Author: Ryan B.
Date: 15:08:26 10/13/05
Go up one level in this thread
Shredder over pruned and lost this time I guess. I cant test it for sure because I don't have a windows computer or Linux ver of shredder but from when I had access to a windows computer I noticed Shredder would sometimes prune out moves that it should not relating to king safety. This is not an eval issue it is a search tree issue. It is rarely Fruits almost non-existent eval function that wins games, it is its amazing search tree stability and accuracy. If you don't believe me try making movei's eval just like fruits. I think it would be called gutting your eval function though. On October 13, 2005 at 15:44:22, Uri Blass wrote: >On October 13, 2005 at 13:32:58, Ryan B. wrote: > >>On October 13, 2005 at 06:26:05, Uri Blass wrote: >> >>>On October 13, 2005 at 03:02:45, Ryan B. wrote: >>> >>>>I think it is a known issue that Toga, just like Fruit that it is based on, >>>>knows very little about stratrgy and plays off being a great searcher. >>> >>> >>>No >>> >>>I think that evaluation is one of the big advantage of fruit relative to >>>opponents and not search. >>> >>>I have examples when Fruit simply evaluate position better than commercial >>>programs. >>> >>> >>> Becuase >>>>it is an open source program you could add to the eval but it will slow down the >>>>search some for each thing added. In computer chess it is still tactics over >>>>strategy, never mind the so called "positional" chess programs. >>> >>>I disagree. >>>I think that there are lot of positions that fruit understand better than >>>commercial programs(see the case of the exchange sacrifice that I mentioned) >>> >>>Fruit is also not a fast solver of nolot 1 or nolot 2 so I do not think it's >>>relative advantage to the commercial programs is tactics. >>> >>>Uri >> >> >>Fruit understands the exchange sacrifice you mentioned because it has no bad >>trade value. I am sure it could also get burned on not having one. Not that a >>bad trade value is good chess knowledge becuase it is just a lazy assumption >>that is usualty right. > >The exchange sacrifice is not the only case > >look at the last game of the following post > >http://www.talkchess.com/forums/1/message.html?455583 > > >Here is a position from that game > >[D]3r1k2/5p1p/3b2r1/5pp1/Pp1B2n1/1P2PB2/4QP1q/R2R1K2 b - - 0 27 > >Fruit has positive score for black when Shredder evaluates it as better for >white and need time to see positive score for black. > >What is the reason that shredder evaluates it as better for white at small depth >if not inferior knowledge relative to Fruit. > >I think that fruit is simply better evaluator than Shredder and >Fruit2.2 can beat Shredder inspite of being often outsearched by shredder >because of a better evaluation. > >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.