DQN, but with NFSP and DMC, learning does not progress and the results are worse than Random. I suspect that information that should not be known to the opposing agent is leaking. In a situation where there are actions that should not be known to the opposing agent and actions that can be known, where should I change from the existing game's Environments?
Hi, I am implementing a new card game. I achieved good learning results with DQN, but with NFSP and DMC, learning does not progress and the results are worse than Random. I suspect that information that should not be known to the opposing agent is leaking. In a situation where there are actions that should not be known to the opposing agent and actions that can be known, where should I change from the existing game's Environments?