Re: Tomsky's Theorem: P(A|A) = 1



On Aug 19, 12:57 am, Jack Tomsky <jtom...@xxxxxxxxxxxxx> wrote:
Since Afonso has given me credit for this profound theorem, I will now provide a detailed derivation.

If P(A) > 0, then P(A|A) = P(A,A)/P(A) = P(A)/P(A) = 1.

In particular, if A is the event that a team has won 62 of its first 120 games, then the conditional probability that it has won 62 of its first 120 games, given that it has won 62 of its first 120 games, is one.

Jack

Here is an interesting extension of your theorem. Suppose my prior
probability that the team has won 62 of its first 120 games is P
(which may not necessarily be equal to 1 -- in fact, I suspect there
may be people somewhere around here that may not assign a prior
probability of 1 to an event that has already happened). Then by
Bayes' theorem:

p(A|A) = p(A|A) * P / ( p(A|A) * P + p(A|not A)*(1-P) )

and since p(A|not A) is 0 (i.e., the probability that the team has won
62 of its first 120 games, given that is has NOT won 62 of its first
120 games is zero), we find that the posterior probability that the
team has won 62 of its first 120 games is 1. Therefore, regardless of
which prior probability I use (as long as P > 0), the posterior
probability must be 1.

Now suppose I consider it impossible that the team has won 62 of its
first 120 games (even though it has already won 62 of its first 120
games) prior to observing that the team has won 62 of its first 120
games. Therefore P = 0. Then:

p(A|A) = 0 / 0

which is undefined. I can conclude from this that there is not
sufficient evidence that the team has won 62 of its first 120 games.

m00es

.



Relevant Pages

  • Re: Deal or no Deal
    ... You keep using the phrase "probability theory" but I feel you don't ... half of all games end ... doesn't matter what sequence the other boxes are opened. ... I want to bet 100 UKP that the money shot box is ...
    (uk.media.tv.misc)
  • Re: Deal or no Deal
    ... You keep using the phrase "probability theory" but I feel you don't actually ... half of all games end ... matter what sequence the other boxes are opened. ... I want to bet 100 UKP that the money shot box is blue. ...
    (uk.media.tv.misc)
  • Re: Deal or no Deal
    ... The guy had 5 boxes left i.e. his own plus 4 others. ... probability theory. ... If you went through them and identified all the games where, at the five box stage, there were four of one colour and one of the other, you would find that the final box was red half the time and blue half the time. ... I want to bet 100 UKP that the money shot box is blue. ...
    (uk.media.tv.misc)
  • Re: Hastings
    ... Below is the probability of each outcome, ... Back here in the real world, knowing that 70% of the games between two ... players will be drawn gives you a huge chunk of inform- ... all ratings and all rating differences, I'll be your most eager reader. ...
    (rec.games.chess.misc)
  • Re: Feller, chapter 1, problem 5c
    ... > which leads to probability of 1/2 - events ac and bc out of the four ... > sample space for no decision is reached in 3 of fewer games: ... the probability of no winner is ...
    (sci.math)