0

Now my task is to use model checking to find a best strategy for card game Love Letter, and I need to model the game using Markov Decison Process first.

I have done a lot of resaerch and I decided to use "what cards are in the deck and what cards are in the other player's hand" to be my state.

But I am not sure which will be the best way to represent the states, because I think the states should include all the information I need and I can make action base on these states that I have got.

Oliver Mason
  • 5,322
  • 12
  • 32
jiaxin886
  • 1
  • 1
  • 2
    I don't know the game "Love Letter"; can you add a reference to it, please? – Oliver Mason Jul 11 '22 at 13:40
  • Please clarify your specific problem or provide additional details to highlight exactly what you need. As it's currently written, it's hard to tell exactly what you're asking. – Community Jul 11 '22 at 13:53
  • I have added a reference about card game Love Letter, thank you :) – jiaxin886 Jul 11 '22 at 14:04
  • 1
    I have read the description on the link. You should note the game as it is usually played requires secrecy (not knowing what is in other players' hands), and that secrecy is a major part of the game. This means that it is not possible to construct an MDP for the game as it is usually played. Are you proposing any changes to game in order to make it into an MDP? It may help to understand your underlying goal here - e.g. are you trying to learn about MDPs and picked this game because you thought it might be a challenge? Or are you trying to write a learning agent specifically for Love Letter? – Neil Slater Jul 11 '22 at 20:54

0 Answers0