I’ve seen a million articles comparing one bandit algo to another, but almost no one seems to have done the work of comparing an A/B test versus a set of bandit algos. You say “It would be great if there was a way to run an A/B test, but not waste time on an inferior version B (or C, D, and E…).” but have no proof or explanation of why that is so great. Can you point to sources, or provide math you’ve done youself? Would be greatly appreciated.