Also one reason a lot of teams can't do more than two options (A, B, C, D, E, F, G, etc. testing) is because you need a TON of traffic for it to be statistically significant.
Statistical significance isn't necessary for deriving value from information! The point of multi-armed bandit is that you use the best information you currently have, while also not taking that information too seriously. In a context where experiments have a cost and you need results, this makes more sense than gathering more and more data until you meet statistical significance thresholds.
Guessing based on uncertain data seems better than guessing randomly? Also if the signal to noise ratio is low my data is likely just noise, but at the same time the decision doesn't matter much. The clearer the difference the more important the decision and the more likely my data is right.
Go take a look at some solutions to the multi-armed bandit problem. A common trait is assigning value (ie. monetary) to each hypothesis. A weaker positive-outcome hypothesis is worth less. The method takes that into account.
