Multiarmed-bandit problem, why can't we use brute force method to tackle this problem since in multiarmed-bandit problem, we can choose which arm to take and get the corresponding reward, however, why can't we conduct a lot of choice of each arm and estimate their probabilities of the reward of each arm just like brute force way?
 A: Because that would make less money.
Imagine that there are three machines. Two of them are terrible and have almost no rewards whatsoever. One of them is amazing and has a huge average reward. If your strategy is to try each one of them a huge number of times to very accurately estimate their rewards, you have wasted a huge number of pulls on the bad machines that you could have used on the good machine.
A key part of the definition of the problem is that you can only pull one lever at a time. This is why people talk about the problem as being about "the exploitation vs. exploration tradeoff". Let's say I give you 1000 pulls, and you want to maximize your overall winnings by the end of those pulls. Each pull you spend on figuring out the probabilities (i.e., exploration) is a pull you didn't spend on whichever machine you thought was the best at the time (i.e., exploitation).
A: Suppose the multiarmed-bandit problem is the game of Go. How reasonable does it sound to just try every possible move on every possible board and estimate their expected rewards?
