Author: Albert Silver
Date: 10:56:42 11/22/04
Go up one level in this thread
On November 22, 2004 at 13:53:04, Albert Silver wrote:
>On November 22, 2004 at 12:14:42, Peter Skinner wrote:
>
>>On November 21, 2004 at 14:27:33, Albert Silver wrote:
>>
>>>Here you're saying you can identify the difference between the play of 2650 and >2700 based on watching the games.
>>>
>>
>>Yes and no. Watching the Gandalf 6.0 results and games I am able to tell that it
>>is improved in the endgame, but still has problems with possible passed pass
>>candidates.
>>
>>The book is also improved; quite a bit over the previous releases. Dan Wulff has
>>done an excellent job there.
>>
>>So based off what I know from Gandalf, can see what changes/improvements have
>>been made I can make a fairly safe estimate that it is around the 50 elo
>>improvement.
>
>I understand, but improvements and changes are rarely if ever linear. This is
>well-known, but experimenting with the Pro Deo settings has made this all the
>clearer. A small change can yield big improvements in some test positions, and
>slow downs in others. And the best test suite settings are pretty much never the
>best playing settings. The point being that if Gandalf has really chanegd so
>much, you should not expect it to simply be like Gandalf 5 plus some
>improvements. It is bound to be very different period. Still, here are some test
>suite results over at CSS:
>http://f23.parsimony.net/forum50826/messages/112946.htm
>
BTW, just to be clear: I have no opinion whatsoever on Gandalf 6. All I'm saying
is that short of solid testing in well controlled conditions, one can't judge
whether it will be 50 or 100 points stronger.
Albert
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.