I am interested in Multiplicative Weights algorithm, it's randomized version of weighted majority algorithm. The idea is very simple, you have a group of experts, each expect give your advice, your algorithm should answer the question which advice to follow, at the end you given the answer if the chosen advice was the good one or not, this information algorithm uses on the next rotation.
I am looking for material on Multiplicative Weights. I had few lectures, the algorithms is very clear and well understood, but I cannot get the idea. For example here we have discussion with adversary, but I don't get why it's so important to consider adversary. I will appreciate any material, preferred video.