It sounds exactly like A/B testing, but using a specific algorithm to determine the winner.

It talks about comparing the current situation to the best situation... But in most A/B testing, A would be the current 'best' and B would be the challenger. Same thing.

It also talks about a reward for certain button-presses, but that isn't actually what you want to optimize. You want to optimize revenue. So it's possible this could send you down the wrong path.

And if it's saying you should pit the current site against the best the site has done historically, that's ridiculous. You couldn't possibly put controls on all of the factors involved. That's why A/B testing is special: All the other factors are guaranteed to be as identical as possible.