Given there are 365 days in a year and $n$ people, the probability that at least two people have the same birthday is given by $$\frac{365^n - 365 \cdot 364 \cdots (365-n+1)}{365^n}.$$ In this calculation, we treat the $n$ birthdays we selected as a sequence where their order matters.
If we think the $n$ dates that randomly selected as a set where the order does not matter, I believe we would have the probability $$\frac{{365+n-1 \choose n} - {365\choose n}}{{365+n-1 \choose n}}.$$
For the total number of outcomes, we model by throwing $n$ balls uniformly randomly into $365$ boxes. Essentially, there are $n$ balls and $365-1$ walls (between the boxes). For example $$ \circ |\;\;\; |\;\;\;|\circ \circ |\;\;\;|\circ |\cdots |$$ would mean one person was born on day $1$, two people were born on day $4$, one person was born on day $6$, and so on. This is how I came up with the ${365+n-1\choose n}$.
For the case that everyone has a distinct birthday, it would be ${365 \choose n}$.
What is wrong with this approach?