In mathematics, the most common type of error metric measuring deviation from an ideal. Least squares are calculated by taking the squares of all the deviations and adding them up together. Squaring the deviations emphasises improvements in large errors over deteriorations in small errors.