[ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

BGonline.org Forums

The next major revolution in backgammon bots

Posted By: eXtreme Gammon
Date: Wednesday, 9 May 2012, at 1:56 a.m.

In Response To: The next major revolution in backgammon bots (Timothy Chow)

If Tournament directors would allow bot to enter, this would be a great motivation for developer to work on that.

That rock-paper-scissors programming contest is amazing. I've no idea how to program a pattern finding program. Though it seems that some of these programs had to be very predictable for others to take advantage of them (OT: my brothers and I were playing a variation of that game with a "well", the well beat rock and scissors, but loses to paper. the asymmetric game was more "interesting")

There are several ways I can imagine approaching the problem for Backgammon.

  • 1- Asymmetric MET (mainly for cube decision, but will change a few checkers play)
  • 2- Specific NN to give an average Total EQ lost for a user of a particular Elo, that value can be added to the Equity to determine the best play against a weak player.
  • 3- Background analyze of the player as the game progress to estimate the player strength and adapt the 1-2 point.
  • 4- Keep a db of player games to determine particular weakness (Always drop too soon, take to late, misplay backgame, etc...) to adjust decisions.

    The problem become VERY complex if you want to take into account that most of the time you will have only little information to estimate the opponent strength. For instance it may be valuable to drop a close cube just so the match is longer and you can gather more move from the opponent.

    Humans are much better at that because they can take into account the opponent body language and other not quantifiable behavior. For instance you are losing and roll a mega joker, snatch a close cube that is a big take and you opponent drops it in a second: a human will register: "After anti joker, the opponent will drop anything, next time, Consider doubling even if incorrect. A bot will have hard time to process that and only will register "player estimates incorrectly that position and/or may drop too early"

    The main question: what is the value of that huge work for developer of backgammon program (ignoring the interesting programming challenge)?

    I see a few:

  • Again: let the bot play in tourney! I would surely like that and work on finding a person to operate XG in most of the BG Tourney in the World! side effect: it will increase attendance!
  • Added value to the program as it will allow player to analyze game taking into account player strength. but that is limited (IMO) to only the fringe of the backgammon players.
  • The program will beat its user even more. Personally, I do not care about that. What matter to me is that the program give the most accurate information and shows the player the best move.
  • Any Others?

    • Messages In This Thread

     

    Post Response
    Your Name:
    Your E-Mail Address:
    Subject:
    Message:

    If necessary, enter your password below:

    Password:
     

     

    [ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

    BGonline.org Forums is maintained by Stick with WebBBS 5.12.