Large gains in the St. Petersburg game (Q927117)

Let \((X_n)_{n\geq 1}\) be i.i.d.r.v., representing the player's gains in a sequence of independent repetitions of the St. Petersburg game (if \(P(X_1=2^k)=2^{-k},\;k=1,2\cdots\)), and let \(S_n=\sum_{i=1}^nX_i.\) Earlier for the so-called polynomial size gains was obtained the following result: for \(\varepsilon>0\) and \(b>1\) one has \[ \lim_{n\to \infty}\frac{\log_2P(S_n>\varepsilon x_n)}{\log_2 n}=1-b,\eqno(1) \] where \(x_n\) is any of the following sequences \(n^b,\) \((n\log_2 n)^b,\) \((n\log_2\log_2 n)^b,\) \(n\log_2 n/(\log_2\log_2 n)^{b-1},\) and \(\log_2\) stands for the logarithm to the base 2. However formula (1) does not hold when \(x_n\) increases faster than a polynomial and to give an answer to this problem is the main goal of this paper. More exactly it was proven that with the same notations as above and if \(M_n=\max\{X_1,X_2,\dots,X_n\}\) then for \(\varepsilon>0\) and \(b>1\) we have \[ \lim_{n\to \infty}\frac{\log_2P(S_n>\varepsilon b^n)}n=-\log_2 b, \;\;\;\lim_{n\to \infty}\frac{\log_2P(M_n>\varepsilon b^n)}n=-\log_2 b. \] In Introduction to this paper are given also some interpretation and application of obtained result and a bref history of St. Petersburg game's results.