Author: Vasik Rajlich
Date: 04:30:23 01/18/06
Go up one level in this thread
On January 17, 2006 at 11:57:41, stuart taylor wrote: >On January 17, 2006 at 11:28:23, Mark R. Anderson wrote: > >>On January 17, 2006 at 10:07:59, Rolf Tueschen wrote: >> >>>On January 17, 2006 at 08:51:58, Vasik Rajlich wrote: >>> >>>>On January 16, 2006 at 20:04:37, Joseph Ciarrochi wrote: >>>> >>>>>I just finished some testing between the betas. It was 4 min with 2 sec >>>>>increments on a decent pentium. Nooman test set. Default paremeters. All games >>>>>were versus fritz9. There seems to be no significant differences, but it is >>>>>critical to not infer too much, given n size is still small and this was against >>>>>only one engine. >>>>> >>>>>best >>>>>Joseph >>>>> >>>>> >>>>> >>>>>Rybka beta 1 58-42 >>>>>Rybka 9b 59.5-40.5 >>>>>rybka 10d 56 - 44 >>>> >>>>Hi Joseph, >>>> >>>>thanks for testing. Bad results are of course just as important as good ones - >>>>otherwise, it's not really testing :) >>> >>>What?? May I politely ask you to avoid such unbelievable nonsense? How could >>>negative results be called "testing of Rybka"? Honest testing would cause >>>positive results. At least this is what I learned in my long university studies. >>>So please, let the programmer of Rybka work in peace and stop confusing him with >>>that kind of nonsense. Thanks! >>> >>>Rolf :) >>> >>> >>>> >>>>Vas >> >>Rolf, >> >>I hope you're joking. It is most definitely *not* the scientific way to only >>include positive results, or only those we like or agree with. That's not >>science, but wishful thinking. The truth, as the saying goes, "is what it is," >>whether we may like it or not. >> >>I believe the reason Vasik wrote that bad results were "just as important as >>good ones" is that *only* such unbiased testing will tell him when he has >>improved his engine, how much, and when he is going down a wrong path after >>implementing a change. Sometimes, changes can be detrimental to an engine's >>performance (we all know that one from our testing), and this needs to be known >>also. It's just as important to know that, as it is to know the good results. >>The truth "is what it is" and not always what you may want it to be. >> >>By the way, if Vasik is reading this .... I agree with a prior post that your >>newest betas (10D and 10) are excellent predictors of good GM moves very early >>in the search. Good skill!! You deserve our applause and admiration for your >>wonderful engine. It was the best $40 I have yet spent on any chess engine. >>Keep up the good work! >> Thanks :) >>Mark A. > >I too, still hope Vasik will never put any slight less work into the longer time >playing strength, nor to compromise any bit of it, to benefit the blitz. In principle there is no need to make a decision about this. If something only works in blitz, then it can be enabled only in blitz. In practice though programmers usually look for some sort of unified principles. Vas >S.Taylor
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.