Smart is a very vague criterion. What you need is people curios and intelligent enough to be able to do independent research and It's very difficult to judge whether someone would be a good scientist before trying them with real PhD work. This process takes years so there's no way to judge how smart or successful a PhD student will be. The master's work can give a hint of how serious the student is but it's not enough because it doesn't require the same degree of independence as a PhD.