Iterative wavefront shaping is a powerful tool to overcome medium scattering and enable focusing of diffusive light, which has exciting potentials in many applications that desire localized light delivery at depths in tissue-like complex media. Unsatisfactory performance and efficiency, however, have been a long-standing problem, and the large discrepancy between theoretical and experimental results has hindered the wide applications of the technology. Currently, most algorithms guiding the iterative search of optimal phase compensation rely heavily on randomness to achieve solution diversity. The lack of clear guidance on the new solution generation process considerably affects the optimization efficiency. Therefore, we propose a probability-based iterative algorithm that the new solutions are generated based on a probability map. With the clearer guidance provided by the probability map and less involvement of randomness, we can obtain optimization results that efficiently approach the theoretical optimal value. Moreover, with the proposed algorithm, we demonstrate higher adaptability in an unstable scattering environment and more spatially uniform optical focusing in the field of view. This study advances the state of the art in the practice of iterative wavefront shaping and can potentially inspire or open up wide applications that desire localized and enhanced optical delivery in situ.