Author: James
Date: 06:39:38 05/15/01
Go up one level in this thread
On May 14, 2001 at 20:36:58, Dann Corbit wrote: >On May 14, 2001 at 20:27:30, stuart taylor wrote: > >>On May 14, 2001 at 20:03:28, Steve Maughan wrote: >> >>>Stuart, >>> >>>> So I'm asking if it would atleast not ever lose a game to 5000, if indeed >>>> 2100, and even, if indeed it would not even possibly lose a match against one >>>> of them? >>> >>>I think the jury is still out regarding CM8000 compared to CM6000, which IMO (I >>>think most would agree) is the strongest of the previous CMs. With regards to >>>CM2100 this is a completely different engine written by the Spracklens circa >>>1990. It was quite weak even in its day and I do remember Rex chess thrashing I can't prove any of that, indeed, my tests from cm5500 through cm8k show that cm 6k is stronger than cm 8k, and that cm5500 is stronger still . >>>it. IMO all of the CM range is considerably stronger than CM 2100. >>> >>>That said, it is irrelevant if CM 8000 loses one game to ANY program i.e. has >>>virtually no statistical meaning. I'm sure in a series of 100 games CM 8000 >>>would lose one match to CM 5000 and probably also one to CM 2100. >>> >>>Regards, >>> >>>Steve >> >>Of course one game means little, but something way above doesn't even do that, >>and when there is -1 point from 100 games, I would expect that to come from 2 >>draws, rather than a game being lost. >> But I can tell you that although my old Turbo King always seemed slightly >>stronger (in my own tests) than all comers (including Novag Scorpio), when I got >>Genius 3 on a 386 computer, it beat TK every game without fail. > >That requires an infinite ELO difference. Perhaps it only won a small sample of >games actually played by you (e.g. 10-0 romp is believable, 100-0 is incredible >but conceivable, 1000-0 is so farfetched as to be absurd) > >>(though novag >>scorpio was slightly better against Genius 3 than that). >> But CM8000 might indeed lose (a match) against CM5000, I was suspecting. > >For a single match, losses are to be expected even for a program with vastly >superior strength (on occasion). That's why the SSDF plays a bucket full of >games before they publish anything.
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.