This season, you'll find the results of my SCHOENE projection system for the NBA exclusively on ESPN Insider and in ESPN The Magazine's NBA preview issue. Here's some more detail on how SCHOENE works.
Player projections
The process begins with player projections. Performance over the past three seasons, weighted by age, generates a baseline projection. SCHOENE uses this baseline projection, adjusted for league averages, to find the most similar past players in 13 categories (including height and weight) at the same age (within six months in either direction). A perfect match would be a 100, while anything better than 95 indicates close similarity and a score of better than 90 shows some similarity.
For 2013-14, the best match is Anthony Morrow of the Dallas Mavericks with former guard Matt Carroll, which scored a 99.4. Two veteran guards, Derek Fisher and Steve Nash, did not have a single player with a score of 90 or better.
For most players, SCHOENE uses a pool of 50 comparable players to estimate how age will affect their performance this season. For players with unique skill sets, this group is smaller, and for a handful like Fisher and Nash, SCHOENE uses the typical aging curve rather than similar players.
In addition to using NBA performance, SCHOENE also projects players with translated statistics from the D-League, the Euroleague, EuroCup or the Spanish ACB using a similar process. The performance of rookies from the NCAA ranks is projected by translating their college performance to its NBA equivalent.
Team projections
The next step is using player projections to generate forecasts for teams. It's not quite as simple as adding up player statistics because of how players interact with each other. On offense, SCHOENE adjusts teams based on their projected ratio of assists to field goals to attempt to account for the value of passing. There is also an adjustment based on whether players are projected to use more or fewer plays than average based on their past history to account for the trade-off between usage and efficiency.
Defense is modeled using a combination of player projections and team-specific factors like shot defense, ratio of 3-pointers to 2-pointers and ability to force non-steal turnovers. Defensive rebounding is also adjusted to account for diminishing returns when combining several good rebounders (or vice versa).
Together, the projections yield a SCHOENE universe in which every point scored by one team must be allowed by another and player and team projections add up. Team win estimates are generated based on Pythagorean expectation for wins.
What's new this year?
In an effort to improve SCHOENE's predictive ability, I made several tweaks to the system this summer. First, I added a regression factor to the baseline projections for players to do a better job of filtering out fluky performances. This improved projections among all players, but especially for those who played fewer than 2,000 minutes or came to the NBA from Europe.
Using the four seasons of data in my injury database, I came up with a new model to project games missed due to injury based on games missed the past three seasons. Players are now penalized more heavily for missing extended periods multiple times, like Andrew Bogut (projected to miss 27 games), Andrew Bynum (23) and Anderson Varejao (34).
At the team level, I adjusted the team defensive factors to consider how much the roster turned over from the previous season. Teams that return a higher percentage of their minutes played are expected to retain more of their defensive characteristics from the season before, while those with heavy turnover are expected to regress to the mean. (There was already a factor regressing teams that change coaches to the mean.)
Lastly, because defensive performance gets regressed to the mean much more than offensive performance, I doubled the spread of projected defensive efficiency to try to avoid underprojecting defensive-minded teams like the Indiana Pacers and the Memphis Grizzlies, who have given SCHOENE a difficult time in the past.
Still, as usual SCHOENE has generated some surprising forecasts. You'll hear about those throughout Insider's NBA preview.
About the name
Following in the tradition of Baseball Prospectus' PECOTA projections, Football Outsiders' KUBIAK system and Puck Prospectus' VUKOTA, SCHOENE is named after a former player -- Russ Schoene, who played four seasons in the NBA in the '80s, including three with my beloved Seattle SuperSonics. Schoene is pronounced SHAY-nee, like danke schoen with an "e" on the end.