Someone wrote n letters to different people, sealed them in…
Someone wrote $n$ letters to different people, sealed them in envelopes, and then wrote the $n$ different addresses on the envelopes at random. Determine the probability that at least one letter will be correctly addressed.
Explanation of Method
Let $A_i$ = "letter $i$ gets the correct address". We want $P(A_1+\dots+A_n)$. The events overlap, so we use the inclusion–exclusion formula. The probability that $k$ given letters are all correct is $\frac{(n-k)!}{n!}$ (Problem 115), and there are $\binom nk$ ways to choose them, so the $k$-th term equals $\frac1{k!}$. The result is a partial sum of the series for $1-e^{-1}$.
Step 1: The model
The addresses form a random permutation of $n$ items; all $n!$ are equally likely. $A_i$ = "letter $i$ is correctly addressed".
Step 2: Inclusion–exclusion
\[ P\Big(\sum_iA_i\Big)=\sum_iP(A_i)-\sum_{i\lt j}P(A_iA_j)+\sum_{i\lt j\lt k}P(A_iA_jA_k)-\dots \]
Step 3: The terms
For $k$ given letters, the probability that all are correct is $\frac{(n-k)!}{n!}$ (the other $n-k$ letters may go anywhere). There are $\binom nk$ choices, so the $k$-th sum equals
\[ \binom nk\frac{(n-k)!}{n!}=\frac{1}{k!}. \]
Step 4: The answer
\[ \boxed{P=1-\frac1{2!}+\frac1{3!}-\dots+(-1)^{n+1}\frac1{n!}=\sum_{k=1}^{n}\frac{(-1)^{k+1}}{k!}.} \]
Step 5: The limit
Since $e^{-1}=\sum_{k=0}^\infty\frac{(-1)^k}{k!}$,
\[ P\ \to\ 1-\frac1e\approx0.6321 \qquad (n\to\infty). \]
| $n$ | $P$ |
|---|---|
| $1$ | $1.0000$ |
| $2$ | $0.5000$ |
| $3$ | $0.6667$ |
| $4$ | $0.6250$ |
| $5$ | $0.6333$ |
| $6$ | $0.6319$ |
| $7$ | $0.6321$ |
Already for $n=5$ the probability is practically $1-\frac1e$. Surprisingly, the answer hardly depends on the number of letters: with $10$ letters or $1000$ letters, about $63\%$ of the time at least one letter is correct.
Step 6: Check for $n=3$
The $6$ permutations of $1,2,3$: $123$ (3 correct), $132$, $213$, $321$ (1 correct each), $231$, $312$ (none). At least one correct: $4$ of $6$, i.e. $\frac23=1-\frac12+\frac16$ ✓.
Result
| Answer | |
|---|---|
| Probability | $\sum_{k=1}^n(-1)^{k+1}/k!$ |
| Limit | $1-1/e\approx0.632$ |