Author: Vincent Diepeveen
Date: 09:59:37 01/18/00
Go up one level in this thread
On January 18, 2000 at 12:28:51, Will Singleton wrote: >On January 18, 2000 at 12:24:00, Vincent Diepeveen wrote: > >>On January 18, 2000 at 11:27:43, Will Singleton wrote: >> >>>On January 18, 2000 at 11:15:19, Francesco Di Tolla wrote: >>> >>>>The formula initially used the clock instead of the logarithm of it. It was even >>>>worse so I suggested to take the log of the clock which, at that time, seemed to >>>>give a better description of threal situation. >>>>I suggested the log because it is quite clear that a doubling of the clock does >>>>give just a little extra elo and not a doubling of it. There are many >>>>"non-linearities" both in the increase of strength with the clock and in the >>>>increase of the elo with the increase of strenght. >>>> >>>>Probably there is no such an easy formula to be found to better describe the >>>>relationships between programs at different clocks, but this is the best we have >>>>found so far. >>>>If you have a better idea (and saying that it sucks is not a better idea) let >>>>all us know. >>>> >>>>regards >>>>Franz >>> >>>Yes, I am always open to suggestions. Come up with a better formula, Vincent, >>>I'll use it. >> >>Remove the silly R(Mhz) formula Will. >> >>>Will > >It's only *silly* if you can give me a better one. Do you not think there is >some relationship between cpu speed and performance? If so, what is that >relationship? When getting above 10 ply i don't see much relationships anymore indeed. I see only that if you're above 2800 rated that you get nailed and raped by 2030 national rated players (2300+ ICC blitz in this case) if they after 30 lost games find finally a line/strategy where your programs sucks at. Very cool representing was the approach taken by a certain person Maup. Zeke had to noplay him for Secret (help abuse #8) and we continue doing that till the new version comes out (tomorrow evening?). Those dudes usually don't take a shot at low rated computers (2400 or something). Therefore it's very tough to say things. Apart from that there is the formula issue. Crafty will never play Maup, as crafty formula doesn't allow maup to match it (rating-400). The account judgeturpin wasn't discovered so far by Maup. These 'trial and cleverness' strategies work obviously only very well if the computer is much higher rated. Very important also, hope to fix that tonight, is contempt factor. Diep didn't have contempt so far, or it does have one, but after some years it didn't work anymore. Those things have a bigger influence than the Mhz of a machine. Vincent. >Will
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.