Projected Player Skill Index
Posted by Neil Paine on October 12, 2010
You probably didn't know this (I admit we didn't exactly advertise it... at all, really), but our Simple Projection System (SPS) has been updated to produce projected stats for the upcoming 2010-11 season. That's good, because today I'm going to use those numbers tocome up with rudimentary "skill indices" for every projected player in the database, if for no other reason than to get a broad view of what each player is good at and how they fit in on their 2011 teams.
If you're curious as to what the SPS actually is, read this first -- or just take my word that it's a very simple method of predicting future performance which meets all the basic requirements for a projection system (account for past performance, regress to the mean, adjust for aging) and nothing more. As simplistic as it is, it actually acquits itself decently against far more complex systems (granted, Basketball Prospectus later improved their method to produce slightly more accurate results than the SPS), so I think it can give us a reasonable set of baseline expectations for future player production.
The next trick is to convert those expectations into skill indices. What I mean by a "skill index" is a unitless, standardized quantity where 100 is average, >100 is above-average, and <100 is below-average -- see Doug Drinen's advanced passing stats at PFR for an example (I've also applied this technique to NBA players' foul-drawing abilities and NFL running backs' yards per carry). For the sake of reference, the resulting scale is essentially like an IQ score, making it easy to understand which numbers are good or bad. For these projected skill ratings, I'm going to run indices on the following stats: Points/36 MP, True Shooting %, Assists/36 MP, Rebounds/36 MP, Turnovers/36 MP (for which an index >100 would represent fewer TOV), Steals/36 MP, Blocks/36 MP, and non-team-adjusted Offensive & Defensive Statistical Plus/Minus.
Here were the leaders in each category:
For the ratings along with each team's current depth chart, click the link below:
October 12th, 2010 at 1:22 pm
Sooo... I can't help but notice the Miami Heat have 3 of the top 10 in projected scoring on their starting roster...
Bodes well for the rest of the league...
Add 1 in the top ten for Assists and 1 in the top 10 for Steals and well, I think we know where this is going...
October 12th, 2010 at 1:58 pm
When it calculates projected DSPM, does SPS realize that Marcus Camby is 10 years older than Dwight Howard?
October 12th, 2010 at 2:20 pm
Yeah, it's basically working off the fact that Howard is projected to take more shots, grab more ORB (greater FGA and ORB are correlated with worse defense), and get fewer steals than Camby.
Obviously the method is not going to be 100% perfect, which is why this really rudimentary DSPM regression only explains 28% of the variation in defensive performance -- by comparison, the OSPM regression explains 65% of the variation in offensive performance. But without a team adjustment, just going off the Simple-Projected per-36 minute rates? I don't think it's that bad, to be honest.
October 12th, 2010 at 7:27 pm
Is there anyone but KG that gets 100+ across the board?
October 13th, 2010 at 4:47 am
since i'm planning for the upcoming fantasy basketball season, i found the SPS kind of interesting. one thing i'm curious about is how would one go about factoring change in pace(among other things)?
i ask because i feel michael beasley's SPS is a bit conservative, because he'll playing for one of the faster offenses after 2 yrs with one of the slowest.
October 13th, 2010 at 9:25 am
#4. - No, KG is the only player who shows up above-average in everything. Love that guy.
#5. - Right, pace factor is not accounted for in the SPS (Justin is trying to keep it really simple), so to factor that in you'd have to divide the projected rates by the Heat pace LY and multiply by what you expect the T-Wolves' 2011 pace to be (I haven't done much work on predicting pace, but with Rambis the coach again, I think it's probably safe to use 2010's pace as an expectation).
October 13th, 2010 at 1:42 pm
Re:4 it's a power forward thing. DJ White, Josh McRoberts and Paul Millsap almost pull it off.
October 13th, 2010 at 2:03 pm
#1 Michael....
Doesn't that mean, baring any significant injury, that anything less than a championship this season should be viewed as a huge underachievement for Miami. After-all no team in NBA history (that I can quickly think of) had 3 of the top 10 offensive players and each offensive player was above average on defense as well (they all have DSPM scores greater than 100).
That would be unprecidented... like having the best record in the league two years in a row (60+ wins) and failing to even make the conference finals.
