I just wanted to affirm what you are doing here. A key point here is that P-Values optimize for detection of effects if you do everything right, which is not c

I just wanted to affirm what you are doing here. A key point here is that P-Values optimize for detection of effects if you do everything right, which is not common as you point out. > Thompson/multi-armed bandit optimizes for outcome over the duration of the test. Exactly.