A note on Banach's match box problem (Q909332)

A certain mathematician carries one match box in his left pocket and one in his right pocket each box containing n matches. When he wants a match he selects a pocket at random. What is the distribution of R, the number of matches in one pocket at the moment when he discovers that the other pocket is empty? This is a well known problem [cf. \textit{W. Feller}, An introduction to probability Theory and its Applications. Vol. I (1968; Zbl 0155.231), p. 166] and \[ P(R=r)=\left( \begin{matrix} 2n-r\\ n\end{matrix} \right)/2^{(2n-r)},\quad r=0,1,...,n\quad with\quad mean \] \[ (*)\quad E(R)=(2n+1)\left( \begin{matrix} 2n\\ n\end{matrix} \right)/2^{(2n)}-1. \] The above results are framed in the following general setup. Suppose that the mathematician wants a match according to a Poisson process with intensity 2 and selects each pocket independently and with equal probability. Thus he wants a match from his left pocket according to a Poisson process; the same holds for the other pocket. The two processes, each with intensity 1, are independent. The waiting time until he finds the left pocket empty for the first time is \(T_ L\); \(T_ R\) is similarly defined. It is noted that the rv's \(T_ L\) and \(T_ R\) are independent and \(\Gamma (n+1,1)\)-distributed. The time between when he finds the both match boxes empty is \(D=| T_ L-T_ R|\). From the mgf of D it is shown that \(E(R)=E(D)-1\) and (*) is derived. Further it is shown that R/\(\sqrt{(2n)}=| X|\) where \(X=N(0,1)\). Using these results, approximations for the df of R are derived. A numerical example concludes the note.