0

There is a unit circle centered at origin on which $3$ points are chosen randomly thus we have $3$ arcs. What is the probability of $(1,0)$ lying on the longest arc?

I tried the intuition that the point may lie in any of the three arcs thus the answer may be $1/3$. Am I correct?

N. F. Taussig
  • 76,571

3 Answers3

2

Let the circle have circumference $1$. The probability $p$ we are looking for then is the expected length of the longest arc. We may assume that one of the three points is at $0$, and that the other two points $x$ and $y$ are independently uniformly distributed in $[0,1]$, in other words: that $(x,y)$ is uniformly distributed in $[0,1]^2$. This leads to $$p=2\int_0^1\int_0^x\max\{y, x-y,1-x\}\>dy\>dx\ .$$ This is an elementary, but nasty integral to compute. Therefore I let Mathematica compute it numerically. The result was $0.611111$. Since the result would have to be a rational number with small divisor this lead to the conjecture that $$p={11\over18}\ .$$ I then drew a figure and computed the integral "by hand". The conjectured value was thereby confirmed.

1

Partial answer: 3 points $(a,b,c)$ chosen uniformly at random from (0,1). Assume $a<b<c$. Let $f=b−a,\ g=c−b,\ h=1−c+a$.Get the distribution of $L=max(f,g,h)$. That is the distribution you want.

Slight simplification: Choose 2 points $(a,b)$ uniformly from (0,1). Assume $a<b$. Let $f=a,\ g=b-a,\ h=1-b$. As before, get the distribution of $L=max(f,g,h)$. That is the distribution you want.

I suspect this will not be too easy.

1

This answer provides in a probabilistic foundation of the answer of Christian.

Let $A$ denote the longest arc and let $M$ denote its length. Then:$$\mathbb P((0,1)\in A)=\mathbb E1_A((0,1))=\mathbb E[\mathbb E[1_A((0,1))\mid M]]=\mathbb E\left(\frac{M}{2\pi}\right)=\frac{\mathbb EM}{2\pi}$$

So the expectation of the length of the longest arc is essential.

Finding RHS comes to finding the expectation of the longest of the $3$ intervals that arise if two points are selected independently and according to uniform distribution on $[0,1]$.

For that see the answer of Christian.

drhab
  • 151,093