Author: Uri Blass
Date: 15:10:49 09/11/02
Go up one level in this thread
On September 11, 2002 at 17:36:44, Rolf Tueschen wrote: >On September 11, 2002 at 09:02:32, Ed Schröder wrote: > >>On September 11, 2002 at 07:51:07, Uri Blass wrote: >> > >[...] > >>>Hope for peace >>> >>>Uri >> >>Well, Thorsten achieved something remarkable tuning the Rebel parameters, both >>his versions perform significant better than the default setting. So far I have, >> >>TC-051 : 51-35 59.2% (+55 elo) >>TC-056 : 43-34 55.8% (+30 elo) >> >>Both matches are still in progress but I think the pattern is set. >> >>Well done Thorsten. >> >>Ed > >Without the exact data, the games, this remains an open question, Ed. Only the >games can speak. The scores alone don't mean the same. The questions of the >readers here are justified. I agree that we still have no proof that 051 or 056 is better than the default version but the results of Ed suggest that they are better espacially when thorsten did not test against the old version. results against Fritz are not going to prove that the new version is better not only because of statistical error but also because it is possible that it scores better against Fritz but worst against another program. It does not mean that the games against Fritz are not important because it is logical to guess that it may be better if it does better against Fritz and deciding about the conjectures to test is important. > >Let me give an example from science. I think it was the Nobel prize laureate >Monod from France who was involved with a terrible error in his laboratories. >Hewas guilty of not having controlled good enough the work of his team members. >What has happened? Well, the error was odd. His assistants had to count certain >items, their frequency in certain solutions. As it could be proven some >assistants counted the items almost always in wrongly enlarged format. How could >this happen if the field with the items was very bright and therefore to be >examined without difficulties? The answer shows a typical case in science, and >therefore certain caution was invented. I remind you of the typically >double-blind experimental setting of medical tests. That means that not only the >clients know what they are given, but also the experimentor himself doesn't know >what he's administering. Meaning placebo or not, medicament X or Y etc. >Look, if these well educated and experienced scientists must respect a basic >caution, it is very telling when Thorsten is simply believing that he has >anything under control. If it were so easy, and if people could control their >unconscious (!) preferences and intentions, we didn't need methods such as the >double blind. It is simply a fact, that we can't control us, no matter how >honest or educated we are. It is simply a human trait, that we are trying to >support our own wishful thinking. So it's not a big surprise if Thorsten, or >anyone else, gets the results he is working for. We could examine his results if >we had the exact data. And also the complete data. Here Thorsten is taking >refuge in making excuses. Perhaps the most complicated excuse is "his" >creativity vs the mere "beancounting" of the uninspired... Here he becomes to >resemble religious sect members. Simply because even if he is 100% right, we >must have the data, the complete data, otherwise his results mean almost >nothing. Not because people believe that Thorsten cheated or something, but >simply because he himself is no longer able to differentiate objective and >subjective. We know this, because everyone would be in the same difficulty if we >experimented like Thorstenin our own kitchen or lab. Therefore the necessary >control. I understand your point. I do not say that thorsten's results are proof that the new version is better. I think that Thorsten's job is to find the version that he believes that it is best and to give information. Other people's job is to verify if the version is really better before it is released. I plan to look at some games or Rebel-Fritz after all this discussion. I did not do it until now because I was too busy in looking at other games of my program or in this discussion. 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.