This paper complements the analysis of Louchard and Prodinger [LP08] on the number of rounds in a coin-flipping selection algorithm that occurs in the presence of a demon. We precisely analyze a very different aspect
of the selection algorithm, using different methods of analysis. Specifically, we obtain precise descriptions of the distribution
and all moments of the number of participants ultimately selected during the execution of the algorithm. The selection algorithm is robust in at least two significant
ways. The presence of a demon allows for the precise analysis even when errors may occur between the rounds of the selection
process. (The analysis also handles the more traditional case, in which no demon is involved.) The selection algorithm can
also use either biased or unbiased coins.