>Somewhere between zero and very depending on how good the person doing the interrogating is and how much the person being questioned knows about the techniques being used against them

With your "very effective" conditions, I'd argue that an innocent person would quite often respond the same as a guilty person, making the process unreliable in a different way.