The confusion I’m trying to resolve is it feels like saving the best for last, if you have priors (assuming no attrition in the pool, which is obviously false). The intuition here is you’re using some % of the pool for calibration purposes, and should not be using your best prospects for calibration.
It would depend how far away the candidates are from each other. 5% apart, 10% apart or 1% apart.
Yes, the original problem assumes that you know nothing. If I were adapting for knowledge, I would be doing something very different and I can’t think of what on short notice because that knowledge could be very variable
The confusion I’m trying to resolve is it feels like saving the best for last, if you have priors (assuming no attrition in the pool, which is obviously false). The intuition here is you’re using some % of the pool for calibration purposes, and should not be using your best prospects for calibration.
It would depend how far away the candidates are from each other. 5% apart, 10% apart or 1% apart.
Yes, the original problem assumes that you know nothing. If I were adapting for knowledge, I would be doing something very different and I can’t think of what on short notice because that knowledge could be very variable