search

dickreuter / neuron_poker

Texas holdem OpenAi gym poker environment with reinforcement learning based on keras-rl. Includes virtual rendering and montecarlo for equity calculation.
MIT License
612 stars 168 forks source link

after 700 k iterations agent just learns to go all in #78

Open neurosynapse opened 1 year ago

neurosynapse commented 1 year ago

Hello,

After 700 k iterations agent just learns to go all in all the time. Maybe the reward architecture should be different?

What is your experience?

Best regards, Roberts

hch1017 commented 1 year ago

Me too. After blind all in, the agent starts to play seriously and even wins dozens of times...... Weird

emirusahin commented 7 months ago

Doesn't that mean the agent is really good?