Here is a clearification of my previous argumentation.

What is important here the distinction between
a simple probability P(X), i.e. the probability for X,
a conditional probabilitiy P(X|Y), i.e. the probability for X, given Y.

We have one car (signified by 1) and two goats (signified by 0). Each permutation has the same probability. The different permutations are:

100
010
001

Let us suppose I choose in the first place door A (the first column) and get the information that behind door B is a goat:

.0.

which decides between 010 and 001 only, but contains no additional information about 100. As a result I have now more information about door C than about door A.

The trick is that the probabilities are not independent (there is no 0.33 chance individually for each door):

100  --> P(A) = 0.33
010  --> P(B) = 0.33
001  --> P(C) = 0.33

is converted to

100  --> P(A) = 0.33
010  --> P(B) = 0.0
001  --> P(C|non P(B)) = 0.66

The condition does not effect P(A), because there was nothing to choose.

Or in other words: What remains constant is P(A) as well as the sum P(B) + P(C), but we know now that P(B) is 0 (the only new information provided) so it follows that P(C) is 0.66.


For those who are still not convinced try this modified example. We have 1 car and 49 goats. I choose in the first place door A. The game master opens 48 other doors:


.00000000000000000000000000000000.0000000000000000

Where is the car? The proability for door A is still 0.02, but for door No 34 it is 0.98. Anybody who still thinks it is 50:50?

Dieter


_______________________________________________ Delphi mailing list [EMAIL PROTECTED] http://ns3.123.co.nz/mailman/listinfo/delphi

Reply via email to