It avoids the disadvantage that only head and tail data are used in item-by-item subtraction.

You can use all the measurement data as much as possible, measure many times, and reduce the random error.

The calculation amount of data processing is relatively small.