home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.math
- Path: sparky!uunet!charon.amdahl.com!pacbell.com!sgiblab!zaphod.mps.ohio-state.edu!sol.ctr.columbia.edu!news.columbia.edu!psinntp!psinntp!scylla!daryl
- From: daryl@oracorp.com (Daryl McCullough)
- Subject: Re: Games with Nonmeasurable Sets
- Message-ID: <1992Nov5.045644.21270@oracorp.com>
- Organization: ORA Corporation
- Date: Thu, 5 Nov 1992 04:56:44 GMT
- Lines: 63
-
- pratt@Sunburn.Stanford.EDU (Vaughan R. Pratt) writes:
-
- >Now where did this depend on the cardinality of the well-ordered deck?
- >The *only* relevant fact, an elementary measure theoretic one, is that
- >only countably many cards can be assigned a nonzero probability.
- >(Proof: every such card is in the finite set of cards assigned
- >probability at least 1/n for some postive integer n, and there are only
- >countably many such sets.)
-
- Vaughan, I think you have completely missed the point of the original
- posting. Let me repeat the key points:
-
- 1. There is one card for every real between 0 and 1.
- 2. Cards are dealt randomly according to the usual Lebesgue
- measure on reals.
-
- So the probability of being dealt any particular card is precisely 0.
- However, for any particular real r, the probability of being dealt
- a card greater than r in the ordering LT is the Lebesgue measure of
- the set
-
- { r' | LT(r,r') }
-
- Since this is the complement of a countable set, it has Lebesgue measure
- 1.
-
- >The resolution of the paradox is that whatever the distribution of a
- >player's cards, once her opponent knows it, her estimated likelihood of
- >winning becomes realistic. This is true not just for the countable
- >deck, as you imply, but for all decks, including Daryl's.
-
- In the case of the uncountable game, the distribution is known ahead
- of time. The paradox *follows* from facts about the Lebesgue measure.
-
- >The one thing I will concede is that the paradox is harder to see
- >through when you state it for a higher cardinal.
-
- >But this was my point in the beginning. The right approach is to ask
- >why there is no paradox at lower cardinals (answer: look at the
- >probability distribution), and then increase the size of the deck to
- >see if anything changes: no it doesn't. The measure disappeared back
- >when the set was well-ordered, so continuing to refer to it is merely
- >a red herring (again, my point).
-
- A completely mistaken point, I'm afraid.
-
- >I don't think this contradicts anything anyone else has said; in
- >particular I think it is completely consistent with Herman Rubin's
- >answer, even if it looks superficially different. It merely proposes a
- >different line of attack on the paradox: ask what role the large
- >cardinal is playing, and attack that question by asking for the
- >smallest cardinal for which the paradox works.
-
- It works for uncountable cardinals, but not countable ones. I think
- the reason is that measures are countably additive, so there is no way
- to give probabilities on the naturals so that for all n, the measure
- of the set of m such that m < n is always 0.
-
- Daryl McCullough
- ORA Corp.
- Ithaca, NY
-
-
-