RSS

Accuracy of Predictions for Wins in NBA in 2010/11

23 Dec

Just before every NBA season most basketball writers, bloggers and authors try to predict what will happen in the next year which is usually fun stuff to read and discuss. But as with mock drafts I’ve always felt there’s a half of the picture missing – accuracy of those predictions.

I’m sure there were many efforts like on APBR forum which dealt with this issue locally but as usual I’m curious about the big picture which in this case means… whole basketball-oriented part of the internet.
I have many questions related to this topic but let’s start with a simple one:
which writer, blogger or author had the most accurate wins predictions last year?

To answer this question I gathered last season’s predictions from most popular sites and other googled sources [btw, it’s not a closed list so if you want to add your’s predictions let me know in the comments] and compared them to actual results by Root Mean Square Deviation which is essentially average difference between each prediction and result in terms of absolute value. If you want to play with numbers or check any single entry just download NBA Predictions Accuracy 2010-11.pdf but if you are interested only in the results here they are…

Prediction Date Downloaded from RMSE
Over/Under Line 10/20/2010 http://thepaintedarea.blogspot.com/2010/10/2010-11-nba-win[…] 6,51
John Hollinger 10/20/2010 http://thepaintedarea.blogspot.com/2010/10/2010-11-nba-win[…] 6,53
Las Vegas odds 10/25/2010 http://pregame.com/pregame_blogs/b/rj-bell/archive/2010/10/25/[…] 6,55
ESPN’s forecast 08/16/2010 http://sports.espn.go.com/nba/news/story?page=SummerForecast[…] 6,56
Accuscore 10/27/2010 http://www.opposingviews.com/i/2010-11-nba-season[…] 6,80
Joe Schaller 10/07/2010 http://high5.rivals.com/content.asp?CID=1136345 6,93
storytellerscontracts 10/05/2010 http://storytellerscontracts.wordpress.com/2010/10/05/[…] 7,10
DSMok1 10/26/2010 http://apbr.org/metrics/viewtopic.php?f=2&t=120 7,25
Roblog 10/26/2010 Archive of http://www.nerdnumbers.com/statsmackdown/ 7,69
Kelly Dwyer 09/27/2010 http://sports.yahoo.com/nba/blog/ball_dont_lie/post/The-NBA[…] 7,81
back2newbelf 10/25/2010 http://apbr.org/metrics/viewtopic.php?f=2&t=120 7,92
Sport’s Skeptic 10/26/2010 Archive of http://www.nerdnumbers.com/statsmackdown/ 8,05
Statistical +/- 10/25/2010 http://www.basketball-reference.com/blog/?p=7896 8,24
Hickory High 10/26/2010 Archive of http://www.nerdnumbers.com/statsmackdown/ 8,76
WoW Network outlook 10/28/2010 http://dberri.wordpress.com/2010/10/28/using-math-to-predict[…] 8,84
Whatifsports.com 10/25/2010 http://www.whatifsports.com/beyondtheboxscore/default.asp[…] 8,89
Win Shares 10/25/2010 http://www.basketball-reference.com/blog/?p=7896 8,92
Arturo’s S.L.S 10/26/2010 Archive of http://www.nerdnumbers.com/statsmackdown/ 9,20
schtevie 10/27/2010 http://apbr.org/metrics/viewtopic.php?f=2&t=120 9,32
Nerd Numbers 10/26/2010 Archive of http://www.nerdnumbers.com/statsmackdown/ 9,51
Kevin Pelton 10/25/2010 http://www.basketballprospectus.com/article.php?articleid=1219 9,58
arturogalletti 10/26/2010 http://arturogalletti.wordpress.com/2010/10/26/2011-nba[…] 9,66
The City 10/25/2010 http://thecity2.com/2010/10/25/2010-11-nba-final-predictions[…] 9,90
NBeh? 10/26/2010 Archive of http://www.nerdnumbers.com/statsmackdown/ 10,41
MIAHeat Index 10/26/2010 Archive of http://www.nerdnumbers.com/statsmackdown/ 10,88
SRS 10/25/2010 http://www.basketball-reference.com/blog/?p=7896 10,89
Wins in previous season - http://www.dougstats.com/09-10.html 12,08

Well, before last season those wise guys knew what were they doing… also it was a good year for ESPN, especially for our winner John Hollinger who had the most accurate wins predictions last year! Congratulations!

The most undervalued teams were Sixers [by avg of 12,32 wins], Bulls [12,24], Spurs [9,6], Grizzlies [8,95], Mavs [7,9], Pacers [6,8], Knicks [6,2] and Thunder [6]. The most overvalued teams were Cavs [17,68], Bucks [9,43], Kings [9], Jazz [7,3], Wolves [6,94], Nets [6,8] and Blazers [6,22].

Obviously injuries to players like Varejao or trades like Deron to Nets had an enormous effect here but hey, that’s an occupational hazard for people trying to predict anything ;-)

And if you are wondering… yes, I intent to make this an annual tradition so if you plan to predict something stupid, look out because you will get caught ;-)

P.S. Week later I collected even more years of data for Accuracy of Win Predictions.

