Another way to explain the sign shift, though, is that the thrust might be zero and experimental errors come out both positive and negative.