Author: Jorge Pichard
Date: 06:37:18 09/10/04
Go up one level in this thread
On September 10, 2004 at 07:46:40, martin fierz wrote: >On September 10, 2004 at 04:40:39, Jorge Pichard wrote: > >>On September 10, 2004 at 04:12:19, martin fierz wrote: >> >>>On September 10, 2004 at 03:20:24, Jorge Pichard wrote: >>> >>>hi jorge, >>> >>>>difference, please take a closer look. >>> >>>take a loser look?? >> >>Please be real, anybody can make a typographical error. What are you implying? >> >>>all i'll say is that your results are useless, as usual unfortunately. how many >>>times do people have to tell you that 10 games prove NOTHING AT ALL until you >>>believe it? >>> >>>cheers >>> martin >> >> >>If you take a closer look and read carefully you will notice that this match is >>still in progress, only ten games out of 20 were played, there are 10 more games >>to play. Plus the number of games is not as important as the percentage. If you >>have a match of 500 games between these two engines and the score ends in 253 to >>247 in in favor of Jonny 2.70 or the other way around, it is NOT proving that >>your match statistically is more convincing than my match of 20 games. I take >>percentage into consideration not NUMBERS of games. >> >>PS: Plus this experiment is NOT just of 20 games, but a series of 20 games >>matches between two identical engines, with the difference that I switched >>opening books for Jonny 2.70, since Jonny doesn' have an opening book. >> >>Jorge. > >hi jorge, > >1) why do you always post results of matches in progress? the end result is what >is interesting, not an intermediate result. I'm not the only one who post matches in progress....... >2) "the number of games is not as important as the percentage" - yeah, right, >having 100% after 1 game is very significant :-) Of course 1 thru 5 games are not significant, but nobody in this forum post only one game and expect to draw a conclusion. Anyway this match that I posted is not 1 games it was a halfway report, and I indicated that in my post. >3) a match of 500 games is not statistically more convincing than one of 20 >games? get real! It all depend what your testing, strength of two programs or simply finding which Arena opening opening book performs better with certain program that do not have opening book. >4) one match of 20 games has a certain error margin (which, FYI is quite high). >comparing two 20-game-matches has a HIGHER error margin than a single 20 game >match, so your argument of having more than 20 games isn't helping. First you said that my 10 games were not significant and now that you found out that I am having different matches of 20 games, and for each match I switch Arena opening book, you try to defend yourself by stating that more than 20 games have a higher error margin than a single match. Obviously you missed the entire purpose of my experiment. >do i hear you asking why i even bother to write stuff like this? i am very >grateful to the testers who have included my program in their matches (among >others patrick buchmann, claude dubois, olivier deville). i really appreciate >the efforts these people make. but all these efforts are just USELESS if not >done properly. Why do you spend so much time making tests if they are >meaningless in the end? why not spend 1% of the time you spend on your tests on >understanding the statistics of such matches?? > >cheers > martin The only person that doesn't comprehend the purpose of my testing and what I'm trying to accomplish with my finding is you. If you don't know why I am doing it, simply ask. Jorge
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.