Author: Robert Hyatt
Date: 18:32:23 07/01/99
Go up one level in this thread
On July 01, 1999 at 17:22:19, Peter McKenzie wrote: >On July 01, 1999 at 15:44:40, Robert Hyatt wrote: > >>On July 01, 1999 at 14:34:26, blass uri wrote: >> >>> >>>On July 01, 1999 at 14:25:12, Robert Hyatt wrote: >>> >>>>On July 01, 1999 at 12:18:30, blass uri wrote: >>>> >>>>> >>>>>On July 01, 1999 at 11:55:46, Bruce Moreland wrote: >>>>> >>>>>> >>>>>>On July 01, 1999 at 10:41:28, blass uri wrote: >>>>>> >>>>>>>I am interested to know results between top ICC programs at standard time >>>>>>>control(if possible 2 hours/40) from the last 2 monthes. >>>>>>> >>>>>>>I guess that the top ICC programs include Ferret,Shredder,Ban,Crafty. >>>>>>>Is it possible to get the games between these programs in ICC? >>>>>>> >>>>>>>Which hardware do they use in ICC? >>>>>>> >>>>>>>Uri >>>>>> >>>>>>There is no such thing as 40 moves in 2 hours on ICC, and you will find very >>>>>>very few games at time controls approximating that. >>>>>> >>>>>>bruce >>>>> >>>>> >>>>>I assumed that there is because otherwise I do not understand what is the reason >>>>>that Bob hyatt said that he is sure that there is at least one program that is >>>>>better than shredder and that he assumed that shredder is a good program but not >>>>>the best. >>>>> >>>>>We have no data to assume that shredder was lucky. >>>>> >>>>>It is also possible that shredder prepared better for the tournament time >>>>>control games. >>>>> >>>>>Uri >>>> >>>> >>>>Quite simple... I have watched shredder and Ferret play for many months. They >>>>are probably in the same ballpark quality-wise. Except that Ferret is 3x faster >>>>with the parallel search. In a match, I'd personally pick ferret, and the >>>>longer the match, the more I'd be willing to bet, because 3x speed advantage >>>>is way-non-trivial... >>>> >>>>That was my point. Not that Shredder is 'bad' at all. Just that Ferret is just >>>>as good on equal hardware (IMHO) and with 3x the hardware it is _very difficult_ >>>>to handle. >>> >>>I do not agree that you can know from watching program playing at fast time >>>control what will happen at tournament time control. >>> >>>The only way to know is to play games at tournament time control. >>> >>>I read that the tester of shredder said about the results against commercial >>>programs that longer time control help shredder. >>> >>>Uri >>> >>>Uri >> >> >>Time doesn't matter _at all_ when you look at the right stuff. If you only >>look at 1-0 or 0-1, then longer games tell you more about how it plays at >>longer time controls. >> >>But if you look at the _moves_ you can pick out 'quality' or the lack thereof. >>You can tell if a program has any understanding of passed pawns, king safety, >>pawn structure, center control, mobility, etc. Even without knowing _how_ the > >I think this is true to a certain extent, but things can get quite deceptive if >a program is being out searched (or it doesn't have the right extensions to deal >with the tactics in a given position). In this case, the program that is being >out searched can appear to lack positional understanding because it makes moves >that lead to a weakening of its position. In this case, the bad positional move >can be caused by EITHER poor positional understanding OR lack of search depth. > >If you can see a programs score during the game, then it becomes easier to judge >the positional understanding of a program. The things I look for are really 'positional' issues. IE playing the e6 Sicilian and then later playing e5 leaving d6 helpless. If you understand weak pawns, you usually don't do that. But you _do_ have to look at the game carefully, of course, as sometimes e5 is forced to ward off even greater positional or tactical consequences. Or white playing Nxc6 against a black Sicilian. Or white (or black) playing f4/g4/h4 with no hope of attacking and with the opponent having much greater attacking chances as a result. IE moves that don't tactically lose, but which give away positional ground. > >>game ended. That is what I look at generally. Knowing that longer time >>controls will repair many of the tactical mistakes, but _none_ of the positional >>mistakes... > >I disagree with this. Tactical and positional play are finely interwoven, >subtle tactical errors can cause positional mistakes. > >Peter You have to study the games carefully. This isn't a 30 second glance and go like Kasparov does. But it is possible to take _any_ program and critique another program's game, using it to spot tactical problems, and using your head to analyze positional considerations. This is an essential requirement to play with GM players. If you can't figure out what you are doing wrong, you have a hole they will drive a truck through. Over and over... And _every_ loss should tell you a story. And many wins have a story to tell if you look carefully. IE Crafty played Nunn several games last night and had a great streak going. The first game was a draw, then 8 wins in a row. Then Nunn worked up an attack Crafty failed to notice, and the score got to -2 as Crafty worked to neutralize things and exchange pieces (-2 to crafty might be an exchange down, a pawn down, or even equal material, depending on how serious the attack looks). It came back by playing an outstanding endgame and won in spite of being dead at -2. But I studied the -2 part very carefully, because it made a serious mistake. And I found out why and fixed it. So that 'win' was still useful. BTW, it has put together some incredible stats over the last 2 days... it played Yasser 21 games and had 4 draws and the rest wins. It played Kamsky 14 and drew 2 (rest wins) and then Nunn with 1 draw in 11 games. I personally think that is impossible to continue, but it was amazing to watch it happen. (there were a few other games mixed in, 'GM Zarnicki (spelling?)' and others would try a game or two but get pissed and leave. :)
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.