Most patients, most of time do fine, despite what a doctor does. 

However, sometimes , due to skill and some good luck, a doctor save lives. 

On the flip side,  even if you are Marcus Welby,  (or Dr McCoy from Star Trek ), a doctor will lose lives. 

But this is not a zero sum game. 

Saving a life does not negate losing a life. 

In fact, saving 100 lives does not negate losing one life.  

These acts are entirely different in character to a doctor. 

The best analogy I can give? 

Landing a 747 one thousand times  perfectly will not negate one crash landing. 

This very human understanding of the practice of medicine ( and piloting 747s!)  is going to be very hard to program into Watson, because it – no matter how advanced its Deep Learning algorithm –  is going to have a tough time learning this lesson.