Imagine we have a cup filled with some **unknown ratio** of white and black balls.
We select balls **one at a time with replacement** and count how many white vs. black to estimate this ratio.
Will **one extra selection** always increase the accuracy of your estimate?