I just need a quick solution check.

100 dice are rolled. What is probability of getting die with number 5 at least twice if there is exactly 20 dice with number 6?

My solution: Let $A=${at least $2$ dice with $5$} and $B=${exactly $20$ dice with number $6$}. $$P(A|B)=\frac{P(A\cap B)}{P(B)}$$ and $P(A\cap B)=P(B)-P(B$ and $0$ dice with $5)$ -$P(B$and $1$ die with $5)={100\choose 20}(\frac{1}{6})^{20}(\frac{5}{6})^{80}-{100\choose 20}(\frac{1}{6})^{20}(\frac{4}{6})^{80}-{100\choose 20}{80 \choose 1}(\frac{1}{6})^{21}(\frac{4}{6})^{79}$ and $P(B)={100\choose 20}(\frac{1}{6})^{20}(\frac{5}{6})^{80}$.

Is this correct?

  • $\begingroup$ Not sure if I understood your scenario but dice rolls are independent.. so $P(A|B)$ is the same as P of getting at least 2 fives in 80 rolls $\endgroup$
    – user126540
    Apr 6 '14 at 10:24
  • 1
    $\begingroup$ @user126540 this would only be true if, among the 80 rolls, the probability of getting a five is $1/5$, rather than $1/6$, because we are given that none of these rolls can be a six. $\endgroup$
    – heropup
    Apr 6 '14 at 10:30

Yes, your calculation is correct.

Let $X$ be a random variable that counts the number of fives rolled. Look at a small case. Suppose we have $n = 3$ dice. What is the probability that, given that exactly one of the dice shows a six, that there is exactly one other die that shows a five? There are $216$ total distinct outcomes, of which there are $\binom{3}{1}\cdot 5^2 = 75$ for which exactly one six occurs. Of these outcomes, how many also have exactly one five? This is also easy to count: $\binom{3}{1}\binom{2}{1} \cdot 4 = 24$. So the resulting probability should be $24/75 = 8/25$. If we were to extend this reasoning to your case, we see that the total number of outcomes for which there are exactly $20$ sixes is $$\binom{100}{20} \cdot 5^{80}.$$ The total desired outcomes for which there is exactly one five is $$\binom{100}{20} \binom{80}{1} \cdot 4^{79}.$$ So the resulting probability is $16 (0.8)^{79}$. You should be able to calculate the other case as well.

Notice that this solution doesn't explicitly calculate the joint probabilities, but instead calculates the conditional probabilities directly through a counting argument--we enumerated the total number of desired outcomes, and divided by the total number of possible outcomes satisfying the given conditions. An alternative method of solution involves recognizing that once the $20$ sixes are given/observed, the only dice of interest, whose outcomes are subject to randomness, are the $80$ remaining ones. But once we know that these dice did not show a six, their individual outcomes are taken from the set $\{1,2,3,4,5\}$, not $\{1,2,3,4,5,6\}$. So we can also express the desired probability as $$\Pr[Y \ge 2] = 1 - \Pr[Y = 0] - \Pr[Y = 1],$$ where $Y$ is a binomial random variable with number of trials (dice) equal to $n' = 80$, and the individual probability of "success" (i.e., rolling a five) is $p' = 1/5$, not $1/6$ as in the unconditional case. So $\Pr[Y = 0] = \binom{80}{0}(1/5)^0 (4/5)^{80}$, and $\Pr[Y = 1] = \binom{80}{1}(1/5)^1(4/5)^{79}$.

  • $\begingroup$ But if I divide my results using formula I wrote I get the same result as you... So, my result is correct... $\endgroup$
    – Meow
    Apr 6 '14 at 11:35
  • $\begingroup$ Indeed, I had made an error when computing your answer. I will revise my post accordingly to reflect this. $\endgroup$
    – heropup
    Apr 6 '14 at 11:40
  • $\begingroup$ Ok =) thank you for effort $\endgroup$
    – Meow
    Apr 6 '14 at 12:37

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.