Author: Vincent Diepeveen
Date: 05:18:47 05/14/98
Go up one level in this thread
On May 13, 1998 at 11:16:01, Ed Schröder wrote: >Posted by Vincent Diepeveen on May 13, 1998 at 06:12:00: > >In Reply to: Re: Deep Blue--Part III posted by Dave Gomboc on May 12, >1998 >at 21:31:19: > >>KK kup was fake. Diep got black in the only 2 important games. >>Just 2 games. > >>And Diep got outbooked by both programs. >>Game against Rebel it was directly lost. Diep's horrible in French. >>Just look at ICC the % of wins of Diep in French. > >Then why play French? > >Because Diep plays the French bad the KK-KUP-I was a fake? > >As far as I remember you were the one claiming to have the best >analysis program in the world and exactly THIS was the reason >that KK organized the first KK cup. > > >>2500 SSDF says nothing. > >Is Diep better then? > > >>Don't you get it, all programs sent to Sweden have played say 10000 >>auto232 games or something. > >>Sweden became one big book tuning event. > >Partly true. > > >>In the future this might change if books get more variaty. > >This for most programs is already true. > > >>Until then this Sweden list is laughable. > >Your opinion, not mine. > > >>How can a program like Fritz5 which loses normally to all those programs >>can get on top? > >Any proof of that? Yep, hand played games against F5. >>How can Rebel9 get stronger than 8, considering that the only >>4 things changed are interface,few extensions to solve some tricks, >>BOOKLEARNING, and with capital letters BOOK. > >Any proof of that? Book different, and booklearning. I received from Jan games where Rebel repeated game after game, so we know that your booklearning is aggressive too. Exactly same score at same depth. Same evaluation at same depth. Faster finding tricks sometimes, needing more nodes for every ply. So engine mainly the same, selectivity either less, or more extensions. Probably that last, you're the one knowing it, your explanation? If i change a lot to my evaluation then most positions get different evaluation. Main conclusion: evaluation the same, but the 'details' way better. So score in Sweden better, no doubt. >>The program didn't play better, yet it got a higher elorating because it >>scored better in matches. > >>That kind of comparision is wrong. It won't play better against humans >>than Rebel8 nor will it play better in tournaments. > >Any proof of that? > > >>A simple 1.d4,d5 2.Bf4 was enough for Diep in aufsess tournament >>to beat it. > >>No doubt that when i send my program to Sweden that i will >>have played some hundreds ot auto232 games too (i'm only >>having a Pentium pro 200 and a laptop 133, so i can't produce >>thousands of games like the pro's). > >Like to help you. Just send Diep and I will test it for you on >40/2:00. Diep is auto232 compatible or? > >- Ed - > > >>Your computerchess insight is laughable. > >>A program just playing 6 games in its life you call God. > >>Where i need to get a rating in Sweden based on cooked lines >>after hundreds of games, or thousands of auto232 games, before >>you believe me? DID EVERYONE MISS IMPORTANT CONCLUSION OF MY MSG? IS EVERYONE BELIEVING DB JUST FOR 6 GAMES, WITH BUNCHES OF BAD MOVES, WHERE I ACCORDING TO THE ONE I REPLIED ON NEED AN ELORATING IN SWEDEN BEFORE HE BELIEVES ME? Of course i understand very well that Ed replied to what i wrote, i didn't expect different. In the past i repeatedly proposed Ed and Jeroen (who asks always for a version of Diep) to exchange version of Diep for version of Rebel. They didn't want this to happen. Greetings, Vincent >>>Dave Gomboc
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.