Author: Johan de Koning
Date: 20:06:09 05/24/03
Go up one level in this thread
On May 23, 2003 at 08:37:20, Kurt Utzinger wrote: >On May 23, 2003 at 06:48:11, Johan de Koning wrote: > >>On May 23, 2003 at 04:48:36, Harald Faber wrote: >> >>>On May 22, 2003 at 21:59:19, Johan de Koning wrote: >>> >>>>On May 22, 2003 at 20:55:15, Axel Schumacher wrote: >>>> >>>>>Hi, >>>>>just to add some more data to CM-personality discussion. Let's face it, >>>> >>>>Let's cut things short. >>>>I don't think your (Kurt's, etc) results are useless, and I never said so. >>>>I don't think your (Kurt's, etc) methods are rubbish, and I never said so. >>>> >>>>This discussion started when Kurt was upset by a, let's say, rather creative >>>>interpretation of my words. And now everyone is PO'ed. :-( >>> >>> >>>Not me. :-) >>>However, if you think you are misinterpreted, why do you leave this >>>misinterpretation open and don't clarify the questionable points? >> >>Basically because I do have a life, and checking CCC 4 times a day for possible >>inaccuracies is not part of it. In fact I'm usually 2 days behind, reading. On a >>fast moving forum like this one it makes little sense to reply after 2+ days, >>and that's the main reason for me not to try. >> >>Usually that's not a problem because there is a huge base of regulars (whose >>lives do include checking CCC pretty often :-) Together they form the spine of >>this forum, providing knowledge, experience, wisdom, and news whenever it's >>needed. Even when a few (in)famous individuals take a day off, or a month. >> >>In this particlar case Kurt reacted quickly to a misinterpretation and many then >>reacted quickly to Kurt, hence the fuss. Unfortunately virtually no one >>questioned the misinterpretation itself, which still puzzles me. By the time >>I caught up the damage was already considerable and my lame attempt at humor >>fell flat. >> >>So now I'm sad and PO'ed at the same time. Which is a rather weird feeling, >>especially after just having played some pretty exciting games at Leiden. >>I'm glad though at least 1 person here is not PO'ed. >> >>>BTW when I changed my systems and HDs I lost your email adress, could you give >>>it to me or write some clearing words to info@harald-faber.de ? >> >><insider joke mode> Backup, Harald, Backup! </insider joke mode> >>Or alternatively, look at poster. :-) > > Under http://www.talkchess.com/forums/1/message.html?297506 I wrote the > following: > > "At Leiden Johan de Koning mentioned he would not believe that the King > plays better with other settings and that there were played too less games > to say that other settings are better then the standard ones. I think that > there is enough evidence to come to an opposite conclusion. We have played > thousands of games at longer time controls with default/other settings but > only published the most successful ones (default does not belong to them). > If an author is not prepared to accept this fact, I see indeed no reason > why CM9-fans should further test the engine with the aim to prove > something. I was just angry and therefore decided to stop testing. What I > do not understand is why there is made such a tumult about this matter." > > It may indeed be that things went from the beginning wrong and then > completely out of control. There is one thing I have learnt in the > meantime: never to react in such a way on the basis of statements > from third party. And it may also be that my first posting and the further > ones were written when I was in an emotional and somewhat irrational state. > In the meantime I have cooled down although I am still unable to find out > where exactly the so called misinterpretation lies. I'm sorry I haven't been very clear on the interpretation issue. Though I don't have my original words on record, my intention was like this: Using autoplayers to improve an engine is hopeless. To be "pretty sure" about 1 small change or addition you'd need thousands of games against a wide variety of oppos (including GMs). And then again after the next change. Life is literally too short for this. Trying to speed up this process (100 x at least) will result in tuning for a very specific task. Apparently good changes will accumulate over the years, with serious inbreeding as a result. Or in other words, as the programmer I'm interested in understanding changes and extensions and adopting the (IMHO) lasting ones. Changing the defaults to eg SKR and renaming it from 3.23 to 3.30 would be easy, but pointless. After all, we already have 3.23 + SKR. There is however little doubt that some alternative settings make the current version perform very well against contemporary programs. I was actually tempted to use SKR at Leiden. But being lazy and slow, I don't have any experience with SKR and the De Gorter book, nor with high selectivity and endgames. (OK, strong engines don't need to play endgames. :-) So I used the exact same set-up that performed fairly well in the Dutch Open last October. Also I would still like to see the SSDF testing the SKR settings if that's OK with them. But due to technical difficulties that's looking like a long term plan at the moment. :-)
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.