Interesting article JCS, but the last time I did correlation analysis of box score stats, steals in many years had negative correlation with winning. Other years they had a small positive correlation. Overall they had a slight negative correlation.
Obviously steals are positive but the typical line of reasoning why they might turn out negative is that attempted steals have a negative value so a guy that gets three failed attempted steals for every steal could hurt his team. This aspect is, of course widely known in basketball circles. If we ever gather stats on failed attempted steals we could find the trade off point - say 1.6 failed steal attempts cancel out 1 successful steal. Going for steals is known as 'gambling on defense' and I think it's well named.
To a lesser extent the same effect shows up in shot blocking. A missed block attempt puts the player in poor defensive position so has a negative value. However, blocked shots never showed up with a negative correlation with winning. That is a league wide average so certainly some players went negative while others made up for it.
I could not follow his reasoning and there may be some nugget of information there, but I'm a mathematician and I understand what correlation analysis is saying and it says that basketball players attempt too many steals.