[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: SV: [Bug-gnubg] Holding games

From: Joseph Heled
Subject: Re: SV: [Bug-gnubg] Holding games
Date: Fri, 20 Jun 2003 21:49:34 +1200
User-agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.2) Gecko/20021202

Henrik Ravn wrote:
Hash: SHA1

Joseph Heled wrote:
- ----
- - No doubt gnuBG 2ply gets this wrong. Not for the first time nor for the last either.

- - We all want GNU to be correct every time on each position, however this is not realistic.

- - In order to make a case we need more than one position. Probably in the order of hundreds or thousands.

- - Even with the above, there is no guarantee we can do something about it. ( In the sense of improving it and not degrading the rest).
- ----

Which raises the question of how to decide whether a new set of weights are actually better than the previous set. How is this done now? Specifically, how can we tell if 14 is better than 13?
be well
- -h-

Two ways. One is my benchmark, obtained using GNU rollouts. Second is to play many matches at 0 ply. The results were given with the email annoncing the new net.


reply via email to

[Prev in Thread] Current Thread [Next in Thread]