From what I can tell, that is the only good predictor we have. See Tversky's work with the IDF or the available CTA methods for studying expertise, e.g. CDM.
Compared to what? Leetcode questions? I'd say an in depth discussion about past projects is the only way to know if someone will be a valuable future member of any team.
Compared to just having a conversation and being a good reader of people. Not everybody can do that, as not everybody is a good programmer, or designer or whatever. No silver bullet, also for recruiters.
