How to access the last game, please! Thank you! about rlcard HOT 11 CLOSED

Hello, @jack86666 May I know what ‘last game’ refers to? Is it the previous game data or the last state of a game?

from rlcard.

from rlcard.

from rlcard.

@jack86666 To start a new game, you can call env.init_game(). Then call env.run(), you will obtain the game data (trajectories) of one game and payoffs. Unfortunately, RLCard will not save the previous game data. If you want to access previous game data, you can simply store the trajectories into a list and access later. Thanks!

from rlcard.

from rlcard.

@jack86666 In this case, you may not use env.run. Instead, you may use env.step which will take one step, and return the next state and the next player. Then you can do whatever you want when the game is in progress.

See the definition of step:

You can also refer to the document to get a sense of the interface design:
https://github.com/datamllab/rlcard/blob/master/docs/high-level-design.md

from rlcard.

from rlcard.

from rlcard.

@jack86666 The action space of Dou Dizhu is too large, thus we have to make abstractions to make it feasible for RL algorithms. Although we have added some rules for kicker, the agent needs to learn how to make 309 decisions. This is still challenging.

from rlcard.

from rlcard.

from rlcard.