Author: Robert Hyatt
Date: 20:38:03 06/29/99
Go up one level in this thread
On June 29, 1999 at 14:45:24, blass uri wrote: > >On June 29, 1999 at 09:49:39, Robert Hyatt wrote: > >>On June 29, 1999 at 01:28:28, Jouni Uski wrote: >> >>>These superfast programs only got 3 resp 3.5 points! Specially P.Conners was >>>one of the favourites before WCCC. I hope someone can give explanations. >>> >>>regards Jouni >> >> >>I think it was an issue of 'robustness'. >> >>I have said many times that the most fortunate thing that ever happened to me >>was winning the 1983 world championship, because our parallel search was only >>completed two weeks prior to the event. And at round-one time, we had not >>played a complete game, although we had done plenty of testing, and we did a >>very simple parallel search for that reason. >> >>This was one of the reasons I expected both Ferret and Shredder to do well, >>because _both_ play lots of games on ICC. > >I do not remember that you expected shredder to do well. >I remember that you did not expect a micro to win WCCC. > 1. I have _always_ said shredder is strong. I had several arguments about ranking commercial programs back in the r.g.c.c days, because I always ranked shredder high after Jakarta. I never changed my opinion. 2. I said "a micro doesn't stand a chance in hell of winning a tournament with all the 'big iron' programs in it. I included in that list of 'big iron' programs the likes of deep blue, Cray Blitz, Cilk, and so forth. Two of the strongest big iron programs weren't there. Shredder did well. But based purely on statistics I would not bet on it if we had the same tournament with the same programs. Because I can think of at least one program that I am certain is better that played in Paderborn. But Shredder certainly deserved the trophy and praise for winning because it is _not_ easy. If you want to write the 'big iron' guys off, feel free to do so. But _never_ turn your back on them... they are _extremely_ dangerous. To any micro program in existance... > > And the first step toward becoming >>a world champion is to play a _lot_ of games and have a bug-free program. The >>"research" guys have always hurt themselves here (me included) because access >>to the 'real hardware' is often quite limited. Which means you do lots of >>testing on 'baby hardware'. But an additional ply or two changes things a >>lot, and without testing, bad moves are not uncommon. > >The research guys could play slow time control games agianst commercial programs >to get these additional plies. That is an awful way to test. If you only goal is to win the WCCC, it might be ok. If you want to beat GM players it most certainly is _not_ the way to tune. What commercial program can carry out any sort of decent kingside attack? None I have seen. So give me IM/GM players to tune against. Because _they_ are the ones _I_ want to beat. But you still overlook _the_ issue. Exactly how would one get access to a 60 million dollar computer to do lots of tuning against commercial programs? With Cray Blitz we were lucky to play in two tournaments a year, one the ACM or WCCC, the other a human event when we could find one, to do what little tuning we could do. It is _not_ easy to get access to such hardware except for special cases like WCCC events. But tuning? No chance... > > >> >>When everyone figures out that ICC is 'there', this will change... :) > >I am not sure about it. >I think that most of the games in ICC are faster than 2 hours/40 moves so if you >want to train for WCCC then it may be better to play many games against the >commercial programs at tournament time control and not in ICC. > >Uri _IF_ your only goal is to win the WCCC, maybe. That has _never_ been mine. And that sort of testing is totally useless if you want to beat IM/GM players. As I have said thousands of times...
This page took 0.02 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.