Gambling systems and multiplication-invariant measures (Q1284501)

Let \(0\leq x\leq 1\) (initial fortune), \(r>0\) (payoff ratio) and \(0<p<1\). Consider the following gambling game: choose a \(y\leq\min(x,(1-x)/r)\), and flip a biased coin with probability \(p\) for heads. If the coin lands head, then your new fortune is \(x_1=x+ry\), if it lands tail, then your new fortune is \(x_1=x-y\). Now choose \(y_2\leq\min(x_1,(1-x_1)/r)\), flip the coin and update your fortune according to the above scheme. Continue playing until you reach a fortune of 1 (in this case you win), or a fortune of 0 (in the case you lose). This strategy the authors refer to as the \textit{Bold strategy}. Let \(F_{r,p}(x)\) be the probability of winning under the Bold strategy with parameters \(r,p\) and initial fortune \(x\). Let \(\mu_{r,p}\) be the measure with distribution function \(F_{r,p}\) and define for each positive integer \(n\) the transformation \(\tau_n:[0,1]\to[0,1]\) by \(\tau_nx=nx\pmod 1\). The authors show that any weak* limit point of \(\{\mu_{2,1/2}\circ\tau_{3^m}\}\) is invariant under \(\tau_2\) and \(\tau_3\). They also (repeatedly) claim that one can probably find such a weak* limit which is not a linear combination of Lebesgue measure and an atomic measure, and thus giving a counterexample to \textit{H. Furstenberg}'s conjecture [Math. Systems Theory 1, 1-49 (1967; Zbl 0146.28502)]. The reviewer is missing any substantial justification to this claim. A natural thing to do would be to calculate the entropies of such measures w.r.t to \(\tau_2\) and \(\tau_3\); if the entropies were to be positive, then certainly these measures do not give a counterexample to Furstenberg's conjecture in view of a result by \textit{D. Rudolph} [Ergodic Theory Dyn. Syst. 10, No. 2, 395-406 (1990; Zbl 0709.28013)].