Author is confusing the map and the territory. A/B testing is comparing the performance of alternatives. Epsilon-greedy is a (good) way to do that. it's better

Author is confusing the map and the territory. A/B testing is comparing the performance of alternatives. Epsilon-greedy is a (good) way to do that. it's better than the most common approach, but you're still testing A against B.