The premise of the 538 analysis was that at the NBA, if you take out a player who scores a lot, you can replace him with someone who scores nearly as much. 'Scoring' is a farily fungible skill set in the NBA. At the very elite level, that's not necessarily the case, and there are some tradeoffs associated with how you score, and what kinds of scoring you get within a given system, and how your particular personnel group can score in ways that are different from another personnel group. But NBA players can score. The value of the steal is that in a league where scoring is fairly constant, you're introducing a statistic that prevents the other team from scoring on a given possession while simultaneously providing your team with a kind of scoring opportunity that is more likely to result in points (fast break, 3-on-2, free throws). That's a major value only because the ability to score is fairly constant.

But that's not the case in college. In college, scoring is not as fungible at the level of an individual team, and certainly is not fungible when you look at all of D1 college basketball the way you'd look at the NBA as a whole. If you take out a Haws (or, more pointedly, take out a Jimmer) there is a much steeper drop-off in the ability to score. What that really means is that the relative value of the steal is much, much lower in college. The value of the steal, then, is not necessarily constant across all levels of basketball, because in large part, it depends upon the value of a steal in relation to the value of scoring at that same level.

In sum, it all comes down to the fact that the spread between the best NBA player and the worst NBA player is much, much tighter than the spread between the best college player and the worst college player. For that reason alone, it doesn't make a lot of sense to apply this analysis using the same relative scoring metric.