search

carlofazioli / cardiathena

A project to study strategies in the game of hearts, using distributed computing, AI, and data analytics.
GNU General Public License v3.0
6 stars 1 forks source link

bad_data #98

Closed davidjha closed 4 years ago

davidjha commented 4 years ago

I could use a closer eye on this. I believe this should fix the issue of have duplicate moves saved. The actual player moves are valid, it is only when saving the data that is the issue.

davidjha commented 4 years ago

The same state gets passed in twice, and the agent could then decide to play a different move on the second call to get_action. The game state vs the saved game state will differ.

At the end of the game, a final state is saved. The final state does not have a player action, so originally 0 was passed in and saved with the final state (final deck, final score). However I think it might make more sense if -1 is the final move because 0 (is the 2 of clubs) is a player move.

irobert4 commented 4 years ago

Oh. Yes, that's a definite oversight.

I'll go ahead and approve this. Hopefully it will also fix the other issue that I noticed in the data as well. I posted a screenshot to slack if you would like to take a look.