← All explorers
Loading explorer…
Read the theory: rl-basics
Read the theory: rl-basics
Four arms, one is best. Compare greedy, ε-greedy, and UCB on the same reward draws — and watch which strategies keep regret low.
Each line is the MEAN cumulative regret over 200 independent bandit problems — how much reward each strategy left on the table versus always pulling the best arm. Greedy's mean climbs highest despite winning most single runs; ε-greedy and UCB bend flatter. Set "Runs to average" to 1 to see why greedy fools you on a single problem.
Illustrative model for building intuition, not a live system.