Author: Vincent Diepeveen
Date: 13:51:24 01/12/03
Go up one level in this thread
On January 11, 2003 at 18:33:44, Robert Hyatt wrote: >On January 11, 2003 at 14:38:33, Vincent Diepeveen wrote: > >>On January 11, 2003 at 09:39:04, K. Burcham wrote: >> >>In 1997 i won easily in blitz from all programs. They knew nothing. >>Bad evaluation. Very passive play. Small search depths. > >That's a crock. I can easily resurrect a 1997 version of Crafty, and can >find some 1997 hardware. In 1997 Crafty was whacking GM players with ease >at 5 0 type blitz games. play that version for money 5 3 against them at 1997 hardware of yours. which was a pentiumpro 200Mhz (single cpu). You'll lose money. Of course hand operate it. Human are way better on chessboards than at a 2d generated small board on a computer. >Whenever you want to show your chess skill let me know and I can set up a >demo on ICC for you. I'd have to think about 1997 hardware, but I ran on >a 500mhz alpha in 1996, you ran on a pro200 at the internet. >so that should give some indication of what the >hardware should be. I have a couple of 750mhz machines which are probably >in the right speed-range for 97. the thing back then got like 6 to 7 ply in blitz. with dubious pruning 8. some mercilous a look like attacks already killed it. You censored everyone who did this or put them at a S list back then (if not invented later). the 97 crafty sucked ass in endgame everywhere. I face now in blitz things that go 10 ply deep. A player of today facing that now will need a game or 20 to adjust to the fact that the thing is missing all kind of simple passer endgames. *any* material equal rook endgame wins against it. After finding that out. then the day after that he will beat it silly. I probably have a person to show you too. Some team member of mine. >>Cars on the other hand advanced hardly from 1997 to 2002. The advances of cars >>from 1997 - 2002 is in no way comparable to the advances in all respects of >>computerchess 1997 to 2002. >> >>We talk about the weakest chains getting stronger by nearly 500 rating points >>*at least*. > > >500? ? ? ? ? Sure your endgame in 97 was *very* weak. If something is real weak at a program you seem to forget how much things improved by some tuning there and especially a bigger search depth. the weakest chain in blitz is tactical the same depth a GM gets in a normal game. So at blitz he won't get through tactical barrier. the 97 thing was insane weak tactical. 7 plies in blitz. The easy test for you to start with is to play your current 4 thread 2.8Ghz Xeon versus a single cpu Pro200Mhz which you had in 1997. No you didn't have a 500Mhz alpha at the internet in 1997. Not at all. See how crafty 10/11 plies smacks crafty 7 plies. It is not even *near* to the tactical barrier that 7 plies. that will already be like 500 rating points. proving my point clearly. >>Let's take openingsbook. the 1997 openingsbook of deep blue was a random book >>with 4000 hand tuned moves. >> >>4000 hand tuned moves is very little. >> >>When compared to todays openings books that means it will get out of book on >>average with -1.0 down in score, if not more. >> > > > >that is a crock. I have a big opening book, and _I_ won't get out of book >"on average with -1.0 if not more"... > > > >>So junior - deepblue i estimate at around 20-0. > >Just like you estimate what can be done with functional languages? With >NUMA hardware? With SMP locks? > >:) > > > >> >>12 points from book. 1 point from middlegame. 5 points from attacking the king >>and mating deep blue (and it not showing any positional problem until it is too >>late. then when it smells it, it sees probably more than junior there despite >>junior searching deeper; singular extensions have that habit) and 4 points in >>endgame. >> >>Best regards, >>Vincent >> > >If you could win games with your mouth, you might have a serious chance. >But the mouth doesn't play the game, unfortunately... > > > > >>> >>> >>>I do not understand the comparision anyway. We do not compare a 1910 car with a >>>2002 car either. >>> >>>You compare deep blue. no nullmove, no good eval (for 1997 standards sufficient >>>though) with the formula 1 cars that we build today. >>> >>> >>>Best Regards, >>>Vincent >>> >>> >>>Vincent, lets use your example since you want to compare Deep Blue to the >>>automobile, and the formula 1 cars. >>>Lets go to the salt flats with a 1997 car, (more realistic time reference), that >>>is capable of a top speed of 2000 miles an hour. Also take a car from 2003, and >>>this top speed is 200 miles per hour. the 2003 car has more knowledge about the >>>goals wanted by the humans, get to the finish line first. the 2003 car is more >>>stable, rides better, better interface, looks better, etc. >>>i think i will put my money on the speed for now. >>> >>>kburcham
This page took 0.01 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.