Err... but the 'mean squared value' with statistical weight equal at all times is what is a good measure of the sinewave. With an ADC that samples at a short time interval, you get 'sum squared value' according to the statistical weight of a series of delta- functions. It might eventually converge to the same value as the desired function, but it only works in one cycle time if your ADC measurements are all taken at carefully selected times (no jitter allowed, no sample-time error relative to the phase of the sinewave).
The four-sample example was a correct Nyquist-limited sampling for the frequency of interest, and would work fine. But, it requires a 4x sample clock synchronized to the sinewave, which is not easy with a modest microprocessor ADC at 200 kHz.