Comment by arethuza

Isn't A/B testing pretty much going to behave like a steepest ascent hill climbing? At each micro decision point you take what looks like the 'best' option but that means you can get stuck in local maxima?