I'm not sure how close to exposure this is, so feel free to edit/delete/salt the earth so that Naught May Grow as appropriate.
In a FASDIU card routine I'm working on at the moment, the opening element of the main routine is a cards-across effect, actually the "Leap of Faith" Bill Malone/Paul Harris piece from The Art of Astonishment.
Now, in fitting with the style I'm working on at the moment, the routine is very hands-off, it's almost like card magic being performed by a mentalist (which I do more of)
I've already pattered about not claiming any supernormal abilities, how the real skill is with misdirection and control of perceptions, and since you can control perceptions, you can direct and control how someone is thinking. (which you can, of course, for a given value of "true")
Therefore, after the two spectators have counted their (10) cards in their hands, I ask someone else, casually, to name a number between one and five in the classic psychological force sense.
Generally they choose three, and therefore we have a miracle (three cards leap from one hand to the other), and in 30-40 performances of this I've missed three times only.
Now. My current out is to mess around with their response - so if they say "4" I'd say "no no, lower than that" etc. until I've forced 3. Since the question was off-hand, and I've talked about control, I can reiterate "See... it's all about control" to the other spectators, and they chuckle. All well and good, and we progress with the routine, still on friendly terms.
I had a thought, which is always a bad sign.
Every time it's failed (which is only thrice, which is why I mention this) they have always said "4" (and I've asked for a lower number), followed by "2" (then a higher number), and then, begrudgingly "3".
I'd love to know if this is a common experience, because if so, the routine could progress as above, but after the mirth diminishes, an index card or whatever could be produced predicting that they would choose 4, followed by 2, followed by 3... which is a little like having your cake and eating it, as you've done the sleights, forced the number, got a laugh and predicted how they were thinking.
Trouble is, with the above the sample size is going to be pretty small. I'll probably come up with something else to help me test this, but I'd thought I'd throw it open to the rest of you in the hopes that anyone else may have the same success rate.