Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Ooops

Author: Robert Hyatt

Date: 14:56:04 06/19/03

Go up one level in this thread


On June 19, 2003 at 17:37:49, Tom Kerrigan wrote:

>On June 19, 2003 at 16:33:02, Robert Hyatt wrote:
>
>>>You told him "pick any 10 consecutive games you want" and he did. So it's _your_
>>>method. Say what you mean and mean what you say.
>>
>>OK.. You are correct.  I was definitely _not_ talking about picking ten
>>games from _one_ handle when using several.  IE ten _consecutive_ games
>>might say something.  The four games (crafty vs stobor) were in the same
>>dates as the ten he included.
>>
>>I suppose if you play me with enough different handles, and pick and choose
>>both the 10 games and the handles you want to include, you could probably
>>find _any_ program that will come out on top of another.
>>
>>But if you look at crafty vs kerrigan only, there is a 20-12 advantage for
>>Crafty ignoring draws.  Throw in "stobor" and it becomes 22-12.  I don't think
>>that shows _any_ kind of "superiority".  And although my "pick any ten games"
>>was poorly thought out, the point remains...
>>
>>It would be much more interesting to do a comparison today, not knowing
>>much about what ws done in those 1997 games.  In 1997 I ran on several
>>machines, from a P6/200 all the way down to a p5/75 notebook.
>>
>>Here is the "10-game windows" from crafty vs kerrigan (which leaves out the
>>four games vs stobor and probably a few as "guest" as well...):
>>
>>games        Win/Lose (from Crafty's perspective)
>>
>>0-9           6-3
>...
>>29-38         10-0
>>
>>(that's all there are)
>>
>>So,out of 30 10-game samples possible from those 39 games, I find two
>>where Stobor was ahead by one point.  games 6-15 had a 4-5 stobor advantage,
>>and games 11-20 had a stobor advantage.  the _other_ 28 "samples" had either
>>an equal score (3 times) or a crafty advantage (25 samples).  The last half
>>of the "samples" are pretty overwhelming, also.
>>
>>The only conclusion _I_ would draw from the crafty vs kerrigan game history
>>on ICC is that Crafty is simply better, at least the configuration playing
>>vs kerrigan was better, whether there was any hardware advantage or not I
>>don't recall back then.  I remember my hardware was a pentium pro 200 at
>>the time (one cpu) although at the 1997 WMCCC I used a 500mhz alpha that
>>year.  Somewhere in 1997 (December I believe) I got the quad processor
>>pentium pro 200 hardware...
>
>
>Uh oh, is little Bobby Hyatt upset because I found 10 games? I can imagine your
>lower lip quivering right now. "But, but, but, that's not what I meant! Uhhhh,
>maybe I was running on my laptop! Uhhhhh, maybe you were using different
>accounts! Uhhhh, best 2 out of 3? C'mon, please?"
>

Nope, not "upset" at all.  "surprised" would be a better word.  But you
found 10.  I actually found _two_ sets of 10 games where you were very
slightly ahead.

But there are 28 _other_ 10 game combinations.  I'm therefore surprised you
would think your single group means much other than that an "exceptional
condition can happen if you try often enough."


>Have some dignity.

Develop some sense...


>
>The fact is, neither of us can prove our case to the other person's
>satisfaction.

Sure we can.  Log on to ICC tonight.  Pick the hardware.  I have PII/400's,
PIII/550's, and a 2.8ghz box.  I can run my dual with only one thread if you
have something in the 2.8ghz range.  Otherwise we can run on older stuff.  I
can probably even run on an older pentium pro 200 if you have something that
slow.

That makes it _easy_ to "prove our case" beyond a shadow of a doubt.  Let
"em play all night...


>
>You certainly can't prove that Crafty has always been stronger than Stobor
>because you've never had access to Stobor, other than my sporadic ICC
>appearances. You can't use a couple dozen ICC games played over a few days to
>prove that Crafty has been stronger than Stobor at every point in the last 8
>years. That's obviously stupid. (Yet you tried anyway.)

At most every sample point I have seen, Crafty was stronger.  If you take _all_
games, I'm not even sure you would find a single case of winning more than you
lost over 10 games, but you might.  But even that doesn't mean "you were
stronger".  I therefore find it very unlikely that at some unknown point in time
you were better, when every time we played you were not.

Of course I can't _prove_ it.  But the data suggests that your program was
_always_ weaker.  It's not a particularly big deal, however, but _you_ brought
it up, not me.




>
>Likewise, I can't use results of automated testing that I used to do vs. Crafty
>to prove that Stobor was stronger because I no longer have those logs and even
>if I did, you could say that I was cherrypicking or falsifying them.

Log on to ICC.  No cherrypicking there..

>
>Since I did say that Stobor has been stronger than Crafty, the burden of proof
>is on me if you call it into question. Since I can't prove it, I'll amend my
>assertion to, "By my recollection, in my opinion, Stobor has been stronger than
>Crafty." You can agree or disagree with me as you see fit.
>
>-Tom

OK.. There we agree...





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.