Learn from completed experience
Monte Carlo return explorer
Same episodes. Two ways to count visits.
← Back to the chapter1. Include a completed episode
Replay three fixed examples. This illustrates averaging; it does not demonstrate convergence.
Changing γ recalculates every included episode. Reset keeps the selected discount.
No episodes included yet.
2. Inspect the latest episode
Waiting for a completed episode.
| Time | State | Next reward | Return G | Included by |
|---|---|---|---|---|
| No visits included yet. | ||||
A repeated state contributes another sample only to every-visit MC.
3. Average across all included episodes
Values are unknown until a state contributes a return.
Values are rounded to three decimals. Each sample is one selected visit's return.