Page 227 of An Introductory Course on Mathematical Game Theory by Gonzalez-Diaz, Garcia-Jurado, and Fiestras-Janeiro gives a formula for the Shapley value $\Phi$ of transferable utility games $v\in G^N$ as: $$\Phi_i(v):=\sum_{S\subset N \setminus \{i\}}\frac{|S|!(n-|S|-1)!}{n!}(v(S \cup \{i\})-v(S)).$$ It goes on to state that "therefore, in the Shapley value, each player gets a weighted average of the contributions he makes to the different coalitions." Saying that this is a weighted average of the values $(v(S \cup \{i\})-v(S))$ implies that $\sum_{S \subset N \setminus \{i\}} {|S|!(n-|S|-1)!}=n!$, but I'm curious as to why this is true.
Can someone please give a detailed proof that $\sum_{S \subset N \setminus \{i\}} {|S|!(n-|S|-1)!}=n!$ ?
Thank you.
Let $A$ be the number of ways of listing the elements of $N$.
First way: $n!$. This comes about by first choosing one number at random and writing it first. Then choosing a second number at random and writing in second, etc. There are $n!$ ways to do this.
Second way. Start by fixing $S\subset N\setminus\{i\}$. Then one way to list out the elements in $N$ is to first list out the elements in $S$ (there are $|S|!$ was to do this), then to append the number $i$, and finally to list out the remaining $n-|S|-1$ elements (there are $(n-|S|-1)!$ ways to do this). Thus, for a fixed $S$, if we insist on listing the elements within $S$ first, then there are $|S|!(n-|S|-1)!$ ways to do this. We can get all possible lists this way as we consider all possible choices of $S$: $$ \sum_{S\subset N\setminus\{i\}}|S|!(n-|S|-1)! $$ Therefore, $$ n!=\sum_{S\subset N\setminus\{i\}}|S|!(n-|S|-1)! $$