Author: Albert Silver
Date: 08:28:00 05/25/03
Go up one level in this thread
On May 25, 2003 at 01:51:53, Uri Blass wrote:
>On May 25, 2003 at 00:55:38, Albert Silver wrote:
>
>>On May 24, 2003 at 11:57:08, Uri Blass wrote:
>>
>>>On May 24, 2003 at 10:01:22, Albert Silver wrote:
>>>
>>>>On May 24, 2003 at 04:34:04, Tony Hedlund wrote:
>>>>
>>>>>[Event "SSDF"]
>>>>>[Site "Tony Hedlund"]
>>>>>[Date "2003.05.24"]
>>>>>[Round "3"]
>>>>>[White "Fritz 8 A1200"]
>>>>>[Black "Shredder 7.04 Classic A1200"]
>>>>>[ECO "E68"]
>>>>>[Result "1/2-1/2"]
>>>>
>>>>Does this mean there will be a separate entry for Shredder 7.04, as opposed to
>>>>the initial 7.0 release? I'm in favor considering the commotion it's increased
>>>>strength has caused, but just curious. Also, will the previous programs the 7.0
>>>>played be using their book learning? If so, I think it's a mistake as this would
>>>>be biased against the 7.04.
>>>>
>>>> Albert
>>>
>>>Do you suggest that shredder classic learning is worse than the learning under
>>>chessbase interface?
>>>
>>>I do not know and I read that the book under chessbase is worse than the classic
>>>book because the book under chessbase allows also inferior lines.
>>>
>>>Uri
>>
>>I was under the impression that only the GUI was different, not the opening
>>books. If the books are different, then my previous comment on books can be
>>ignored. Otherwise a program that had already learned from opening problems
>>against 7.0 would now avoid them against 7.04 if they used the same book.
>>
>> Albert
>
>If I understand correctly the book of the classic interface is more restricted
>to good lines.
>
>I doubt if book learning against shredder7 is going to help much because
>shredder7.04 is different program than shredder7 and what is good against 7 may
>be bad against 7.04
Maybe, maybe not. A lot depends on how the book learning is done, and how
different the evaluations between the two engines are.
If the learning is done via results, then the difference between the programs
will increase the changes in the results book learning would produce, but that
makes little sense to me. A game could be lost via weak playing having nothing
to do with a bad book after all.
If the learning is done via the evaluation it perceives after leaving the book,
an evalaution that is persistent over several moves, then it's another story. An
engine shouldn't simply discard a line because it left the book thinking it is
worse, since its evalaution could be wrong. A few moves later, by playing moves
this might change. If it doesn't then it should be excluded even if the line
isn't bad objectively since if over several moves the games continues poorly,
the engine at the very least has shown itself unable to play the positions
properly.
You could be right stating that book learning with 7.0 won't benefit against
7.04, but significant holes in books have certainly been found, and if a big
hole was found in Fritz 8's book (for example) in a game against Shredder 7.0,
and it no longer repeats this against 7.04, then it is certainly benefitting
from its book learning.
Albert
>
>Uri
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.