October 13th, 2010 at 8:43 pm
Does this take into account change in personnel? The ratings for the Heat guys seem to imply that it considers them all alpha-dogs on their own team.
October 14th, 2010 at 9:17 am
No, it makes no attempt to account for the player's 2011 teammates. Obviously in cases like Miami that's going to mean scoring is overestimated, but cases like that where it makes a big difference are rare. Besides, one cool thing about projections is that it captures a player's past demonstrated skill level, and the "true scoring skills" of James/Wade/Bosh are not going to decline just because each player's pts/36 does.
October 14th, 2010 at 12:39 pm
#10 - Good point, Neil. It gives some indication of what the amigos are capable of sans each other. Wade is the only one we'd expect to see any drop off in, right? He's going to be 29.
October 14th, 2010 at 12:53 pm
I've started work on some comprehensive aging curves. I'm getting the peak age to be 26.
October 14th, 2010 at 12:56 pm
Dwight Howard's been working with Hakeem and Pat on post moves all summer. Plus he said no more mr. nice guy. Mentally throw another 5 points on his SPMs.
October 14th, 2010 at 1:55 pm
"Neil Paine Says:
October 12th, 2010 at 2:20 pm
Obviously the method is not going to be 100% perfect, which is why this really rudimentary DSPM regression only explains 28% of the variation in defensive performance -- by comparison, the OSPM regression explains 65% of the variation in offensive performance."
No kidding. It has Golden State with one of the most elite defensive starting 5!
Still, great work. And hopefully it's on to something with Miami as the only team with 5 starters + first 2 guys off the bench all above-average defensively. But Joel Anthony is still a black hole on offense...
October 14th, 2010 at 2:11 pm
Also, remember that the SPS numbers aren't adjusted for pace or team defense, so a team like Golden State is going to be massively inflated by their tempo. I'm going to do some work in the next week, though, on this theory I have, that coaches carry with them a predictable offensive and defensive team adjustment to SPM (i.e., Gregg Popovich will always lower your DSPM by 0.7 pts/100, etc). If true, we can add a team adjustment to the raw O/DSPM numbers projected by the SPS, and come up with some predicted efficiency levels for the 2010-11 season that way.
October 14th, 2010 at 5:12 pm
James would've had a godly year had he stayed in Cleveland. I would've expected around 32 8 8 this year had he stayed.
October 14th, 2010 at 5:24 pm
@Neil "greater FGA and ORB are correlated with worse defense" I understand what you are saying here, but I don't think this makes sense to incorporate into a metric.
Players have to have certain skills from a set of valuable basketball skills, offensive and defensive. If a player has sufficiently good skills in a particular area, especially if it matches a need for a team, (yes I'm thinking of Bruce Bowen)then they get to play in the NBA, despite weakness in other areas of the game.
It is unusual for most players to have a complete skill set (see KG comment), but it seems wrong to assume that each player in the NBA is going to have some sort of average skill set - that's exactly the opposite of what occurs, teams require players with complementary skills, so often recruit players with specific abilities - Mike Miller anyone?
Further, many players have considerable value because they are typical of their position but they have excellent skills in a specific atypical area - Centers who can shoot 3s - shooting guards who rebound and block shots etc.
October 14th, 2010 at 7:13 pm
"It is unusual for most players to have a complete skill set (see KG comment),"
It's unheard of, actually. KG doesn't. If you look at the list, a lot of the top players are below average in the same thing (LeBron, for example, his only below average stat): Turnovers per 36. Those stay low when you aren't part of the offense that handles the ball.
Let's be clear: Steve Nash is dead last. Having a low turnovers per 36 isn't an indication that you are good at passing or avoiding turnovers, it's a sign that you aren't trusted with the ball in the first place.
October 19th, 2010 at 2:18 am
"Dwight Howard's been working with Hakeem and Pat on post moves all summer. Plus he said no more mr. nice guy. Mentally throw another 5 points on his SPMs."
I don't know if you're being serious, probably not but I'm not going to believe Howard has improved his skill level until he shows it in an actual regular season game. If it was so easy that one could just go to Hakeem and come away with a whole new bag of tricks, there would be a horde of Olajuwon clones because he's been teaching for awhile. I really don't understand this whole Hakeem mentor thing: so nobody else has taught Howard counter moves or how to shoot off the glass before?