Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: The Computer-Chess Player And The Mathematician (was: Waltzing Matil

Author: fca

Date: 07:49:09 08/18/98

Go up one level in this thread


On August 18, 1998 at 08:23:37, Dan Homan wrote:

>On August 18, 1998 at 07:26:14, fca wrote:

>>On August 17, 1998 at 10:17:11, Dan Homan wrote:

>>>It seemed clear to me that you defined the trials as separate units, on
>>>re-reading I see that "20 in a row" could stretch across trials.

>>As Dann Corbitt pointed out :-)

'Corbit', not 'Corbitt'.

>The odd thing is that the intersecting case has nothing to do with
>chess.  We don't measure match results this way.  While your question
>was worded such as to allow this meaning... I find it odd that you
>would insist on this interpretation in the context of chess matches.

Bruce Moreland and I were and still are discussing exactly these sequences in
the other sub-thread.  I snip:

-----------------------------------------
Bruce Moreland wrote on 10:51:44 8/17/98
[agreeing with my preceding post]
Right.  A match always has a defined starting and ending point.  Usually when
someone plays a computer on the net, they play several games in a row, with
little break between them.  If someone won the first one then lost the next
nine, I would say that they had lost nine out of ten, not that they'd lost
(nine) in a row.
-----------------------------------------

and more further down that thread on exactly this matter.

What constitutes a start and what constitutes the end of a trial = match is
hence a very very valid point, and exactly my intention to highlight.  Please
remember Bruce's original statistical questions arose at least partly from his
working Ferret on the servers, where anyone can challenge him at any time, and
sequences that cannot be strictly regarded as 'matches' occur.  But inferences
may be drawn from them on all sorts of bases (e.g. "I beat Ferret in a match of
3" without stating that the 2-1 "match" result was simply one of a sequence of
15 games against Ferret which Ferret won 11-4).

Does the player have an advantage in being able to terminate on success?  When
quoting a result, who determines the trial start and end?  All very valid and
chess-relevant.

Nothing odd at all, therefore.  Subtle, yes.  Intentional, yes.  Relevant, YES.

><snip>

>>This is an absolutely correct observation by Don, IMO.  Let me expand:

>Dan, not Don - You have repeated this mistype intermittently in this
>thread and it seems deliberate - I find it rather rude.

Sorry, Dan.  I cannot see any rudeness, though, and certainly none was intended.
Dann is also part of this thread.  Possibly contributing to my mistake.  The CCC
format does not permit me to back up and then down a different subthread (which
was necessary here, and still is, where one is quoting from someone in a
non-preceding post) while preserving a partly written answer. Hence I was
working from memory...

>>The question was (deliberately) a "bad question" - and that was the main point
>>of the post.  Statisticians will be familiar with "bad questions".

>So what you are saying is that you have deliberately asked poorly defined
>questions

No. At least one *bad question*.  The ambiguous definition (all heads or all
tails; continuity of trial sequences as discussed above) is a wholly differrent
matter.

Unfortunately, you snipped all that.  The B/G question (1) is a statistical
pitfall, representing a completely ***well-defined*** question which is
nonetheless bad (1/3 being an absurd answer).  At least one earlier question
here was analogous.

>, sent Dann, Bruce, and myself e-mail inviting us to answer them,
>and then took the opportunity to critique (in a very patronizing manner)
>our necessarily incomplete answers?

Incompleteness was not the issue.  Badness of the question was.

>Nice!

(0) Nothing patronising intended.  The point is "Stats are hard".  To me too.
(1) I have not made any comment or criticism re Dann's answers.
(2) I found one mathematical error in your answer (^10 instead of ^20).  Perhaps
you feel I should not have corrected it.  Sorry anyway.  btw I made a not
dissimilar mistake too, which I also corrected.
(3) Your reply to Q(4) was in the *hundreds of thousands*.  Bruce's reply was
*1*.  Is it surprising I should explore the reason why?  The reason being that
Bruce actually IMO answered the exact question asked (which was IMO partly a bad
question). You (Dan) perhaps answered what *should* have been asked.  Exactly
illustrating the statistical tightrope that has already been highlighted by
others, and which is (as Bruce pointed out) chess relevant.
(4) All the answers Bruce gave were correct IMO (given the ambiguity in
questions). He did not reply to the definitely *bad* Q(3), and I believe you
(Dan) also (rightly IMO) chose to reply to another (unasked) question rather
than Q3.  Which is not to say that Q(3) does not have an answer, but that I
question what possible use the answer could be.  (like with the 1/3 BG answer).

As you related two mathematician/physicist jokes/anecdotes in your last post,
and had made humorous references to there being a catch and rising to the bait,
I am perhaps to be forgiven for assuming I was not about to be flamed!

Kind regards

fca

"Still smilin', though"


PS: Assuming I am either stupid (always one's privilege to make such an
assumption) or in some way mocking (which I was not) just might make one miss
another possibility.



This page took 0.03 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.