Author: Don Dailey
Date: 14:17:28 06/10/98
Go up one level in this thread
On June 09, 1998 at 10:58:24, Robert Hyatt wrote: >On June 09, 1998 at 02:38:56, Ed Schröder wrote: > >>>Posted by Christophe Theron on June 09, 1998 at 00:55:16: >> >>>>Say one game takes 60 hours = 2.5 days. 30 games will take 75 days. >>>>Using 2 autoplayer pairs the total NPS tournament would take 30-40 >>>>days which looks pretty acceptable to me. >> >>>>Interesting enough? >> >>>>Alternatives? >> >>>>Or better play the normal (10 game) 40/2:00 matches? >> >>>>- Ed - >> >>>Sorry Ed, I think this is a complete waste of time. >> >>>But maybe this has to be done one time, just to verify that we learn >>>nothing from such matches, and then definitely throw away such ideas. >> >>Why do you think the NPS tournament is a waste of time? >> >>My view, in total 30 >>4:30 vs 45:00<< games will be played. Imagine >>the following possible scores: >> >>a) 4:30 vs 45:00 5 - 25 --> Speed is decisive. >>b) 4:30 vs 45:00 10 - 20 --> My expectation. >>c) 4:30 vs 45:00 13 - 17 --> Speed is not decisive. >> >>>Please keep on doing the good work on normal 40/2:00. >> >>I will, I am expecting a positive answer for sponsoring CCL. This >>means I will get 6 top Pc's for the use of CCL. I then can move CCL >>to these six machines (so more games) and use the PII-266/64 I am >>using now for CCL for the NPS tournament. >> >>- Ed - >> >> >>> Christophe > > >I think the result will be interesting. For years, we've heard Larry >Kaufman's "doubling = 70 rating points." at a 10 x speed, that is >over 3 doublings, which should produce at least a 200 rating point >improvement, if Larry's formula is anywhere near correct. And this >means that a 30 game match ought to end up at 24-6 or so. I think it >will be interesting to see if it does. I believe it was 60 rating points per doubling. It may be that he modified this number later and it was at one time 70. In general I believe this number get's smaller as hardware increases. It may only be 50 per doubling now, or even less. - Don >the only flaw is that there are *two* degrees of freedom in the >statistical >analysis. speed *and* different program opponents. It would also be >useful to play X vs X' where X' is simply X with 10x the time per move. >Then the only variable is time, and you need fewer games to produce an >acceptable error. > >But handicap games between different programs will also be >interesting...
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.