Author: Ed Schröder
Date: 22:32:24 03/21/01
Go up one level in this thread
On March 21, 2001 at 11:50:55, Vincent Diepeveen wrote: Although I am not in agreement with every statement you made this is a cool posting Vincent! Ed >On March 21, 2001 at 09:53:07, Jorge Pichard wrote: > >Well first of all, today programs all search way deeper as deep blue, >so i don't see the deep blue issue, but regarding the subject >question: > >the problem is picking the move from a combination of programs. >If we call the picking the move person or entity or program a 'hirn' >(from 3-hirn, which means translated from german to US 'brain') then >the hirn is the weakest link. > >Suppose the hirn is 2200 rated in selecting moves, not to confuse with the >hirns own rating. > >Supposing that the programs play 2600 fide rated, then obviously there >is a major problem as 2200 < 2600. So the level will be UNDER the level >of the programs. > >In general most hirns will give weaker play, especially if you replace the >hirn by a simple program with a few rules. > >Suppose we have 2 programs from equal strength, but different >programs. > >And suppose the rule inside the hirn is next: > > 'play the move of the program with the highest score' > >MOST LIKELY THAT WILL PLAY NOT BETTER! > >>Since Deep Blue had a tremendous calculating power, I was wondering if a >>combination of the best current programs available such as: Deep Fritz, Deep >>Shredder, Deep Junior and Gambit Tiger II with all the Deeps using a minimum of >>a PIII 2X 1,0 GHz and and G-Tiger II using a minimum of a P III 1,2 GHz could >>have a better chance to beat Kramnik than using a single program. Of course > >Now let's pick things from a different spot: "can we beat anyhow kramnik". > >Whatever combination you're not going to beat kramnik anyway, unless >he's playing the match in the same kind of suicide-chess way as kasparov >did against Deep Blue. Without commenting too loud on Deep Blue, as we >must compare with 1997 programs and not with the high level standard >of today, it's obvious that the big guilty person in this match was >Kasparov. He CAN be blamed for losing the match by pathetic moves. > >But what is best way to show based upon what happened so far in >computer-human why kramnik would be an ideal player to face computers? > >Best is probably to look at the 2 matches from 1997 and 2001, where >kasparov played deep blue and where v/d wiel played rebel. The only >slow level matches from the last 5 years actually... > >If kasparov would have shown the same kind of games like v/d wiel recently >showed against rebel, then things would have been different. > >v/d wiel had 3 disadvantages against the computer which kasparov nor kramnik >never will have and never had: > a) v/d wiel was badly out of shape the weeks before the match he > had lost from dozens of 2200 players. his rating also dropped > a lot. > b) the tactical weakness. When kasparov has a big dip he'll lose perhaps > from Van Wely (2700) a game, but not based upon tactics. > However if v/d Wiel is badly out of shape tactically spoken, > then he blunders away piece after piece. Of course not the first 20 > moves. The big suffering for the human player always happens after > some time has passed, like 3.5 hours and more... > c) v/d wiel actually will have problems playing his default openings > against the computer. His default opening is caro-kann with black. > Now i'm also a caro-kann player and i can assure everyone that with > caro-kann i lose in blitz nearly everything against the computer. > On the other hand with French, which i never play in my life, i have > a very good score against the computer with black. Even in blitz i > sometimes manage to win it, which is quite incredible as i am not aware > of any standard tricks in french. Kasparov on the other hand has > a splendig openings repertoire to play against the computer. I mean > najdorf is completely deadly for the computer. It will be lost bigtime > long before it realizes it and it most likely won't sacrafice material > for complicated positions which is nearly always needed to win > najdorf in an objective way for white. If white never sacrafices in > a najdorf game, then usually black is either a complete beginner or > black was the one who sacraficed himself to mate white... > >There are just a few factors to consider where v/d wiel has the edge over >kasparov, but in no means this is a big plus for v/d wiel > a) not paranoia/crazy. v/d wiel is in no way crazy or paranoia, whereas > kasparov definitely is. His stupid accusations against the Deep Blue > team are already enough proof of this. Later he didn't even correct > this. As a world champ it is NOT very polite to accuse the opponent > from something just to DISTRACT the audience from what a bad games he > showed. The games are really