Order is the whole point
In the last guide we learned the multiplication principle: when a task happens in steps and the number of choices at each step is steady, you multiply the choice-counts together. A permutation is what you get when you point that engine at one specific question — how many ways are there to put things in a definite ORDER? Three runners A, B, C can finish a race as ABC, ACB, BAC, BCA, CAB, or CBA: six orderings. Order is not a detail here; it is the entire question. ABC and ACB use the same three people yet count as two different outcomes.
Why does order matter so often? Because in the real world, a reshuffle frequently changes the meaning. First, second, and third on a podium are not the same as second, first, and third. The password 'cat' is not the password 'act'. The seating plan that puts Grandma at the head of the table is a different plan from the one that puts her at the foot. Whenever swapping two items produces a genuinely new outcome, you are in permutation territory — and that is exactly the permutation count we are about to build.
Arranging everything: the factorial
Start with the simplest version: line up all n distinct items in a row. The first slot has n candidates. Once it is filled, the second slot can be any of the remaining n-1. The third has n-2 left, and so on, until the last slot is forced — only 1 candidate remains. The multiplication principle says: multiply. So the number of full arrangements is n times (n-1) times (n-2) times ... times 2 times 1. That product is so common it earns its own symbol, n!, read 'n factorial'.
The crucial thing to notice is that the COUNTS 5, 4, 3, 2, 1 never depend on which item you actually placed first. Whoever you seat in chair one, exactly four people remain for chair two. That steadiness is the hidden condition from the multiplication principle, and it is precisely what lets us multiply. Five people lining up for a photo can do so in 5! = 5 times 4 times 3 times 2 times 1 = 120 ways. The growth is ferocious: 10! is already 3,628,800, and a shuffled deck of 52 cards has 52! orderings — a 68-digit number, vastly more than the atoms making up our planet.
Arranging only some: P(n, k)
Often you do not arrange all n items, only k of them in order — say, hand out gold, silver, and bronze among 8 sprinters. This is a so-called k-permutation, written P(n, k). The reasoning is identical to the factorial, except you stop early. The first medal has 8 candidates, the second 7, the third 6 — then you halt, because you only have three slots. So P(8, 3) = 8 times 7 times 6 = 336. You multiply a falling run of exactly k factors, starting from n.
There is a tidy closed form. The falling product 8 times 7 times 6 is just 8! with the tail 5 times 4 times ... times 1 = 5! divided away. In general P(n, k) = n! / (n - k)!. The (n - k)! in the denominator cancels exactly the factors you did NOT use. As a sanity check, arranging all n items means k = n, and P(n, n) = n! / 0! = n! / 1 = n! — which is exactly the factorial we already found, and the spot where the 0! = 1 convention pays off.
P(n, k) = n times (n-1) times ... times (n-k+1) [k falling factors]
= n! / (n-k)!
P(8, 3) = 8 x 7 x 6 = 336 (medals among 8 runners)
P(26, 4) = 26 x 25 x 24 x 23 = 358800 (4-letter codes, no repeats)
P(n, n) = n! / 0! = n! (arrange everything)Two honest traps
The first trap is identical items. The clean count n! assumes every item is DISTINGUISHABLE. The word HELLO has five letters, but two of them are L's, so 5! = 120 over-counts: swapping the two L's gives a string that looks exactly the same. The fix lives in the next idea up — permutations of a multiset — where you divide n! by the factorial of each repeat count. For HELLO that is 5! / 2! = 60. The lesson for now: plain P(n, k) and n! are only valid when nothing repeats.
The second trap is the one that trips up nearly everyone: confusing a permutation with a combination. The deciding question is one line long — does ORDER matter? If a reshuffle creates a NEW outcome (rankings, passwords, seatings, race orders), count permutations. If a reshuffle gives back the SAME outcome (a hand of cards, a committee, a set of lottery numbers), count combinations instead, which is just a permutation with the orderings divided away. Always ask this question first; it is the fork that the next guide explores in full, and the basis of the four sampling models.
From permutations to probabilities
Counting is not the goal in itself; it is fuel for probability. Under the classical definition, when every arrangement is equally likely, P(event) is just (favourable arrangements) / (total arrangements) — and both the top and bottom are permutation counts. Suppose four people draw straws to sit in a row and we ask: what is the chance Ann ends up first and Bob second? Total orderings of 4 people: 4! = 24. Favourable orderings: Ann and Bob are fixed in slots 1 and 2, and the other two fill the rest in 2! = 2 ways. So P = 2 / 24 = 1/12.
- Decide whether order matters. Here it does — being first is different from being second — so count permutations.
- Count the total equally-likely arrangements: all 4 people in a row, which is 4! = 24.
- Count the favourable arrangements: pin Ann to slot 1 and Bob to slot 2, then arrange the remaining 2 people, giving 2! = 2.
- Divide: P = favourable / total = 2 / 24 = 1/12. The permutation counts on top and bottom do the work.
This same machinery powers the famous birthday problem, where 'no shared birthday among k people' is counted as an ordered selection of distinct days, 365 times 364 times ... — a falling product, exactly P(365, k) in disguise. Permutations also underlie the multinomial coefficient once we start splitting items into labelled groups. So this small idea — that an ordered arrangement is a falling product of choices — keeps reappearing as we climb. Master the question 'does order matter?' and the formula writes itself.