 
20 Comments

Posted by on December 23, 2011 in Expanding Horizons

 

Tags: , , , , ,

20 responses to “Accuracy of Predictions for Wins in NBA in 2010/11

  1. ek

    December 23, 2011 at 07:25

    Good job. I think Hollinger is close to beating O/U in each of the last three seasons as well.

     
    • wiLQ

      December 23, 2011 at 14:59

      Thanks, but how do you know about John’s accuracy in the last 3 years?
      Did you calculate those numbers before?

       
  2. ascreamingcomesacrossthecourt

    December 24, 2011 at 13:36

    Hollinger doesn’t have a great reputation on the internet because people only focus on PER, but he’s a great judge of the NBA.

     
    • wiLQ

      December 24, 2011 at 19:15

      True, true. BTW, and I worried that my blog’s name is too long ;-)

       
  3. Crow

    December 24, 2011 at 17:47

    WoW Network outlook was about 30% off the best and got 15th place. I wonder if and how much better the new Wins Produced with the defensive rebounding adjustment would have done. Do you want to check WiLQ?

    At APBRmetrics I finished slightly ahead of DSMok1 so I guess that would put my blended metric projection in 8th place on this list, or in the upper-third.

    If I reading it right The TrueHoop Network’s Painted Area listed and may have considered the picks of Hollinger, Kevin Pelton and Kelly Dwyer but its over/under line does not appear to a simple average of them (try a average of the 3 for Boston, Atlanta, etc. and see that it does not match the over/under line) and probably is not explicitly a hybrid of them, contrary to what Henry Abbott said.

    I wonder how that 3 author equal blend actually did. Want to check it? Kevin Pelton’s metric based projections have not typically been strong over the years. At APBRmetrics his projections have finished 10th of 14, 5th of 5, and 7th of 17. If the blend did well it was probably helped more by Hollinger and by blending itself rather than being particularly helped by blending those 2 others.

     
    • wiLQ

      December 24, 2011 at 19:00

      “I wonder if and how much better the new Wins Produced with the defensive rebounding adjustment would have done. Do you want to check WiLQ?”
      Oh, no. I don’t have their tools, I don’t know how they calculated their predictions [especially regarding distribution of minutes/games] and I just don’t want to put numbers into their mouths.

      “If I reading it right The TrueHoop Network’s Painted Area listed and may have considered the picks of Hollinger, Kevin Pelton and Kelly Dwyer but its over/under line does not appear to a simple average of them”
      I agree with your point. I also interpreted it as a simple Over/Under line from Vegas at the time and then he used 3 different predictions to argue whenever [potentially and theoretically] you should bet over or under.

      And what’s the point of checking a blend of 3 authors when I measured them individually?

       
  4. Crow

    December 24, 2011 at 21:36

    Maybe one of the Wins Produced authors will check how much the defensive rebounding adjustment affected their predictive power last season themselves.

    “And what’s the point of checking a blend of 3 authors when I measured them individually?”

    To see the impact of the blending. I think it affects the end result and is not just the average of the error of the 3.

     
    • wiLQ

      December 25, 2011 at 22:57

      “To see the impact of the blending. I think it affects the end result and is not just the average of the error of the 3.”
      Wow, that’s a huge surprise for me.
      RMSE for average win predictions between Pelton, Dwyer & Hollinger was… 6,2525!
      Is it because this move hides the worst individual mistakes?

       
      • Crow

        December 26, 2011 at 03:41

        In part, yes.

         
        • Crow

          December 26, 2011 at 03:43

          If you blend good materials, there is a good prospect the average will be even better,

           
          • Crow

            December 26, 2011 at 04:03

            Dwyer had a pretty good year last season.

             
  5. Crow

    December 25, 2011 at 19:11

    All Wins Produced based metrics finished in the bottom half. RAPM, SPM and Advanced SPM (and Holinger using PER and / or all stuff) all did better, dealing with the same set of uncontrollable factors mentioned at Wages of Wins (without direct mention of the poor predictive power of WP last season).

     
    • wiLQ

      December 25, 2011 at 23:18

      “dealing with the same set of uncontrollable factors mentioned at Wages of Wins”
      That’s true but those factors affected authors very differently. For example, if you assumed great year by Varejao you would suffer bigger loss due to his injury than if you assumed average season from him etc.

       
  6. Crow

    December 25, 2011 at 19:12

    correction: Holinger using PER and / or “other” stuff

     
  7. B

    December 27, 2011 at 01:16

    You’ve actually got an error in the predictions listed for Hollinger. He picked the Pacers to win 34 games, not 31.

    http://insider.espn.go.com/nba/trainingcamp10/insider/columns/story?columnist=hollinger_john&page=PacersForecast1011

    That bumps his RMSE up to 6.46. I actually managed to get a 6.28, but unfortunately I didn’t post mine anywhere (just emailed amongst friends). Oh, well.

     
    • wiLQ

      December 28, 2011 at 00:19

      “You’ve actually got an error in the predictions listed for Hollinger. He picked the Pacers to win 34 games, not 31. ”
      That’s what you get when you try to save some time by not doing everything yourself… ;-(
      Because technically it’s not my error but blog’s thepaintedarea which was my source for this one but you are right, it’s a mistake. Thankfully it doesn’t change anything important – Hollinger was the winner anyway and correction would only make him look better.

       

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 )

Google+ photo

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

Connecting to %s