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

BGonline.org Forums

of course backgammon has been solved

Posted By: Casper van der Tak
Date: Wednesday, 6 April 2011, at 6:36 p.m.

In Response To: of course backgammon has been solved (LowRayTeeng)

The strongest bots use neural nets. These have been trained by playing against themselves. First play is random, that the bots get feedback and learn, and the games become more and more similar to normal BG games. In later training they will get to play from selected positions so as to address some of their experience limits.

As a consequence, bots become very good at certain positions that arise easily & naturally out of the opening position, or other positions that have been used for training purposes. But that is only a very small subset from all positions. For the vast majority of positions, bots are clueless, and for some of these, through logical thinking, humans can do a lot better than bots. For example, late hit and containment positions - almost all checker born of, and where a lot of checkers have been hit and send back - a human will figure to build a prime, GNU for instance will stack more and more checkers on the mid (the record of GNU against me containing a single checker was 10 on the mid). A bot rollout for a position where the bot is clueless should not be believed - the results are meaningless.

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.