Author: Christopher R. Dorr
Date: 12:58:56 03/30/99
Go up one level in this thread
On March 30, 1999 at 13:47:54, blass uri wrote: > >On March 30, 1999 at 08:18:33, Christopher R. Dorr wrote: > >>On March 29, 1999 at 22:02:24, Andrew Dados wrote: >> >>> >>>On March 29, 1999 at 21:35:06, Christopher R. Dorr wrote: >>> >>>>On March 29, 1999 at 18:08:28, blass uri wrote: >>>> >>>>> >>>>>On March 29, 1999 at 13:33:20, Christopher R. Dorr wrote: >>>>> >>>>> >>>>>>In short, I believe the premise of this thread is somewhat flawed. Crafty has >>>>>>not (against humans) been demonstrated to be significantly weaker than anything >>>>>>else, especially at speed. >>>>> >>>>>The results of mark young showed that Hiarcs7 is better than parallel crafty at >>>>>10 min+20 seconds or faster time control(5+5 to 5+12) >>>>> >>>>> >>>>> >>>>>Here are the results by mark young: >>>>>Results: >>>>> >>>>>Matches with Bob Hyatt. >>>>> >>>>>Hiarcs7(PII450) Vs. Crafty(Quad Xeon PII 400) Hiarcs7 wins +5 -2 =1 TC >>>>>10min 20 sec incs. >>>>> >>>>> >>>>>Server games with Hiarcs7 against Crafty. >>>>> >>>>>TC - Blitz from 5 min 5 sec incs. to 5 min 12 sec. incs. >>>>> >>>>>Hiarcs7(PII 450) wins >>>>> >>>>>+28 -5 =7 >>>>> >>>>>Avg. speed of the crafty's per game was 668 MHz. Most games were against >>>>>multi-cpu Crafty's. >>>>> >>>>>If crafty is not weaker then how do you explain the result of mark young against >>>>>crafty? >>>>> >>>> >>>>If Crafty is not stronger, how do you explain the higher rating on ICC for >>>>Crafty? Even at standard, when response time is diminished in importance. There >>>>is conflicting data here. >>>> >>>>I remember the late 80's early 90's when Fritz1 outscored Zarkov 2.5 >>>>significantly in computer-computer matches, but was weaker when playing against >>>>people. It happened then, and it happens now. >>>> >>>>I'm not saying that Crafty *isn't* weaker, just that *I'm* not convinced it is. >>>>A Single processor Crafty (Mofongo) is over 3050 in blitz on ICC. better then >>>>*everything* else. How can you conclusively say that Crafty is weaker when it is >>>>the highest rated *anthing* on ICC? >>>> >>>>I've played just about everything out there, and honestly feel that Crafty may >>>>very well be up there with the best of them. >>>> >>>>Chris Dorr >>>> >>>> >>> Case of Mofongo is easy: see it's formula. It does not play computers.... My >>>experience shows that if you exclude computers from opponent pool your rating >>>goes some +100-200 points higher... Comparing ratings on icc seem to make sense >>>only when both handles have same formulas... Now see other crafty clones without >> >> >> >>But isn't this the point? When I hear someone say 'program X is the strongest >>there is." I generally understand this to mean *agains humans*. To me, 10,000 >>games of Fritz vs. Crafty don't mean nearly as much as 100 games of Patrick >>Wolff vs. Crafty. Against humans, Crafty has demonstrated itself to be right up >>there with the *very best*, at least at blitz. > > >The comparison is not fair if you exclude computers from crafty's opponents and >do not exclude computer from commercial's opponents. > >I understood that programs usually can get 100-200 higher rating if you exclude >computers but I am not sure if I understood right. > OK. But Crafty itself does play computers, and is *still* rated 3000. So are you saying that if it didn't, then it would be a 3200+? If so, that is not a good case for saying that it is not as strong as Fritz. > >I believe that crafty should be weaker than the commercial programs against >humans because humans can learn crafty's evaluation function and know the >weakness of crafty when they cannot do the same for commercial opponents. > >I believe that humans do not do their homework in learning Crafty because it is >not important for them to win because there is no big money prize for winning. > But there's no big money for beating anything on IC, so this caveat should apply equally to all programs on there, no? >Uri > > > > > I believe 'Why does Crafty score >>poorly against Fritz?' and 'Why is Crafty weaker [against humans, or even >>against a wide variety of opponents, including both humans and comps] than >>Fritz, if indeed it is so?" to be two very different questions. >> >>>formula restrictions... >>> Examples are Tryagain (K2-300Mhz; 2656) Razzle(K2-400mhz; 2636), >>>KillerGrob(PII 450; 2818)....and, for comparison, say junior (Ban): PII-333; >>>2930 (also without restricting formula)... >>> >> >>And again, the Grand Daddy of all Crafties 'Crafty' has set a rating record of >>3177 at blitz, better than *any human* or *any computer*, and it does play >>computers. It's currently hanging out at around 3000. >> >>And you might say that it isn't a fair comparison because it's on a quad Xeon. >>But toss Fritz on a quad....you'll get no improvement. So this arguments is >>basically saying that we should compare Fritz on it's optimum system (some >>PIII500), but not Crafty. > >junior (Ban) is running on pII-333 and not on the PIII500 so the comparison >2930 For Ban against 3000 for Crafty is not fair. > >Comparison is not fair if one program is running in cheaper machine unless both >programs are running in the optimal machine under some price. > >Uri I believe that the comparison *is* fair. Crafty is optimized to play on an SMP machine. Fritz isn't. If we play them on a single processor machine, this is giving an unfair advantage to Fritz, as this is it's optimal platform. And how do we compare them? Computer-computer testing does not (IMHO) honestly reflect strength (necessarily) against humans. 'Why doesn't Crafty win a match against on a single processor machine?' and 'Why doesn't Fritz win a match against Crafty on an SMP machine?' are flip sides to the same question. They are *both* different from 'What's the strongest machine out there?' Chris
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.