Combinations for Unordered Selections

When Does Order Not Matter?

Suppose a club needs to choose three students to serve on a committee. If the same three students are selected, the committee is unchanged whether the names were considered in the order Maya, Leo, Sam or Sam, Maya, Leo. The outcome is the group itself, not an ordered list of its members. This is an unordered selection.

Counting unordered selections is useful whenever the result is a group with no distinct positions: a committee, an unranked team, or a set of representatives. If positions or roles are distinct, such as president, treasurer, and secretary, changing who holds a role changes the outcome. That is an arrangement problem rather than an unordered selection.

The objects must be distinguishable for this count: selecting Alex, Bri, and Chen is a different committee from selecting Alex, Bri, and Devon. The selection size \(r\) is fixed, and \(r\) must be an integer from 0 through \(n\). A selection of zero objects has one possibility—the empty selection—and selecting all \(n\) objects also has one possibility.

The Combination Formula

For integers \(0\leq r\leq n\), the combination formula is

Here, \(n!\) is the factorial of \(n\), with \(n!=n(n-1)\cdots 2\cdot 1\), and \(0!=1\). The formula is the result of removing the different orders of the same selected group from a permutation count. It is the appropriate count only when each outcome is a group, rather than an ordered list.

A short counting argument clarifies the distinction. An ordered list of \(r\) distinct objects from \(n\) can be made in \(n!/(n-r)!\) ways. Each group of \(r\) objects appears in exactly \(r!\) such lists, one for every possible ordering of its members. Thus dividing the ordered count by \(r!\) leaves one count per group:

This division is valid because every group has the same \(r!\) possible orders. It is not valid to divide by \(r!\) if different outcomes have different numbers of corresponding orders; the combination formula describes selections of distinct objects with no roles or ordering.

Comparing Different Selection Sizes

For a fixed set of \(n\) available objects, changing the group size changes the number of combinations. A group that is larger by one member is not necessarily more numerous than a smaller group: the count depends on how many distinct groups can be formed at that size. Compare counts by using the formula with the same \(n\) and the relevant value of \(r\).

A useful relationship compares consecutive selection sizes. For \(0\leq r<n\), divide the number of selections of size \(r+1\) by the number of size \(r\):

In the final step, cancel \(n!\), use \((r+1)!=(r+1)r!\), and use \((n-r)!=(n-r)(n-r-1)!\). Therefore, the number of selections grows from size \(r\) to size \(r+1\) when \(n-r>r+1\); the two counts are equal when \(n-r=r+1\); and the count falls when \(n-r<r+1\). This gives a quick way to compare adjacent sizes, while the formula gives the actual counts.

Choosing a Group or Choosing Who Is Left Out

There is a useful symmetry in combination counts. Choosing \(r\) objects to include from \(n\) is equivalent to choosing the \(n-r\) objects to leave out. Each included group determines exactly one excluded group, and each excluded group determines exactly one included group. Consequently,

This identity can simplify calculations when \(r\) is close to \(n\): it may be easier to count a small group of excluded objects. It also explains why the combination counts for group sizes on opposite sides of the midpoint match.

Common Mistake: Counting Orders as Different Groups

A common error is to use a permutation count for a committee or team without distinct roles. For example, an answer of \(8\cdot7\cdot6=336\) for the three-person committee from eight members counts ordered lists: it distinguishes A-B-C from B-A-C. But those lists describe the same committee when membership alone matters. Each committee has \(3!=6\) orders, so the correct count is \(336/6=56\).

The correction is to identify exactly what makes two outcomes different before calculating. If changing the order or swapping labels leaves the selected group unchanged, use a combination. If the positions are meaningful, such as assigning three different officers, the outcomes are not merely unordered groups, and the combination count alone does not answer that question.