A New Proof and Extension of the Odds-Theorem

Abstract

There are $n$ independent Bernoulli random variables $I_{k}$ with parameters $p_{k}$ that are observed sequentially. We consider a generalization of the Last-Success-Problem considering $w_{k}$ positive payments if the player successfully predicts that the last "1" occurs in the variable $I_{k}$. We establish the optimal strategy and the expected profit in similar terms to the Odds-Theorem. The proof provided here is an alternative proof to the one Bruss provides in his Odds-Theorem (case $w_{i}=1$) that is even simpler and more elementary than his proof.