Author: Sandro Necchi
Date: 12:37:12 06/28/05
Go up one level in this thread
On June 27, 2005 at 16:14:26, Uri Blass wrote: >On June 27, 2005 at 13:19:49, Peter Skinner wrote: > >>On June 27, 2005 at 10:56:28, Uri Blass wrote: >> >> >>>It was not a non stable version but a version with bugs. >>>The reply to your question why was already given. >>> >>>SMK did not know that it is buggy version. >> >>In fact in a thread below Sandro stated it was a beta version that had bugs. And >>that it was 150 elo weaker. One _has_ to assume that they _knew_ it was beta, >>and that it had bugs. > >No > >I believe that >Stefan believed that it is the best version that he had in that time. >He made some change that he believed to be an improvement and he did not notice >that he deleted some part. > >I think that this was a failure of both stefan and sandro. No, I did my best to convince Stefan, but he believed in the result of the 2004 WCCC and on the blitz tests. As you know I look the programs play and evaluate them depending on how they play and other things I have learned in more than 28 years experience on computer chess and chess programs. Most programmers believe in tests and on statistics. I believe in what I see...and how I do it... I have been able several times to estimate almost correctly how much a new version was better out of a few games. I already told you, but you do not believe this. I do not care if people believe me or not. I get fun on doing this and I'll do it until I will get fun. It was my system against std. testing systems and once again I was correct. I tell you another secret: do you remember when we lost to Junior in 2001 WCCC (I think) a won game only because the program understimated a passed pawn? Well, just before the 2004 WCCC I told Stefan I was worrying about our program (Shreder 6 preliminary version) not handling the passed pawns correctly in many positions...he said that I was exaggerating a little on this...who was correct? Even a very smart guy like Stefan is sometimes difficult to convince and only evidence can make him change his mind. I do not like to have to demostrate things as I say things only when I am sure, no matter how I do to get these info. My system may be an old system, but I like it and so I do not want to change it as has been correct so far... > >Sandro could probably do a better job in explaining stefan that Shredder is >weaker by showing him positions that the new shredder is weaker than the old >shredder. You should know that programmers are not so easy to convince when they like to believe in something... Sandro > >If we want to talk about examples then I think that the fail low of shredder >against falcon in WCCC could be a good example and based on my memory commercial >shredder is clearly better in solving the fail low. > >I read in the israeli chess newspaper that shredder needed almost an hour to >solve some fail low in that game and Falcon used that time to find the reason >that shredder failed low. > >Maybe Falcon could not find some good move in case that shredder played it >faster. > >Uri > > > > >> >>>Junior is a positional program? >> >>Yes it is. Junior and Hiarcs are probably the best positional chess programs >>around. >> >> >>>I do not believe it for a moment. >>>Junior is a fast searcher. >> >>Exactly. Faster search = deeper depth = more positional play. >> >>>Fast searcher cannot have a lot of knowledge in it's evaluation so Junior >cannot be a positional program. >> >>Programs can become positional without knowledge but by deeper search. This has >>been stated many times on this very board. > >deeper search can help to find better positional moves but I cannot define fast >searcher as a positional program and there are a lot of things that you cannot >find by deeper search. > >For example if you do not understand the fortress concept then there are cases >when no search can help you to avoid a draw by fortress. > >Shredder shows some understanding of the fortess concept and there are fortress >position that it evaluates as 0.00 or close to it(of course this understanding >can be improved and there are a lot of fortress positions that shredder does not >understand but at least stefan tried to do something in that direction). > >Junior shows no understanding of the fortess concept. > >Uri
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.