Author: Vincent Diepeveen
Date: 06:16:56 05/10/04
Go up one level in this thread
On May 10, 2004 at 07:49:18, martin fierz wrote: Let's quote Seirawan here : 13 ... g5? A terrible positional concession. Deep blue thrusts its protecting pawn shield away from the body politic. Now black's king will lack protection in the upcoming middlegame. what prompted this mistake? Again, deep blue realized that after white threatened 14.g4,bg6 15.nxg6 fg, black would be forced to part a bishop for a knight and this time, its pawns would be doubled. This cure, however, is worse than the disease! After 13..Be5 14.Bxe5 (14.d4!?) Nxe5 15.g4,bg6 f4, ned7 17. nxg6 fxg6, white has earned an advantage. After the text, a positional player like myself would consider black's position to be strategtically lost. Another important insight into how computers play is revealed by this move. Computers are programmed with a simple rule of thumb, it is good to push pawns. deep blue cannot calculate any negative consequences for this move on its horizon and it therefore plays according to its program pushing a pawn with a tempo to boot! In later games we will also see that deep blue doesn't like doubled pawns on g7,g6 nor at g2,g3 despite that this is sometimes good. The programmers simply have forgotten to put this simple exception into the deep blue knowledge. >On May 08, 2004 at 18:55:16, Djordje Vidanovic wrote: > >>On May 08, 2004 at 12:14:42, Anthony Cozzie wrote: >> >>>On May 08, 2004 at 11:51:12, Robert Hyatt wrote: >>> >>>>On May 08, 2004 at 10:50:57, Anthony Cozzie wrote: >>>> >>>>>On May 08, 2004 at 07:18:27, Vasik Rajlich wrote: >>>>> >>>>>>On May 08, 2004 at 04:34:40, Sune Fischer wrote: >>>>>> >>>>>>> >>>>>>>>>You are absulutely right. >>>>>>>>>It is obvious that humans already solved chess so they know if a move is a >>>>>>>>>blunder or not a blunder so you can be sure that all the question marks are >>>>>>>>>correct. >>>>>>>>> >>>>>>>>>It is also obvious that the number of mistakes is what decides the game so if >>>>>>>>>your opponent did 2 mistakes you can let yourself to do one mistake like letting >>>>>>>>>him to force mate and you are not going to lose. >>>>>>>>> >>>>>>>>>:_( >>>>>>>>> >>>>>>>>>Uri >>>>>>>> >>>>>>>>You know, Uri, I have never seen you do anything but post how other people are >>>>>>>>wrong (never with any reasons of course). Many other people have noticed your >>>>>>>>unending flood of negativity. It is difficult to consider this post as anything >>>>>>>>other than a flame. It appears I am going to have to take off the kid gloves >>>>>>>>and dispose of you. >>>>>>> >>>>>>>Isn't it natural to only post if you disagree? >>>>>>> >>>>>>>Anyway, I suspect Uri has a point. >>>>>>>It's not unusual for computers to play "unatural" moves, just think of the >>>>>>>Hedgehog Junior played against Kasparov. >>>>>>> >>>>>>>All the time the GM's were saying how strange Junior's moves were, how "it >>>>>>>showed no understanding of the position" blah blah blah. >>>>>>> >>>>>>>So please explain why Kasparov suddenly had to fight for a draw after 10 >>>>>>>questionmark moves from Junior! >>>>>>> >>>>>>>-S. >>>>>> >>>>>>I never thought this day would come - but I agree with Uri here. :-) >>>>>> >>>>>>Sports aren't about beautiful play. Sports are about winning. If someone is >>>>>>playing ugly, and winning, then it's your sense of aesthetics which needs to be >>>>>>reviewed. >>>>>> >>>>>>Computers have a long history of winning ugly. In the recent Fritz-Kasparov and >>>>>>Junior-Kasparov matches, the machines made many many more "mistakes" (according >>>>>>to human opinion) than Kasparov. But - if these mistakes aren't punished - are >>>>>>they really mistakes? Is it a mistake to leave Shaq wide open for three point >>>>>>shots? (Or send him to the line for "free" throws?) It's impossible to speak >>>>>>about objectivity here. You can only look at the results. >>>>>> >>>>>>Vas >>>>> >>>>>Let's take a look at some of the moves the annotator didn't like: >>>>> >>>>>[D]r2q1rk1/pp1n1ppp/2pbpn2/3p3b/8/1P1PPNPP/PBPN1PB1/R2Q1RK1 b - - 0 10 >>>>> >>>>>Zappa plays the obvious 10 ...e5. Deep Blue played 10 ...h6. I won't call this >>>>>a bad move, but it's clearly a pass move. >>>> >>>>That isn't very convincing. Did you look at _your_ PV? move 4? :) >>>> >>>>Order doesn't mean much to alpha/beta as it scores positions, not moves as they >>>>are played. >>>> >>>>First impression is that h6 and e5 transpose to the _same_ position... >>>> >>>> >>> >>>I will accept that my 4 ply search plays pass moves some of the time :) Zappa >>>uses pure R=3 now, and perhaps the evaluation isn't quite good enough for it. >>> >>>anthony >> >> >> >>In my opinion 10...h6 is not merely a pass move or waste of time. White may >>plan to push the g-pawn to g4 and drive the black bishop to g6 aiming to >>exchange his knight for the bishop later, playing the knight to h4. Thus >>10...h6 gives black refuge. Besides, the bishop positioned at h7 would be very >>useful later on, perhaps after the push you mentioned that Zappa plays right >>away, exerting pressure on the e4 square... >> >>Just an idea. One glance at the diagram... Caveat emptor :-) > >hi vas, > >here's why ...h6 is wrong: just look what happens in the game: after Qe1 and >Nh4, white is threatening g3-g4 and after ...Bg6 Nxg6 fxg6 it's clear that black >has gone wrong. without ...h6 this just never happens, you just go back to g6 >and after Nxg6 recapture ...hxg6 with a solid position. that's why deep blue >went ...g5 later, which it had to do, else the abovementioned line happens. so >...g5 is not a blunder IMO, but before: Bc7 and Qa5 are ridiculous moves, doing >nothing at all. it could have played Bg6 before Qe1+Nh4, for example. or gone >for one of the standard plans with ....a5-a4 or e7-e5. all makes much more sense >than what was played. >i guess playing ...h6 and ...Bg6 is just a little slow compared to doing >something active, so i would also call ....h6 a mistake, but i'd call Bc7 and >Qa5 to be the real culprits here. > >cheers > martin
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.