A counting argument is a model of how to build one valid outcome. The main challenge is making sure your construction includes every outcome exactly once. Lists and trees reveal that structure; multiplication and addition compress it.
Multiply stages, add disjoint cases
If each of first choices can be followed by second choices, there are complete outcomes. For example, 4 routes to a station followed by 3 routes from the station to a museum give 12 route pairs.
The multiplication principle does not require probabilities or independence. It requires the stated count of available choices at each stage. When choices are restricted by earlier decisions, update the number available.
If you separate outcomes into cases that cannot happen simultaneously, add the case counts. If cases overlap, separate or subtract the overlap before adding.
Example: a first-position restriction
How many three-digit even numbers can be made from 0, 1, 2, 3, and 4 without repeated digits?
The units digit must be 0, 2, or 4. Split into cases because choosing zero last changes the number of choices for the first digit.
If the units digit is 0, the hundreds digit has 4 choices and the tens digit then has 3. This case gives numbers.
If the units digit is 2 or 4, there are 2 choices for it. The hundreds digit cannot be zero or the used units digit, leaving 3 choices. The tens digit then has 3 choices from the remaining digits, including zero. This case gives numbers.
The cases are disjoint, so the total is A single product such as would ignore the different restriction on the hundreds digit.
Example: count the complement
A four-symbol code uses A, B, and C, with repetition allowed. How many codes contain at least one A?
There are unrestricted codes. Codes with no A use only B and C, so there are Subtract:
Counting the complement is useful because “at least one A” combines several cases: exactly one, two, three, or four As. The complement is one simple condition. The subtraction is valid because every code either contains A or does not, never both.
A table can prevent mistaken symmetry
For two dice, an outcome is an ordered pair: first die, second die. There are 36 cells in a six-by-six table. The pairs and occupy different cells even though their sums agree.
Do not assume all sums are equally likely merely because they can be listed. The sum 2 occurs in one cell, while 7 occurs in six. Counting distinct labels and counting equally likely elementary outcomes are different tasks.
A tree need not have equal branch counts
Suppose one route to a museum allows 3 return routes, while another allows only 2. A tree has 3 endpoints under the first branch and 2 under the second, for 5 trips. Multiplying 2 by 3 would invent an unavailable route. When the number of later choices depends on the earlier choice, count each branch separately or split into cases with a constant number of choices. The tree is a reasoning tool, not merely a decoration for an already chosen product.
Try it yourself
How many three-letter codes use A, B, C, D with no repetition and contain A?
Answer: There are codes overall. Excluding A leaves so 18 contain A. Alternatively, choose A's position in 3 ways, then fill the remaining positions in ways: again 18.
What to remember
Define outcomes precisely, handle special positions first, and distinguish overlapping from disjoint cases. A second counting method is an excellent check when it describes the same outcomes uniquely.