ven a placebo has 5% chance of having statistical evidence of being better than another placebo
This is by defnition of alpha, confidence level,
which is conventionally set at 5%. This is probability
of type I error -- error of rejecting null hypotheset
(no effect).