Author: Vincent Diepeveen
Date: 19:12:39 09/08/01
Go up one level in this thread
On September 08, 2001 at 14:07:22, Gian-Carlo Pascutto wrote: >On September 08, 2001 at 13:50:01, Vincent Diepeveen wrote: > >>Secondly we have a lot of data about DT's speed. And it for example >>didn't find h3 which is in the LCTII testset against Karpov. >> >>So it lost from karpov. >> >>You must add the both positions. >> >>The h3 trick against karpov is just a 12 ply thing and they didn't >>even find that one. > >Yes, but which DT was it? This game was played after the Karpov >game if my dates are correct. > >The DT/DB guys made quite impressive speed improvements coming >from 50knps with Chiptest to 200M with DB. chiptest was 500k nps AFAIK. chiptest == DT1 note that cray blitz wasn't so impressive when talking about speed. A P3 from nowadays is faster than that old machine, but of course when getting only a few ply you get toasted against something searching deeper. Both hardly used nullmove. DT never, chiptest with R=1 (Bob?) if i remember bobs description well. So that was a simple walk over, especially considering how bad evaluations were tuned & tested in these days. >I don't care it didn't find the h3 trick. Perhaps their endgame >eval was lacking there. Perhaps they just had bad luck. Perhaps >they had a bug. It's a tactical 12 ply, 13 if you don't do checks in qsearch perhaps. >The question was a post where DB saw something the current >programs can't see. This is one. No this isn't one. This c5 is dead old. And all we know is the memory from Robert saying they expected a move after which it was +2.xx for them. No hard proof. I see sometimes my own program expect a bad move for the opponent which is then also +2.xx. c5 is not a proof of that. Interesting is to see the games played against kasparov. In the mainlines printed by DB you see all kind of tactical horizon things which you can analyze at home. REgrettably there are so little persons here checking out th elog files, that i can imagine Uri's dissappointment there very well. yes i did some big analysis of the log files with several programs, and yes i did checkout many things and i am always amazed about strongplayers how much they know from all games kasparov plays against Kramnik. They even know exactly the move number where kasparov made a mistake and they still can put up the endgame in several berlin defenses, some also know exactly the games from kasparov-karpov from quite some years back. Etcetera, but they never know a single move from the games played between kasparov-deep blue. Not to mentionned that they were hardly analyzed. If i ask to a GM why, he usually says: "never checked them out", and some others (who also write in newspapers so they HAD to check them out): "these games were so bad that they weren't worth analysing them". Considering the impact of the match i don't think that true. But it's the choice they all made! >-- >GCP
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.