Williams is probably better for mathematicians coming at probability with an already solid mathematical understanding rather than practitioners who want to try to understand the underlying theory.