Spectrum selection is a key issue in Dynamic Spectrum Access (DSA). The purpose of the selection is to minimize interference with legacy devices and maximize the discovery of opportunities or white spaces. There are several solutions to this issue, and Reinforcement Learning algorithms are among the most successful. Through simulation, we compare the performance of the Q-Learning algorithm to our proposal which is based on an Evolution Strategy. Our proposal outperforms Q-Learning in most scenarios, and has the further advantage of not requiring any parameterization since the parameters are automatically adjusted by the algorithm.