Simplifying the problem:

When n is small, the estimation is noisy, unreliable; the bias/variance tradeoff is swamped by the estimator variance.

When n is large it does not make a difference.

nitter.moomoo.me/adad8m/status/1660696787194363906#m (https://nitter.moomoo.me/adad8m/status/1660696787194363906#m)

https://nitter.moomoo.me/nntaleb/status/1660781955896467461#m