around 2400 level at most, from tactical > viewpoint. From strategical viewpoint kasparov played like 1900 at most. > Now 99.99% of the world population for sure doesn't have 1900 so won't > notice that... > b) kasparov won the first game. Actually kasparov after the match said: > "i shouldn't have won the first game that easy". This is very true, > when talking about the PERSON kasparov. > Deep Blue does about everything wrong with pawn structure what one > can do wrong, even 1997 programs understood that with perhaps exception > of the blunder g7-g5 which most progs still play. In fact by playing > so deadly passive (just 3 rows) any 2600 player would have completely > humiliated kasparov if the 2600 player would have had black. If deep > blue would have been a box with a 2600 GM inside who doesn't fear > tactics, then kasparov would have lost that first game for sure. > Speculations are always dubious to do, but very sure kasparov offered > deep blue the first point for free. Perhaps in an attempt to attract > more spectators? > So kasparov simply played a bit suicidal in order to get more audience > seemingly. Resigning in a drawn position is only confirming this. In > real matches you play till the last pawn. So then you play qe3 and > you either find Re8 then and draw the game, or you don't find Re8 > and you lose the game. Kasparov already resigned long before that > stage... ...definitely v/d wiel is not so stupid. He is a lot smarter > as kasparov. Basically because of his low rating (low for GM): > list april 2000: Wiel J.T.H. v/d M 2551 17* > list september 2000: Wiel J.T.H. v/d M 2483 47* > Someone from 2483 (dropped 70 points within 47 games) > risks of course to not get invited anymore if he loses by a big margin. > Kasparov doesn't care. He has 28xx anyway so he cares shit in > this respect. > c) Now playing 3 rows hoping your opponent to rape its own pawn > structure is a very well known way to attack a computer, most likely > kasparov didn't know this as he didn't try any other trick which was > shown in Aegon tournaments, but in aegon tournament it was very common > for white players to open with a near to 3 row system: d4 c3 e3 and then > wait till computer kills itself and then strike. V/d Wiel clearly > based his games against Rebel upon experience from the Aegon tournaments, > whereas Kasparov obviously played without caring what to play against > the computer. Even worse, kasparov played only systems from Karpov > versus it. > d) All those systems share that they just shuffle with pieces. Now even > Seirawan in ICCA journal june 1997 says several times: "perhaps kasparov > is not used to this kind of play and therefore plays this horrible > strategic mistake". Tactical blunders are of course usually decisive in > games. Above 2400 we talk usually about a level where those horrible > moves don't get played. A few bad moves usually don't hurt a game too > much, but a move that's taking the wrong strategical decision is > completely going to lose games for you! > V/d Wiel didn't make any obvious strategical mistake against Rebel in > the first few games. > Kasparov made plenty against Deep Blue. Kasparov made actually only > 1 game several tactical mistakes and that was in game 6. > e) Kasparov technique is kind of bad when looking at his 28xx rating. > Most 2500+ gm's are technical better in endgames as Kasparov. Without > claiming a world title yet myself, i'm pretty sure that if i would make > after many hours of play a bit less tactical blunders (sometimes i > make one), then i would probably be a big problem for kasparov in > technical endgames. Now v/d Wiel has dropped in rating, but i'm pretty > sure he's technical better as kasparov. However, also v/d Wiel suffers > from the same tactical blackouts as i do. > Playing for many hours without blundering away material is something > which not too many people on this planet can do. for > 2200 players > they CAN play a game without blunders. Above 2600 sure tactics is > not the problem. Actually one of the greatest tacticians in the world > is a female! She's called Judith Polgar! > >In short we have a few DISADVANTAGES of v/d Wiel which Kasparov doesn't >have nor Kramnik has. We have a few ADVANTAGES of v/d Wiel which kasparov >doesn't have, most of them psychological. > >If i find kasparov so bad to face a computer, what made kasparov for >so long the #1 of the world and what are still his strong points that >let his rating raise and raise? > >Ok this is easy: > a) if you come to the board to kasparov, you feel the eyes of kasparov > in your body as if you play against someone who thinks you are a > complete beginner. Fearing complete humiliation you sit down and > already lose > the game psychological against kasparov. Kasparov even very intimidating > always puts down the his