That’s the same thing, isn’t it? Outliers don’t go well with optimizing, so you remove them first. That’s 101 of data sanitation, otherwise your model/message will over- or under-fit.