A icosahedral die has each face marked with different numbers from 1 to 20. Assume that each face is equally likely to occur on a single roll, the outcome is a random variable $X \sim DU(20)$

1) If a die is rolled twice, find the pdf of the smallest value obtained, say $Y$

2) If the die is rolled three finds, find the probability that the largest integer value is 3

Okay so DU just means "Discrete Uniform Distribution"

The answers are apparently $f(y) = (41 - 2y)/400$ for $y=1,2,\dots,20$ and 19/8000 respectively.

I am trying to work backwards and I've noticed that they probably some how multiplied 1/n twice to get that 400 in the bottom.


1) We calculate the probability that the minimum is $y$. This can happen in three ways: (i) We get a $y$ on both rolls; (ii) We get a $y$ on the first roll and something bigger than $y$ on the second; (iii) the other way around.

For (i), the probability is clearly $(1/20)^2$.

For (ii), the probability first is $y$ is $1/20$. There are $20-y$ numbers greater than $y$, so the probability of being greater than $y$ on the second is $(20-y)/20$. Thus the probability of (ii) is $(20-y)/400$.

The probability of (iii) is the same as that of (ii).

Add up. We get $\dfrac{1}{400}+\dfrac{20-y}{400} +\dfrac{20-y}{400}$.

There is a nicer (I think) way to solve 1). We have $Y=y$ iff (a) both tosses are $\ge y$ and (b) it is not the case that both tosses are $\ge y+1$. The probability of (a) is $\left(\dfrac{21-y}{20}\right)^2$. The probability of (b) is $\left(\dfrac{20-y}{20}\right)^2$. Subtract, and, if desired, simplify.

The second way may seem harder. But note that the cumulative distribution function of $Y$ is most easily arrived at in this way. Similar ideas, for $\max$ as well as $\min$, will continue to be useful.

2) Here is a nice way to solve the problem. The probability they are all $\le 3$ is $\left(\dfrac{3}{20}\right)^3$. The probability they are all $\le 2$ is $\left(\dfrac{2}{20}\right)^3$. Subtract.

Alternately, we can divide into cases: all tosses are $3$; there are two $3$'s; there is one $3$. The relevant probabilities are not hard to calculate. For example, the probability of two $3$'s is $(3)(1/20)^2(19/20)$. Add the three answers we get.

  • $\begingroup$ Could you explain (2)'s first method? I don't understand why it is $(3/20)^3$, where does the 3 inside come in? $\endgroup$ – Hawk Dec 4 '12 at 2:07
  • $\begingroup$ We are tossing a die $3$ times. The probability that on the first toss we get a number $\le 3$ is $3/20$. Just like probability we get a number $\le 8$ is $8/20$ ($20$ possibilities, of which $8$ are favourable). We have same probability for second toss, third toss, so multiply. (Unfortunately, the $3$'s come from two different places, $\le 3$, which gives $3/20$, and $3$ tosses, which gives the exponent.) I will be away for several hours, so probably cannot answer further questions until roughly $4$ hours from now. $\endgroup$ – André Nicolas Dec 4 '12 at 2:14
  • $\begingroup$ Wait, so why are you subtracting? $\endgroup$ – Hawk Dec 4 '12 at 7:05
  • $\begingroup$ We found the probability they are all $\le 3$. But that includes having everybody being $\lt 3$, in which case the largest integer would not be $3$. If $a$ is probability all $\le 3$, and $b$ is probability all $\le 2$, then probability the maximum is $3$ is $a-b$. $\endgroup$ – André Nicolas Dec 4 '12 at 7:09

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.