Author: Uri Blass
Date: 02:00:54 09/10/01
Go up one level in this thread
On September 10, 2001 at 04:35:08, Gian-Carlo Pascutto wrote: >On September 10, 2001 at 02:31:25, Uri Blass wrote: > >>I suspect that it was productive in some test suites but counter productive in >>games and they simply did not know it. > >How can you know that?? > >Because you _think_ so?? > >Please give some good arguments to back up all these >claims, because so far I have only seen bogus 'I think' >stuff that was trivial to refute. > >For the record, Bruce uses SE and he definetely thinks >they help also in real games. Bruce does not use SE in the way that DB used it. I do not know about one programmer who use SE with the same algorithm of deeper blue. I need to explain how is it possible that deeper blue could not see one impressive move that top programs need a long time to find inspite of it's hardware advantage and if singular extensions that were used were counterproductive in games then it may explain the facts that I see. I guess that giving programs hours to analyze Kasparov-deeper blue games can help programs to see some good moves that deeper blue simply missed in the games I already found that they can see Kh1 that is winning in game 2 and they even need only few minutes to see it and I expect them to find more moves if I continue the experiment. I analyzed for hours only few positions from the match and I hope that if other people help me then we can find some tactics that deeper blue simply missed to get a more convincing evidence(Kh1 is not tactics and the difference is only positional difference from programs point of view). I believe that a lot of tactical things that programs can see in games and not in test suites are not based on singular extensions. 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.