home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: rec.gambling
- Path: sparky!uunet!walter!news
- From: Jonathan Rosenberg <jxr@thumper.bellcore.com>
- Subject: Re: Not the Monty Haul problem
- Message-ID: <1992Nov23.023241.10312@walter.bellcore.com>
- Sender: news@walter.bellcore.com
- Nntp-Posting-Host: bambam.bellcore.com
- Organization: Bellcore
- Date: Mon, 23 Nov 92 02:32:41 GMT
- Lines: 52
-
- In my last post, I gave a long analysis of the cute problem posed by
- Jeff Tang. I've been rethinking what I said & I've realized the
- problem's even trickeir than I thought.
-
- All of my comments at the beginning of that post were ok, but I think
- they missed the point (though some of the reasoning will help you
- understand the rest of this). In particular, at the end of my message I
- said:
-
- On rethinking this, I realized that you might argue that the problem is
- well-defined & that the chooser just is not told the distribution used.
- That seems to be ok. In that case, answer #1 is correct. Answer #2 has
- flawed logic because it is calculating an average of repeated trails
- that are not like the original trial. I.e., answer #2 assumes that in
- each trail the envelopes will contain either 0.5 & 1 units or 1 & 2
- units. But, that ain't the original problem. The original problem says
- that you have no information about the amounts in the envelopes (other
- than 1 is twice the other).
-
- I think this is closer to the truth: the problem is well defined & we're
- just not told the distribution. Answer #1 is, in fact correct. But my
- explanation of why Answer #2 is incorrect is just wrong. (I believe the
- other 2 explanations of why #2 is bogus are also wrong. I don't think
- it has anything to do with "utility values" or with "changing the unit
- size".)
-
- The problem with #2 can be found in the calculation of the average:
-
- Answer 2: Well, the other envelope has either .5 or 2 units. Since there's a
- 50% chance of each of these, ...
-
- The problem is the assumption that the probability of the other envelope
- holding .5 units is 0.5 (and, likewise that the probability of it
- holding 2 units is also 0.5). Believe it or not, this is just not
- possible. This assumption is equivalent to saying that it is possible
- to choose a random number uniformly from an infinite sequence.
-
- This takes a little while to see, but think about it. You'll see that
- the 0.5 probability assumption implies that there is no upper bound on
- the number of units that can be placed in an envelope. If it is
- possible to put N units in an envelope, it must be possible to put 2N
- units in an envelope. Recursively, if there can be 2N units in an
- envelope, there can be 4N units. And so on, ad nauseum.
-
- The assumption that P(N/2) = P(2N) = 0.5 then implies that all numbers
- in the infinite range are equiprobable. As stated in the earlier post,
- this is not possible. Therefore, the calculation of the average is
- incorrect. In fact, since we don't know the distribution of the units,
- we can't calculate an average.
-
- JR
-
-