This proof of the fact that $n \choose r$ = $n-1 \choose r-1$ + $n-1 \choose r$ is correct but very baffling to me.
Proof: Let X be an n-set, and suppose x is a chosen member of X. The set of all r-subsets of X can be split as follows:
U= r-subsets which contain x.
V= r-subsets which do not contain x.
An r-subset is in U if when x is removed from it we obtain an (r-1)-subset of the (n-1)-set X-{x}. So:
|U|= $n-1 \choose r-1$
On the other hand, an r-subset is in V if it is an r-subset of the (n-1)-set X-{x}. So:
|V|= $n-1 \choose r$.
Addition of U and V gives the result.
But: Surely, |U|= $n-1 \choose r-1$ is only the case when we remove x. How can we remove x without changing the result?