Author: Jorge Pichard
Date: 09:33:50 04/18/02
Go up one level in this thread
On April 18, 2002 at 10:06:12, Christophe Theron wrote: >On April 18, 2002 at 08:03:14, Jorge Pichard wrote: > >>On April 18, 2002 at 05:54:53, Jorge Pichard wrote: >> >>>On April 18, 2002 at 00:23:35, Christophe Theron wrote: >>> >>>>On April 17, 2002 at 22:26:36, Jorge Pichard wrote: >>>> >>>>>Novag Turquoise USCF estimated Rating of 2294 has won the first Game against >>>>>Chess Tiger 14.9 For Palm. Since the Novag is operating on a H8 RISC processor >>>>>of only 26.6 Mhz, I decided to match it using my wife Palm Vx overclocked to 28 >>>>>Mhz. This will be a ten games match at G/60. >>>> >>>> >>>> >>>>Once again: MHZ MEANS NOTHING. >>>> >>>>There is absolutely no logic in your choice. >>>> >>>>If you read one of the post below you will understand that a ONE MHz Pentium is >>>>as fast as a 16MHz DragonBall (approximately). >>>> >>>>So now tell me, where is the logic in your choice of the 28MHz Palm??? Maybe the >>>>H8 is 16 times faster than the Palm, when the MHz are equal. Who knows? >>>> >>>> >>>> >>>> Christophe >>> >>>Christ, Once again: MHZ MEANS NOTHING, therefore I decided to test it against >>>the Vx instead of my Prism. Next match will be against my Prism 54 Mhz, so I can >>>have a reference point of comparison. >>> >> >>PS: I came with a better idea, and that is to play two games with white same >>opening one with Palm Vx at 28 Mhz and the other game with the Prism at 54 Mhz >>Back to Back, and two games with black same opening again with both Palms back >>to Back. You are probably wondering why is this necessary, since MHZ MEANS >>NOTHING. But at least I am comparing Both Palms against the same opponent with >>the same opening, so if it loose twice, you can only blame it on the Opening >>Book, NOT the Slow Palm Vx. Also I didn't mentioned the MHz as a point of >>comparison, but the estimated USCF rating of 2294, which to me is closer to >>2200. >> >>Pichard. > > > >You cannot do a reliable comparison with only a few games. > >It is preferable to have more games played by a single entity in order to get >better statistical relevance. > >What is the point in the comparison between 28 and 54 MHz? > >If you want to know the elo difference, it is easy: > > elo_diff = 70 * log(speed_ratio) / log(2) > >where speed_ratio = (TigerMark at 54MHz) / (TigerMark at 28MHz) >(and not 54/28). > >It took years to establish this formula, and it works very well in computer >chess. > >With a few games, you could as well conclude that running CT Palm at 28MHz is >better than running it at 54MHz. Statistically it can happen, and it will even >happen quite often. > >We do not even have a measure of the strength of CT Palm for a fixed speed, not >even in the 100 elo range, and you want to get 2 measures at the same time? > >It is not possible. > >I'm telling you this because I fear that you waste your time by trying to draw >more conclusion that it is possible from the games you play. Focus your work on >establishing ONE measure, and your time will be well used. > > > > Christophe Christophe, I am NOT trying to get two measures at the same time, since it is required at least 100 games to come out with a concrete estimated ELO. I am simply trying to provide games where the two Palms would be matched against the same opponent with an estimated rating already established by the USCF to be 2294. And you are correct to say that the Novag Tuquoise could be several times faster than even my Prism, since it is using an H8 RISC Style processor, but don't forget that David Kettinger program is almost 5 Years Old, and is much weaker than your program on comparable hardware. Therefore, it is obvious that the H8 RISC Style processor is several times faster than any available Palms. Pichard.
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.