When it comes to who actually makes it onto the basketball court on any given night in the NBA, it is not always a matter of better, but more of fit. You see this all the time in the league. If a team's rotation was decided by a one-on-one tournament, then we'd see a different list of names showing up each night in the box scores. Instead, coaches deploy the combinations that can best execute their plan on the court. Not coincidentally, the list of best fits is usually very similar to the list of best players, but not always.
You see a lot of bad analytical arguments that miss this basic point, that make a mistake of judging the component parts of a team as if it played in a vacuum. This blind spot is why stat geeks get accused of spending too much time with their noses buried in spreadsheets and not enough time actually watching the games. Of course the two activities can and should go hand in hand, because there are innumerable things that happen on the basketball court that matter and which are not counted in the same way we keep track of field goal attempts and rebounds. Mark Cuban knows this, which is why the Dallas Mavericks are one of the smartest organizations around.
A good projection and evaluation system doesn't just assign a catchall metric for player value and leave it at that. At Basketball Prospectus, our bottom-line metric is WARP and we use it all the time for all sorts of purposes. But even we recognize that so much more goes into building a roster than simply assembling the 13 highest-WARP players you can find.
So we work hard to study all of the underlying categories that make up WARP and how they interact with each other. Our projection system, SCHOENE, seeks to model the game as it's actually played on the court. The inputs to the system are basketball data and countless hours of watching, and playing, the game we all love. It's the game that drives the model, not the other way around.
Consider the case of the point guard position for the Houston Rockets. For 25.1 million obvious reasons, it's Jeremy Lin's position to lose. SCHOENE would agree, forecasting the following projected winning percentages (the per-unit companion of WARP) for the point guards at Kevin McHale's disposal: Jeremy Lin .613, Courtney Fortson .526, Toney Douglas .462, Scott Machado .446.
These are four very different kinds of players. Lin, as we all remember, is a dynamic pick-and-roll, drive-and-kick kind of a point guard who uses his ability to penetrate the lane to get to the rack, the line and to set up spot-shooting teammates. Fortson is more of a pure water bug type, more adept at getting his own shot in the paint than kicking out. Douglas, as New York Knicks fans remember, is really only a point guard in size and shape. He's basically a 3-point guy with a tendency toward streak shooting.
Machado, the undrafted assist dynamo from Iona who agreed to a partially guaranteed contract with Houston recently, is the only pure playmaker on the roster. While he projects to have the lowest overall value among the four, his standout skill might make him a more useful player than the other two players behind Lin. And, depending on what lineup McHale rolls out, there might be situations in which Machado fits better with a certain group than even Lin.