Some interesting data here. Basically it looked at the WP by a player's opponent, and compared their number to the average of that team's opponents. 

http://wagesofwins.net/2011/08/11/beating-their-opponents-introducing-wins-prevented/

Some real surprises, before I discuss my reaction to the data:

First of all, Curry ranks first(!) in point guards! He's considered a substantial defensive plus (which sounds absurd, but there's a caveat coming which I'll discuss in a moment.)

Monta comes in 22 among SGs, but is also a slight defensive plus (same caveat applies - wait before it before you jump to conclusions). 

Dorell comes in 18th among SFs, and is a slight defensive plus. 

David Lee comes in 33rd among C's (they have him listed as playing 61% of his minutes there, so count him as a C) - and is a substantial defensive minus. 

The big caveat here is what those defensive numbers are measured against - and ends up not being league averages, but rather team opponent averages at that position.

In other words, Curry is a substantial defensive plus because he was significantly better than the other guys we had out there playing defense at the point guard position (some combination of Law, Lin, Monta, and Bell). 

Monta comes in as slight defensive plus because he was a little better than the other guys we had playing SG - Reggie Williams et al. 

And Lee comes in as a substantial defensive minus because he was a lot worse than the other guys we had playing C (Udoh, Biedrins et al). 

This makes some sense as a way to evaluate players defensive contributions, because it avoids punishing a player for being surrounded by poor defenders. This, however, basically makes comparisons across teams pretty pointless - eg, Rondo is clearly a much better defender, but because his team is a much better defensive team, his relative defense shows up in the stat sheet less. 

The other caveat has to do with positions. It's not clear exactly how Arturo did this, but he says he did it game by game, which means that Curry may be look better than he was by virtue of the early-season defensive switches the team sometimes employed. (Sometimes! If your response to this is along the lines of "it's BS because Monta always took the tougher assignment," go directly to jail, do not pass go, do not collect $200 - Monta sometimes, mostly early in the season, took the tougher assignment even when it was a PG. Sometimes. Not always.). 

So really the rankings are really more along the lines of, "What's the impact on a team if they lose this player?" rather than making comparisons of absolute value. He's looking for a way to evaluate, as he says "players who a better defender than you would expect given their team and/or system."

I suspect someone will look at this data and say, "It says Curry's a better defender than Howard, he's on crack." But of course it doesn't say that at all. But it's interesting that, in addition to Curry, Kevin Martin (who I generally think of as a near-Monta-level defender) scores really well here - who's backing him up?

But even with a player like Monta who ends up with a more plausible ranking, there's the issue of this data not really being consistent with the +/- data we have. eg, +/- lists Curry as a slight negative on defense, and Monta as a huge one. 

Interesting. Generally, I trust +/- more than this data because doing the position adjustment this data requires on a play-by-play basis defensively creates some rather huge room for error. Nevertheless, this is an interesting and ambitious attempt to start analyzing defense, and while there's clearly a way to go yet, it's worth looking at.