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

BGonline.org Forums

Bot Comparison after 60/500 games.

Posted By: Michael Depreli
Date: Friday, 30 October 2009, at 6:36 p.m.

In Response To: Bot Comparison after 60/500 games. (MaX)

The pattern of GnuBG 2/3/4 ply errors on wrong doubles/takes/passes is just strange. one of the following:

- we need more games - there's something very bad in gnubg code - the idea of supervised training where 0ply is trained to mimic 2ply (instead of 1ply) is a bad idea, at the origin of the odd-even crap.

I had a look back at the original Depreli 626 positions. At the time the current gnubg build was 0.12 and on 2-ply it suffered from the same doubling and passing too soon problem that the current build has on 3-ply. This 2-ply problem was eliminated from build 0.13 which had a new neural net. I don't have any data for 3-ply from the old builds.

Why don't you ask Joseph Heled about the odd even effect he will surely know the answer. I know he's still around and replying to the mailing list.

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.