Author: Gian-Carlo Pascutto
Date: 15:05:02 09/17/04
Go up one level in this thread
On September 17, 2004 at 17:57:15, Dann Corbit wrote: >I suppose your question is a good one, because we would have to ask: >"What is the control?" >In order to answer that question clearly, we would also need to benchmark other >approaches like 0x88 to see if there is any connection. Correct. His paper assumes in some parts that bitboards are inherently a faster approach than the classic one. Given that there have been at least 500 flamewars about that issue on this board, I don't think I have to explain that stating that without proof is a bit 'controversial' to say the least. >>It seems he set his goal too high. Just a well-working parallel program is >>already quite hard. I believe this may have been discussed before the paper >>started even, as he was a poster here? >> >>I won't argue that these people are doing good things but to call this >>scientific seems to be quite far-fetched to me. > >Accurate reporting of observations is good science. "We set out to test how x relates to y. We conclude that it should be investigated how x relates to y." His reporting is good. It's just what he set out to do and ended up concluding that is somewhat silly. >When it comes to the actual findings in reports, very frequently that is where >some really bad science originates. Very true, which is why I'm so happy that he included his source code - at least these results can be verified. -- GCP
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.