Ready to Slide Rule

A deep dive into NBA statistics

Enhanced Offensive Ratings – Factoring in MORR (Miss Off Reb Rate)

New blog HoopNumbers.com recently posted Missed Offensive Rebound Rates here. (in other words, the chance that a player’s missed FG is rebounded by his own team)

One of the approximations that the Offensive Rating calculation makes is estimating the number of possessions a player gives away by missed shots leading to defensive rebounds. The estimate is based on the team’s average offensive rebounding rate. With the MORR numbers available, we can recalculate offensive rating using MORR instead of the team average.

Let’s take a look at how this changes things. For these samples I’m looking at last year’s data for players with 1600+ minutes played. In the charts below, OffRtg is the player’s Offensive Rating, OffRtgMORR is the offensive rating with MORR factored in.

First, the players that got the biggest boost:

Blake and Sweet Lou each got a bump of almost 5 points, pretty substantial.

Here are the players taking the biggest hit:

These numbers pass the laugh test for me, as I do think Okafor and Villanueva were overrated by their base Offensive Ratings.

Finally here are the players ranked by the new adjusted offensive rating (OffRtgMORR), the top two remain the same (Gasol and Paul), but things mix up somewhat after that

For spreadsheet lovers, the raw data is here.

Update: Fuller view of the raw data is here.

Advertisement

September 3, 2009 - Posted by | Data Mining | , ,

8 Comments »

  1. Recalculating offensive rating using MORR instead of the team average is a good start but why not look at each and every component of OR and see if they really all hold up as is or if other assumptions might need to be updated.

    Comment by Moreplease | September 4, 2009 | Reply

  2. MP – Not really sure what you mean.

    Also, for those without Excel, I’ve uploaded the data to google docs: http://spreadsheets.google.com/ccc?key=0AlrBc94qfCiRdFBxRnQyN0hESjViQW1rbjgtSDIwdVE&hl=en

    Comment by battaile | September 4, 2009 | Reply

  3. What I mean is that Dean Oliver’s offensive rating formula for players has some other components that might not as good as they could be good in my view or upon the review of others. I haven’t seen much critical review of the formula, like has been given to other popular formulas.

    Comment by Moreplease | September 4, 2009 | Reply

    • Ah, I see what you’re saying. To that end, I think the next step is getting the numbers for how many times a player is scoring on an assisted shot, which is another approximation the formula makes. This is probably something I will tackle in a future post, as I’m pretty sure that data is available.

      Beyond that, you’re really looking at critiquing some of the math/theory behind the Offensive Rebounding vs Assist weighting, which frankly is out of my depth. Also, there’s the ever present issue of assists being highly subjective which plagues any box-score based stat, but I don’t think that can be addressed without moving towards stats like Adjusted +/-, floor on/off etc.

      Comment by battaile | September 4, 2009 | Reply

  4. The math/theory behind the Offensive Rebounding snd Assist weighting is a unique personal opinion of how they should be weighted and in my mind is very sketchy and certainly not the last word. I won’t be using OR anymore or DR.

    Comment by MP | September 4, 2009 | Reply

  5. No, I don’t think its the last word but he does provide a rationale for how the weights are determined by his formulas that is much more compelling than any linear weighting system that I have seen. Not perfect, but the best I’ve seen so far.

    Comment by battaile | September 4, 2009 | Reply

  6. If you fix the assists (by far the sketchiest part of OR) like you did the offensive rebounding I’d use your version.

    Comment by MP | September 5, 2009 | Reply

  7. I agree about fixing the Assist Estimate.
    I’m sure the assist numbers are out there, but I would say that some of this PbP data type values are unreliable, eg. the Basketball Value PbP database are about 80 games short for the 2008/09 season, and some plays like OR are out of order – so a simple query (eg. one that picks up the “play” after a shot attempt) may not be accurate. Given the large number of events (>80,000) you aren’t going to always check.

    Also, the “weights” may need a little rejink, as it seems that the players who pull down the most OR, seem to systematically have reduced ORtg, except for those like Noah who rebound their own misses a lot.

    Comment by Dave | September 6, 2009 | Reply


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Connecting to %s

Follow

Get every new post delivered to your Inbox.