White RoomNEW

Point Nine Two Five

A bandit agent has 4 arms and uses ε\varepsilon-greedy action selection with ε=0.1\varepsilon = 0.1. On the exploration branch it samples uniformly over all 4 arms, including the greedy one. Exactly one arm currently has the highest estimated value.

What is the probability that the agent selects that greedy arm on a given step?