= Repeated singleton-removal allocation
Fix $n\geq2$ bins. In repeated <uniform balls-in-bins allocation>, throw each remaining ball independently into those same $n$ bins, remove every ball that is alone in its bin, and repeat with the survivors. Let $B_j$ be the number remaining before round $j+1$.
Conditional on $B_j=b\geq2$, a specified ball is alone with <probability> $(1-1/n)^{b-1}$. Its survival <probability> is therefore $1-(1-1/n)^{b-1}$, so <linearity of expectation> gives
$$
\mathbb E[B_{j+1}\mid B_j=b]=b\left[1-(1-1/n)^{b-1}\right].
$$
For $b=0$ or $1$ the next count is zero. Here $n$ denotes the fixed bin count, not the current number of balls; confusing those two quantities changes the process.
The survival <indicator random variables> are generally dependent. For two initial balls and two bins, they survive together exactly when they choose the same bin, of <probability> $1/2$ each round. Hence $\mathbb P(B_j=2)=2^{-j}$ and $\mathbb EB_j=2^{1-j}$ for $B_0=2$. In particular, replacing the conditional formula by the same nonlinear <function> of $\mathbb EB_j$ would not be justified by <linearity of expectation>.
Back to article page