Sum of Three Largest Dice
You roll $5$ standard fair six-sided dice and sum the $3$ largest values showing. What is the probability that this sum equals
Hints
- What is the maximum possible value each die can show? What does that force the three largest dice to be?
- If all three largest must be $6$, reframe the problem as counting how many sixes appear among the $5$ rolls.
- The number of sixes follows a $\text{Bin}(5, 1/6)$ distribution. Compute $P(N \geq 3)$ using the binomial PMF.
Worked Solution
How to Think About It: Each die maxes out at $6$, so the only way the three largest sum to
Quick Estimate: Each die has a
Approach: Compute $P(N \geq 3)$ exactly using the binomial PMF.
Formal Solution:
Let $N$ be the number of sixes among the $5$ dice. Since each die independently shows a $6$ with probability
The sum of the three largest values equals
$P(N \geq 3) = P(N = 3) + P(N = 4) + P(N = 5)$
$= \binom{5}{3}\left(\frac{1}{6}\right)^3\left(\frac{5}{6}\right)^2 + \binom{5}{4}\left(\frac{1}{6}\right)^4\left(\frac{5}{6}\right) + \binom{5}{5}\left(\frac{1}{6}\right)^5$
$= 10 \cdot \frac{25}{7776} + 5 \cdot \frac{5}{7776} + 1 \cdot \frac{1}{7776}$
$= \frac{250 + 25 + 1}{7776} = \frac{276}{7776} = \frac{23}{648}$
Answer: The probability is $\dfrac{23}{648} \approx 0.0355$, or about $3.55\%$.
Intuition
The key insight is recognizing that a constraint on order statistics (the three largest values) can often be translated into a simpler counting condition on individual outcomes. Here, "the top three sum to 18" sounds like it needs order-statistic machinery, but since each die caps at 6, the only way to hit 18 is for all three largest to equal 6 -- which just means at least three dice landed on 6. Once you see that, the problem collapses to a standard binomial tail probability.
This pattern comes up constantly in quant interviews and in practice. Whenever you see a condition involving extremes or order statistics, ask yourself: can I rephrase this as a condition on counts? It almost always simplifies the math dramatically. In trading, the same logic applies when you think about joint tail events -- the probability that multiple positions all hit their max loss is easier to reason about by counting how many positions are in the danger zone rather than working with the joint distribution directly.