True. But I'm not measuring their performance relative to their estimates. I'm measuring performance relative to their peers and their previous performance. So

True. But I'm not measuring their performance relative to their estimates. I'm measuring performance relative to their peers and their previous performance. So giving me a padded estimate won't make much of a difference. Obviously, judgements are deeply subjective, since no two programming tasks are ever identical or even alike, but if you code a lot like, me you have a pretty good idea of what constitutes good.