share

Explanation of Composite Score Player Rating System

Thu, 09/18/2008 - 1:27pm


By Jon Nichols


2008 Composite Score Rankings


The Composite Score rating system consists of two components, Offensive Composite Score (OCS) and Defensive Composite Score (DCS). Each is made up of a combination of three advanced statistical metrics and then adjusted based on a player's position and playing time. The rating system is always under construction, so the system used today may not be as good as one you could see in a few weeks.

 
 


Kevin Garnett
 

First, let me explain OCS. It is a combination of three statistics: the Offensive rating system developed by Dean Oliver (which can be found at basketball-reference.com), PER production (found on the team pages at 82games.com), and offensive adjusted plus-minus, found at basketballvalue.com. I then adjust offensive rating and PER production for position. Next, I add up the three z-scores for the three stats for each player and multiply them by 10. After adjusting for playing time (whether or not most of it is against starters or benchwarmers), Offensive Composite Score is complete.

Offensive ratings are an incredible measure of efficiency. PER also measures efficiency but takes into account players with higher usage rates. +/- is a great indicator of how well a player is fitting into his team and helping them offensively.

The average score is set at 0.

 

  • -40 and below: Terrible
  • -40 to -20: Very bad
  • -20 to 0: Below average
  • 0 to 20: Above average
  • 20 to 40: Very good
  • 40+: Elite

 

Player rank percentiles were also included to get a relative value of each player.

Next is an explanation of DCS. DCS uses the mirror statistics of the three used for OCS. Instead of Offensive rating, it uses a very similar Defensive rating. Instead of PER, it uses counterpart PER, which is basically the PER accumulated by a player's defensive matchups. Finally, instead of offensive adjusted plus-minus, I use defensive adjusted plus-minus. The scores are on the same scale as OCS.

Composite Score (CS) is a combination of OCS and DCS.

Counterpart PER, and PER were found at 82games.com. PER, or Player Efficiency Rating, was developed by John Hollinger. Offensive and defensive ratings, which were developed by Dean Oliver, were obtained from basketball-reference.com. Adjusted plus-minus can be found at basketballvalue.com. Player names and positions were obtained from dougstats.com. I used an adjustment for starters based on the formula developed at the APBRmetrics message board (http://sonicscentral.com/apbrmetrics/index.php).

Glossary of terms on stats pages:

OCS - Offensive Composite Score

OCS Rank - How a player ranks offensively among all the players in the league that qualify (at least 500 minutes played)

DCS - Defensive Composite Score

DCS Rank - How a player ranks defensively among all the players in the league that qualify

CS - Composite Score (OCS + DCS)

CS Rank - How a player ranks overall among all the players in the league that qualify

 

Comments are greatly encouraged and can be sent to [email protected]

Borko_Popic
Registered User
Joined: 06/22/2008
Posts: 1
Points: 2
Offline
This is great stuff man, I

This is great stuff man, I like how you combined the different formulas to get your composite score... I guess it definitely helps to be on a good team.

Good work

mathieu
Registered User
Joined: 07/24/2008
Posts:
Points: 0
Offline
Question

Hello,
This is an interesting piece of work.
But I have a question.

First one fact: Bruce Bowen is ranked 69 in DCS.
It is quiet obvious that he is still one of the best defensive player in the league.

I suppose it is related with the fact that he does not guard a specific position but the best player on the floor. This is also probably the reason why Ginobili is ranked #5 in DCS.

Then the fact that you have 4 players from Boston in the defensive top 7... Of course team defense is probably more important that individual skills... but 4 out of 7 seems to indicate that de DCS does not measure what it is supposed to measure.
How Paul Pierce would have ranked in DCS before the big trade?

What do you think about these issues? Is there a way to find find an accurate measurment for player defense?

I must admit I don't believe statistics can give a correct evaluation on players defense, but it is still interesting to have a bit of defense in players ranking.

Jon_Nichols
Registered User
Joined: 06/13/2008
Posts:
Points: 0
Offline
You're probably right about

You're probably right about Bruce Bowen.  The fact that he's given the toughest matchup every night while Ginobili is given an easier one could very well be the difference.  As for the Celtics being ranked high, it's a difficult problem to solve.  DCS does reward players on good defensive teams, but on the other hand, you can't punish a player for looking good on a good defensive team.  After all, good team defense requires good individual defense too.

In regards to Jarron Collins, steals and blocks don't mean everything.  In fact, getting a lot of steals could often be a sign of a bad defender who gambles too much.  

mrsaturdaynight
Registered User
Joined: 09/22/2008
Posts:
Points: 0
Offline
sort of there

you need to factor in more stuff - like the composite scores of a players team mates as well as the OCS vs DCS in particular player matchups - as with the Bowen example above the fact that he guards guys with very high OCS scores should boost his DCS score

RSS: Syndicate content