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

BGonline.org Forums

GNUBG RO Blunder(BGB/XBG right)

Posted By: Robert-Jan Veldhuizen (Zorba)
Date: Sunday, 28 June 2009, at 1:59 p.m.

In Response To: GNUBG RO Blunder(BGB/XBG right) (Andreas)

GnuBG has always had trouble with more extreme, highly assymetric positions.

What is extreme? It's hard to say exactly, but always be suspicious of GnuBG's results with: more than 5 checkers back; pip counts over 200; more than 6 checkers on some important point; complex timing issues; very long-term containment; (future) primes that are (at least partly) on opponent's side of the board.

This position qualifies for many of these, so it's not all surprising GnuBG gets this wrong. Since the basics of the position won't change much over the next few rolls, if GnuBG plays it wrong now, it will probably keep playing it wrong for many turns to come, so the rollout result will be wrong too.

It's hard to find out where the neural net goes wrong here, but in my experience, it seems that GnuBG is too worried about the extremely huge lead in the race leading to fantastic "timing" for White. Therefore, it probably 'thinks' getting hit will help. But this is mostly speculating.

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.