Author: Gian-Carlo Pascutto
Date: 12:13:51 04/10/02
Go up one level in this thread
On April 09, 2002 at 18:16:34, Christophe Theron wrote: >I think the surprise comes from the fact that many people overestimate what >computing power can do. > >It's good and refreshing to point out that with the right algorithms a slow Palm >can kick a PC in the ass. > >Computing power is not all. Don't expect it to cover algorithmic deficiencies. > >People take it for granted that even an average amateur program will beat them >at chess. And that a Palm is so slow that they will beat it easily. > >Obviously this is not true. > >I think it is worth remembering. You have a good point here, but I will illustrate with some approximation math why I think Sjeng 12.11 would win an infinitely long match: Looking at the latest SSDF list, we see that CT14.0 is almost exactly 100 points higher rated than Crafty 18.12. I also estimate that Sjeng 12.11 is not more than 100 points weaker than that Crafty version. So, that gives 200 ELO points ratings difference that must be overcome with hardware differences. On SSDF we also see that CT14 wins 81 elo points from a tripling of speed, or about 56 points for a doubling. I'm going to assume similar numbers hold for halvings. CT for Palm does about 800 nps on that given hardware. It does about 70knps on my Celeron 366, so that's about 83knps on a Celeron 433. That makes that CT is about a factor 100 hundred slower on the Palm than it is on the PC. (says a lot about efficiency as the CPU is only 8 times slower in Mhz) A factor 100 is about 6 or 7 doublings of speed. So the hardware can compensate for about 6*50 = 300 to 7*50 = 350 ELO points. I conclude that CT for Palm is about 150 ELO weaker than Sjeng 12.11. In a direct matchup, it's certainly not an unusual result that it will win the first two games. I realize this is fuzzy math, but I prefer to go on this rather than to make bold statements based on the result of 2 games. -- 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.