From order to no-order: dividing out the overcount
In the previous guide you counted permutations: arrangements where order matters. The number of ways to line up k people chosen from n is P(n, k) = n! / (n - k)!, built from the multiplication principle and factorials. But very often we do not care about order at all. A handshake between Ana and Ben is the same handshake as between Ben and Ana. A committee of three is the same committee no matter what sequence we named the members. When order stops mattering, a permutation count overcounts — and a combination is what you get after you divide that overcount away.
Here is the trick made concrete. Suppose we choose 3 letters from {A, B, C, D, E}. The number of ordered arrangements is P(5, 3) = 5 * 4 * 3 = 60. But each unordered choice — say the set {A, B, C} — was counted once for each way of ordering those same 3 letters, and there are 3! = 6 such orderings (ABC, ACB, BAC, BCA, CAB, CBA). So every group of 3 got counted exactly 6 times. To get the number of groups, divide: 60 / 6 = 10. That division — permutations of the whole, divided by permutations within each group — is the entire idea of a combination.
The binomial coefficient: n choose k
The count we just discovered has a name and a symbol. The number of ways to choose k items from n, with order ignored, is the binomial coefficient, written "n choose k" and read aloud the same way. Following the divide-out-the-overcount logic, it equals P(n, k) divided by k!, which simplifies to a clean factorial formula. This single quantity is the binomial coefficient, and it shows up everywhere from card games to the binomial distribution you will meet a few rungs up.
C(n, k) = P(n, k) / k! = n! / ( k! (n - k)! ) Example: C(5, 3) = 5! / (3! 2!) = 120 / (6 * 2) = 10 Multiplicative form (fast by hand, no big factorials): C(n, k) = (n * (n-1) * ... * (n-k+1)) / (k * (k-1) * ... * 1) C(52, 5) = (52*51*50*49*48) / (5*4*3*2*1) = 2,598,960
Two facts are worth committing to memory because they catch errors fast. First, C(n, 0) = 1 and C(n, n) = 1: there is exactly one way to choose nothing, and exactly one way to choose everything (the empty set and the full set each count once). The 0! = 1 convention is precisely what makes the formula deliver these correctly. Second, the symmetry C(n, k) = C(n, n - k): choosing which k items to keep is the same act as choosing which n - k items to leave behind. So C(52, 5) = C(52, 47) without any new arithmetic.
Pascal's triangle: counting by a single recursion
Binomial coefficients are not isolated numbers; they interlock through one beautiful recursion: C(n, k) = C(n - 1, k - 1) + C(n - 1, k). The combinatorial reason is a clean either/or argument. Pick one special item, say item number n. Any k-subset either contains it or does not. The subsets that contain it must choose the remaining k - 1 from the other n - 1 items: that is C(n - 1, k - 1). The subsets that exclude it must choose all k from the other n - 1: that is C(n - 1, k). Every subset falls into exactly one bucket, so the counts add.
Stack these coefficients in rows and you get Pascal's triangle, where each entry is the sum of the two directly above it. The edges are all 1 (the C(n, 0) and C(n, n) facts), and the interior fills itself in by addition alone — no factorials required. This is more than a curiosity: it is genuinely how you would compute these numbers without overflowing, since you only ever add small integers rather than divide enormous factorials.
row n=0: 1 row n=1: 1 1 row n=2: 1 2 1 row n=3: 1 3 3 1 row n=4: 1 4 6 4 1 row n=5: 1 5 10 10 5 1 C(4,2) = C(3,1) + C(3,2) = 3 + 3 = 6 (each entry = sum of two above)
Why "binomial"? The binomial theorem
The coefficient earns its name from the binomial theorem, which expands a power of a two-term sum: (x + y)^n is a sum of terms C(n, k) x^k y^(n-k) as k runs from 0 to n. The connection to counting is not a coincidence — it is the whole reason the same numbers appear. When you multiply out (x + y)(x + y)...(x + y) with n factors, each term in the expanded product is made by picking either x or y from each factor. To get x^k y^(n-k), you must pick x from exactly k of the n factors — and the number of ways to choose those k factors is precisely C(n, k).
This viewpoint hands you identities for free. Set x = y = 1: then (1 + 1)^n = 2^n equals the sum of all C(n, k) for k from 0 to n. In words, the total number of subsets of an n-element set is 2^n, because each subset is a yes/no choice on every element — and summing the subsets by their size k reproduces 2^n. That is a small but real preview of how counting arguments and algebra reinforce each other; the same coefficients that count committees also weigh the terms of a polynomial.
Worked example, and a trap to avoid
Let us put it to work on a classic: how many 5-card poker hands contain exactly two aces? The standard 52-card deck has 4 aces and 48 non-aces, and a poker hand is unordered, so combinations are the right tool throughout.
- Choose the two aces from the four available: C(4, 2) = 6 ways.
- Choose the remaining three cards from the 48 non-aces (so the count of aces stays exactly two): C(48, 3) = 17,296 ways.
- Combine the two independent choices with the multiplication principle: 6 * 17,296 = 103,776 hands.
Notice the discipline that kept this correct. We split the task into independent sub-choices (which aces, then which non-aces), counted each with a combination, and multiplied. We did not multiply by any ordering factor, because within a hand the cards are unordered. A common slip is to mix the two regimes — for instance computing C(4, 2) and then arranging the three other cards in order — which double-counts hands that are actually identical. Keep order in or out consistently across the whole problem.