The thing that would actually help an underperformer improve is to teach or (even better) show them how to do a better job and the same is true for models.