wristwatch besides the board, to measure what > time it takes to humiliate you. > b) openings preparation. the first 15 moves or so Kasparov plays very quick. > Now and then he takes half a minute to remember an openings line. Very > soon after you leave your home preparation you discover after some > investment in time that you have a lost position here. Actually you > are the second one to discover that, as kasparov during all that > time investment already obviously knew you were lost as he was smiling > like a happy collie-dog. > c) killerinstinct. At the moment you feared that a certain move would > cause you much troubles, either objective or subjective, then kasparov > already plays that move. Sometimes within a second... ...everyone remembers > kasparov quick knock-out of karpov in the famous live television broadcast. > This is how kasparov always strikes. quick, mercilous and like a fanatic. > d) no perpetuals. As soon as kasparov thinks he has a won position he > puts back on his watch to his wrist. Now that's a clear sign for beginners > of around 2600 who play him to resign AT THAT MOMENT. > >None of the above advantages were used against deep blue! > >Even if kramnik doesn't show his openings preparation he'll kick the hell >out of the computer as > a) Kramnik is technical number 1 of the world, great chess heroes like > Capablanca were best in their days but would get beaten > simultaneously blindfolded in endgames when facing Kramnik. > b) Kramnik is by a large margin the best ever in positional shuffling, > even complete useless demonstration games he wins by a few shuffle > moves. Very well i remember a 1 hour a game exhibition on german > broadcast versus Judith Polgar. 1.Nf3 a few shuffle moves and it was > 'game over'. > c) Kramnik is strategical best in world by a large margin > d) Kramnik is very good in tactics. Note that many 'shufflers' are > tactical not bad at all. This applies to Karpov too. > >So in short at 40 in 2 level Kramnik is the ultimate nightmare against >computers. > >At blitz i don't know whether the advantages of kramnik are going to >work for him. In blitz general shuffling works against you. Especially >the tactical details of shuffling need to get calculated very accurate >as without detailed tactical planning having little space means you >get beaten quickly by a few combinations! Even if you see the combinations >nothing is going to stop you from losing then! > >But we already know that humans play worse in blitz as they play at 40 in 2, >so it would be cool to see a 40 in 2 match, but a 40 in 2 match >against kramnik objectively seen is completely suicide for the computer, >*whatever* computer now or in the future, with exception of course of a >10^40 database. > >The only one that can beat kramnik in a match against a computer is >the same person as who beated kasparov. And that was the person himself. >Only kramnik can beat kramnik. > >If he thinks he is just playing a few suicide-chess games for money, >without realizing he needs to play a few GOOD games, then he might lose >if he blunders away pieces. > >My own expectation would be that kramnik plays on safe. He's going to win >the match. After he has nearly won the match or has enough points to win >the match he might not want to humiliate his opponent to much (that's like >humiliating a big sponsor, basically also a factor where kasparov >had to deal with) otherwise he'll lose the pretty income for next year >perhaps... > >>there would have to be a human monitoring the most selected move to be the >>chosen move against Kramnik. By the end of the GTiger II vs Web, we should have >>a good idea on much advantage if there is any of using several programs versus >>using a single program. Please take a look at this interesting position >>where the Web with black pieces take the initiative of the attack. > >The average rating on the world is 1500, so a web game versus a program >is going to play 1500 rated moves theoretically spoken. Especially >if you play based upon 'voting'. > >That's in no way a compare for >kramnik versus a program. The > 1500 player always wins of course >a web game, in this case tiger. > >The only exception being here that if a real strong player gives >advice on which move to play, then the level of the game gets near to >the level of the strong player. Of course many <1500 players still will >vote for a bad move. So sometimes a horrible moves is getting made despite >perhaps superb advice, see kasparov versus world when a strong playing girl >was simply not allowed to post here advice on the website, which at the same >day caused the 'world' to play a completely idiotic blunder. Not even >close to common sense even of an average national player... > >>http://www.rebel.nl/gt2-web.htm
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.