tanliyon / gym-xiangqi

This repo sets up the environment to play Xiang Qi (chinese chess) following the OpenAI Gym framework.
GNU Lesser General Public License v3.0
32 stars 6 forks source link

Remove Jiang points #72

Closed hojoungjang closed 3 years ago

hojoungjang commented 3 years ago

Description

We decided NOT to reward agents for making a Jiang against its opponents because this could be misleading in some way. We are not experts in Xiangqi and there are rather very deep hidden meanings and values to each move in a game like Xiangqi and Chess, so we think having a very basic and general reward system is better than having a sophisticated and specific reward system.

Therefore, currently, we only assign rewards when pieces are captured and when illegal move is made (negative reward i.e. penalize).

Following change is made:

Type of change

Not exactly a bug fix but an update!

How has this been tested?

GitHub Actions: CI