Quote:
Originally Posted by Shrike
I don't know why the 4X average triangle thing would work, and in fact I opined in the comment section at stackexchange that it wouldn't. Interesting that it appears from the simulation that it does.
Consider the ordered problem where you generate (x1,x2,x3,y) and y must fall within the triangle formed by x1, x2, and x3. Assume that the area of the circle is 1. It's easy to see that the probability that y falls within this triangle is the area of the triangle:
Let I be the indicator random variable that y falls within the triangle. Then, P("y falls within the triangle")=E(I)=E(E(I|x1,x2,x3))=E(area of triangle formed by x1,x2,x3). \Box
Now, if order doesn't matter, there are 4 choose 3 ways (i.e. 4 ways) to rearrange the set to the ordered problem. Alternatively:
P("(x1,x2,x3,x4) have point within triangle")=P("x1 is within triangle" or "x2 is within triangle" or "x3 is within triangle" or "x4 is within triangle").
Note that the events "x1 is within triangle", "x2 is within triangle", "x3 is within triangle", and "x4 is within triangle" are disjoint because only one of the 4 points can be within a triangle formed by the other three. Thus, the probability of the union is the sum of the probabilities.
P("x1 is within triangle" or "x2 is within triangle" or "x3 is within triangle" or "x4 is within triangle") = P("x1 is within triangle") + P("x2 is within triangle") + P("x3 is within triangle") + P("x4 is within triangle").
The final expression is 4*E(area of triangle), i.e 4*average area of a triangle. \Box