Re: [Pan-users] Strange line counts

[Brian Morrison]

On Sun, 14 Mar 2004 22:37:22 +0100 in
address@hidden Christophe Lambin
<address@hidden> wrote:

> On Sun, 14 Mar, 2004 at 00:01 +0100, Brian Morrison wrote:
> > I've noticed a few posts in alt.sysadmin.recovery where the poster
> > seems to manage to have two headers paired, in other words they are
> > there twice (shades of Catch 22!).
> > 
> > One of these is Lines: and the other is X-cache:
> > 
> > When this happens Pan decides that the actual number of lines in the
> > article is somewhere about 57224 lines, even when the true number is
> > 11.
> 
> Where do you see this? In the header pane? Pan doesn't base this on
> the Lines header, but on what the newsserver reports in the XOVER
> command. So, this looks like a server problem.  Got an example
> message-id + group?

Yes. I've pasted in the article that causes this to happen for me below
but note that every post in this group by the author has this problem.
Note that there are two Lines: headers and two X-Cache: headers. Pan
reports this one as having 57394 lines.....

Approved:             not very carefully
Date:                 15 Mar 2004 23:30:14 GMT
From:                 address@hidden (Alan J Rosenthal)
Lines:                44
Lines:                44
Message-ID:           <address@hidden>
Newsgroups:           alt.sysadmin.recovery
Organization:         Dynamic Graphics Project, University of Toronto
Path:                
news.globalcrossing.net!nntp1.roc.gblx.net!nntp.gblx.net!nntp.gblx.net!
priapus.visi.com!orange.octanews.net!news.octanews.net!green.octanews.n
et!news-out.octanews.net!news.glorb.com!newsfeed2.telusplanet.net!newsf
eed.telus.net!cyclone.bc.net!torn!utnut!utcsri!devnull
References:           <address@hidden>
<address@hidden>
<address@hidden>
<address@hidden>
<address@hidden>
<address@hidden>
<address@hidden>
<address@hidden>
<address@hidden>
<address@hidden>
<address@hidden>
Subject:              Re: Further confusing the masses
X-Cache:              nntpcache 2.4.0b5 (see http://www.nntpcache.org/)
X-Cache:              nntpcache 2.4.0b5 (see http://www.nntpcache.org/)
X-Nntp-Posting-Host:  atlas.dgp.toronto.edu
X-No-Ahbou:           yes
Xref:                 news.globalcrossing.net
alt.sysadmin.recovery:319538
MIME-Version:         1.0
Content-Type:         text/plain

"Shmuel (Seymour J.) Metz" <address@hidden> writes:
>at 01:56 AM, address@hidden said:
>>A classic example is the 3-doors game.
>
>Yes. What is classic is the ambiguous formulation of the question,
>allowing you to legitimately debunk any numeric answer.

True, but no one seized upon that issue.  Everyone seems to have been
assuming the usual version of the underlying problem.

The issue is the algorithm employed by Monty Hall.  The rule everyone
seems to have been assuming here, and which is meant to be implied by
the
standard problem statement (but I agree with Shmuel, the standard
problem
statement is definitely inadequate here), and for which it's
advantageous
to switch in accordance with the reasoning posted here, is:  Whether you
guess correctly or incorrectly on your first guess, Monty always opens
one of the two remaining doors.  If you guess correctly, he opens a
random
other door with equal probability; if you guess incorrectly, he opens
the
other door which presents a goat.

Another algorithm for Monty which is compatible with the abbreviated
problem
statement is:  If you guess correctly, he opens a random one of the
other
two doors.  If you guess incorrectly, he does not open another door and
you don't have the option of switching.  In this case, you should never
switch.

Any intermediate situation can be constructed by changing the
probability that
he will open a door when you guess correctly to a value other than 0 or
1.

Another possible Monty algorithm is that he opens a random one of the
three doors (i.e. he might open the one you chose), chosen independently
of your choice.  In this case, if his door is not the same as yours
and his door shows a goat, then if I'm not mistaken your odds are even
("50-50" in luser-speak).  There are any number of possible Monty
algorithms.

Here's another intro probability problem I quite like:
You have three cards: one is red on both sides, one is black on both
sides,
and one is red on one side and black on the other.  You put them into a
hat, and draw a uniformly-randomly-selected card and put it on the table
with a uniformly-randomly-selected side showing.  The top side shows
red.
What is the probability that the bottom side also shows red?  Again, for
those who find the result "unintuitive" (which would reveal someone
else's
misunderstanding of statistics), this is testable by the monte carlo
method,
or by betting for money against your lusers.

-- 

Brian Morrison

bdm at fenrir dot org dot uk