The new implementation of filtering outliers has side-effects

[ivoflipse]

Without filtering: With filtering:

For some reason the standard deviation goes completely nuts after I filter out contacts

[ivoflipse]

The problem was: force_over_time = np.zeros((len(self.forces), interpolate_length)) which is correct for as long as you fill up force_over_time with all elements in self.forces. If you start filtering them however... You end up with one or more zero-based time-series which throw off the calculations. So changing it to: force_over_time = np.zeros((len(self.forces)-len(filtered), interpolate_length)) fixed the problem.