3

Suppose I have a stick of L length . It is divided into uniformly randomly into n parts.

What is the probability of those n parts to form a closed polygon?

It will be helpful if anyone show the steps to proceed with this problem.

ironman
  • 115
  • What are your thoughts on the problem? What have you done so far? Can you did it for a triangle for example? – saulspatz Jul 12 '18 at 17:17
  • yeah I did it for triangle. But in case of polygon I got many inequalities to solve using graph eg. 5 sided polygon. I seems to me like multidimensional . – ironman Jul 12 '18 at 17:20
  • 4
    If the longest part is more than half the length, you can't form a polygon; if it's shorter, you can. – saulspatz Jul 12 '18 at 17:22
  • For a triangle, did you use the argument about dividing an equilateral triangle into quarters, or some other approach? – saulspatz Jul 12 '18 at 17:31
  • I take x, y , (L-x-y) as lenght and use graph to plot inequalities like x>0,y>0 and plot the line (L-x-y) and find the area under curve – ironman Jul 12 '18 at 17:37
  • What answer did you get? (You should put your analysis of the triangle in the body of the question, not the comments.) – saulspatz Jul 12 '18 at 17:40

1 Answers1

8

To start with, you might look at this youTube video (starting a at 1:00) to see a different way of computing the answer $1/4$ for the triangle problem. This readily generalizes to higher dimensions.

Let us set the length of the stick to be $1$. If we break the stick into $n$ pieces, we get $n$ positive numbers whose sum is $1$, which we can think of as an ordered $n-$tuple, or as a point $(x_1,\dots,x_n)$ in $\mathbb{R}^n$ that satisfies $0\le x_1\le1,\ i=1,\dots,n\text{ and }\sum x_i=1.$ These points form the standard n-1-simplex, the $n-1-$dimensional analogue of a triangle. For example, the standard $2-$simplex is the triangle in $\mathbb{R}^3$ with vertices $(0,0,1),(0,1,0),(0,0,1).$ The standard $3-$simplex is a regular tetrahedron sitting in $\mathbb{R}^4,$ and so on.

Note that we can form a polygon out of the $n$ pieces if and only if the longest piece has length less than $1/2.$ The way this relates to the simplex is that the sum of the distances from any interior point to the faces of the simplex is equal to the altitude of the simplex, namely $1$. This is proved just as in the case of an equilateral triangle. The volume of an $n-$simplex is proportional to the product of the altitude and the ($n-1-$dimensional) volume of the base. (The constant of proportionality is $1/n,$ but we don't need that.) Since all the faces are congruent, the result follows.

So the points inside the simplex corresponding to divisions that result in polygons are those within $1/2$ of all the faces. For any face, a hyperplane parallel to that face and at distance $1/2$ cuts off a simplex of inadmissible divisions at the opposite vertex. The excluded points form a simplex similar to the standard simplex, but of half the height, so its volume is $2^{-(n-1)}$ times the volume of the standard simplex. (Recall that we're working with an $(n-1)-$simplex.) Since there are $n$ faces, the volume remaining, divided by the volume of the entire simplex is $$\boxed{1-\frac{n}{2^{n-1}}}$$ and that is the probability you seek.

I wrote a little python script to test this by simulation, and it confirms my answer:

from random import random, seed
seed()
successes = 0
trials = 100000
pieces = 7
cuts = pieces-1
for _ in range(trials):
    points = [0]+sorted([random() for _ in range(cuts)])+[1]
    stick = [points[i+1]-points[i] for i in range(pieces)]
    if max(stick) < .5:
        successes += 1
print('actual ',successes/trials, 'expected', 1-pieces*.5**(pieces-1)) 
saulspatz
  • 53,131
  • thank you sir for taking your precious time and writing a detailed answer. thanks – ironman Jul 13 '18 at 05:14
  • 1
    My pleasure. What really took time was writing the python script; I can't believe how many times, and in how many different ways, I did the sampling wrong. – saulspatz Jul 13 '18 at 12:41