# Wrong Reasoning about the problem of breaking a stick in $2$ points and build a triangle with the $3$ parts.

For homework I was asked to solve this classical problem "If you break a stick at two points chosen uniformly, the probability the three resulting sticks form a triangle is 1/4" and ok, it must result $1/4$. But I can't figure out why. This is my reasoning (obviously wrong somewhere!). Let $X,Y$ be the points (picked uniformy at random on the stick) There are $3$ case:

1) $0<X<Y<1$ with probability $\frac{1}{2}$ to happen

2) $0<Y<X<1$ with probability $\frac{1}{2}$ to happen

3) $0<Y=X<1$ which has probability $0$ to happen.

Consider the first case: ($X$ indicate now the length of the segment $[0,X]$) the triangle inequalities leads to:

$X<(Y-X)+(1-Y) = 1-X \longrightarrow X<\frac{1}{2}$

$Y-X < (X) + (1 - Y) \longrightarrow Y-X < \frac{1}{2}$

$1-Y < Y \longrightarrow Y > \frac{1}{2}$

which leads to a probability of "success" of $\frac{1}{8}$

and the "global" probability of this event is $\mathbb{P}(\text{case} \ 1)*\mathbb{P}(\text{ I can build a triangle in this case})=$ $\frac{1}{16}$. for the second case is the same. the third case has probability $0$, so it doesn't give any contribution to the probability. According to my reasoning the global probability is $\frac{1}{8} \neq \frac{1}{4}$.

I can't find why my reasoning is wrong. I don't want a solution of the problem, but WHERE and WHY I've made an error. I know there are a lot of questions about this problem, but I need a correction in this reasoning not a complete solution, and I didn't found an answer in the other questions :) Thank you in advance.

-
sorry, $y-x < x +1-y$is equivalent to y-x<1/2$right? But it doesn't change the final value i get i think – Riccardo May 13 '13 at 21:09 That is one problem. Another problem is that you seem to assume certain things are independent when they are not. For example, when you assume that case 1 holds, the probability that$X < \frac 1 2$is greater than$\frac 1 2$. You need to tease this apart a little more carefully. – dfeuer May 13 '13 at 21:10 @dfeuer ok, this make sense, i start working on this observation! – Riccardo May 13 '13 at 21:12 If$X$is the smaller of the two, the probability that$X<1/2$is more than$1/2$. This is not an unconditional probability... – Thomas Andrews May 13 '13 at 21:19 @ThomasAndrews, i can't figure out why it is more than 1/2. It make sense ok, but i can't "prove" it let's say – Riccardo May 13 '13 at 21:25 ## 1 Answer There's nothing wrong with your (corrected) calculation. The triangular region specified by the three inequalities$x<\frac 12$,$y<x+\frac 12$, and$y>\frac 12$, does indeed have area$\frac{1}{8}$. However, the total area under consideration is the portion of the$[0,1]$square subject to$x<y$, which has area$\frac 12$. Hence the conditional probability is their ratio, or$\frac 14$. You've made things more complicated by splitting into cases this way. Looking at the$[0,1]$square, there are two triangular regions where the stick-breaking leads to a triangle; their total area is$\frac 14$. - yes, so in the end the solution is$\frac{1}{2}*\frac{1}{4} + \frac{1}{2}*\frac{1}{4}$right? i think you are right i made more complicated, but if i didn't split in cases i wasn't able to solve it... How do you find the 2 regions? – Riccardo May 13 '13 at 21:19 The trouble with his original approach is that it seems to work only by accident. I haven't quite figured out why it works in this case. – dfeuer May 13 '13 at 21:22 If$\max(x,y)<\frac 12$it's not a triangle. If$\min(x,y)>\frac 12$it's not a triangle. If$|x-y|> \frac 12$, it's not a triangle – vadim123 May 13 '13 at 21:26 @defeuer, it's unclear what his "original approach" is. If it's to cube$\frac 12$, then of course it's coincidence that it works. But if it's to calculate the area of the triangle and (incorrectly) divide by 1 instead of$\frac 12\$, it's fine. – vadim123 May 13 '13 at 21:27
@vadim123 thanks for the explanation, I've understand the error and the easy way to do it :) – Riccardo May 13 '13 at 21:30