✅ Technical Explanation

Click here to reveal the explanation

**SampleAverage** (step size $1/N$): - ✅ **Unbiased** — in expectation, converges to the true mean. - ✅ **Full convergence** — estimates stabilize as $N \to \infty$. - ✅ **Variance shrinks** — more samples → less noise. **WeightedAverage** (constant $\alpha$): - ⚠️ **Biased** — never exactly matches the true mean. - ⚠️ **No full convergence** — estimates fluctuate forever. - ✅ **Responsive** — adapts quickly to changes in the environment. 👉 This is why SampleAverage shines in stationary bandits, while WeightedAverage is better in drifting ones.