Today’s guest post — the WoW Journal NBA Forecast for 2008-09 — is by Erich Doerr . Erich first contacted me prior to the 2006 NBA Draft with a statistical preview in hand. Each subsequent year has seen improvement in the depth and breadth of his analysis. Outside of his basketball writing, Erich does consulting work for major software products by day and has started a fledgling sports-themed Open Source software initiative by night.
As the NBA season begins, each team begins with zero wins, zero losses, and playoff dreams. Where will these teams likely be six months from now? With helpful input from Dave Berri, I have developed a rudimentary model to predict team performance with Wins Produced.
Projection Challenges
Projecting the NBA field is difficult and there are many challenges to overcome, including player development models, minute allocation methods, and coaching changes. In particular, we have a number of choices when it comes to projecting Wins Produced. Should we just use last year’s performance? How about what a player did in prior seasons? And if these seasons are used, should the seasons before last year be weighted in some fashion? Obviously the answer to these questions will impact your projections.
And then there is the whole issue of allocating minutes. How many minutes a player ultimately see depends on the decisions of his coach and the player’s health. Neither of these factors is easy to forecast.
In sum, there are a host of challenges to overcome in making these projections. Here are some details on how these challenges were addressed.
Projection Details
This year’s Wins Produced team projections are the results of a Monte Carlo analysis on the NBA schedule and team based WP numbers. This method produced three sets of Wins Produced projections:
- Using last year’s minutes played and last year’s productivity
- Using Kevin Pelton’s projected minutes played with last year’s productivity and Berri’s rookie projections
- Using Pelton’s projected minutes played with a three year weighted average of productivity and Berri’s rookie projections.
- Using Pelton’s projected player statistics
Within the Monte Carlo model, division and playoff odds are generated, and the results are summarized here:
2008-2009 Wins Produced Projections – Wins and Playoff Odds
Other Statistical Prognostications
In the spirit of sabermetrics, I have roamed basketball’s statistical analyst community and gathered other projections for the coming season. There are many diverse opinions on the coming season, even within stat-favoring prognosticators, but certain themes emerge.
NBA 2008-2009 Statistical Projections Summary
It’s important to note that there is a very high correlation between the projections of the various models. There are no two models that do not have at least a 0.80 correlation, and many have a correlation with another model in excess of 0.90.
Given the similarity in projections, it’s not surprising to see that all of the models cite Boston as the definitive top dog in the East, while the vast majority also like the Lakers out West. All expect the East’s winning percentage to improve over the West’s (.473 last season), but not overtake the lead.
The major difference between popular opinion and the stat heads would be Indiana’s forecast (this issue was noted in this forum last week). The Vegas line lists them at 35 wins, but all the included forecasts have Indian above 36 wins. In fact, one of the Wins Produced projections indicates Indiana may be a contender for a home court playoff series. By the stat-heads, Vegas’s Over/Under line is also too pessimistic on Boston and the Lakers.
At the other end of the league stands the teams that are not likely to make the playoffs. Of these, which teams are likely to top the draft boards? Last year, the top draft slots went to an injury-decimated Heat squad and a lottery lucky Chicago franchise. If these projections are to be believed, the odds should favor Memphis and Oklahoma City as the two clear cut cellar dwellers.
The prognosticators:
- Kevin Pelton is a former employee of the former Supersonics. He currently writes for BasketballProspectus.com and favors Wins Above Replacement Player as his metric of choice.
- John Hollinger brings statistical analysis to the masses via ESPN.com and ESPN’s Insider service. He created Player Efficiency Ratings (PER) and also studied economics in his undergraduate years.
- Mountain is a frequent commenter on the Wages of Wins Journal and a prominent member of the APBRmetrics forum.
- Accuscore is a budding sports statistical analysis company creating models and running Monte Carlo simulations across major sports. Their content regularly appears on ESPN.com.
- Wylie21Davis is a poster on APBR metrics and author for ArmchairGM.com.
- MikeG is a APBRmetrics poster.
Note that Accuscore and Kevin Pelton are not done releasing their season predictions at the time this post was published. I will update the projections as they become available and review the results after the season.
And finally I’d like to thank Dave Berri for his input and access to this forum. Thanks for reading and enjoy the games.
– Erich Doerr
The WoW Journal Comments Policy
Our research on the NBA was summarized HERE.
The Technical Notes at wagesofwins.com provides substantially more information on the published research behind Wins Produced and Win Score
Wins Produced, Win Score, and PAWSmin are also discussed in the following posts:
Simple Models of Player Performance
What Wins Produced Says and What It Does Not Say
Introducing PAWSmin — and a Defense of Box Score Statistics
Finally, A Guide to Evaluating Models contains useful hints on how to interpret and evaluate statistical models.
Mountain
October 29, 2008
I look forward to reading your projections Eric.
Thanks also for including us in your survey. The summary will be handy to review and I may comment more later.
I’ll add that “Wylie21Davis ” is a very versatile stat analyst and writer. If anyone is looking for either or both he’d be a good one to touch base with.
MikeG is in fact the most prolific APBRmetrics poster and regularly and generously provides carefully crafted and unique current and historical statistical analysis.
Tball
October 29, 2008
You mentioned the three teams on whom Vegas seemed pessimistic, but not the team for whom they were optimistic – OKC. What data exists for team relocations? Do NBA teams see a ‘bounce’ in the win category after moving from a high stress, last season environment to a low stress, highly supportive (typical of new cities, anyway) fan base? Had the Hornets and Grizzlies outperformed expectations in their inaugural seasons in new digs?
Evan
October 29, 2008
TBall — don’t you think there is way too small a sample size to ever know? Plus, there’s the possibility that some moves are helpful and some are hurtful.
Evan
October 29, 2008
A quick note to Erich: I think these models are great, but I’d include more of a fudge factor to increase variance so that teams don’t hit 100% likelihood of making the playoffs through a Monte Carlo simulation. Obviously no team is ever 100% to make the playoffs, when you consider possible trades, injuries, coach malfeasance in allocating player minutes, plus flatout variance (like the Bulls last year)
Anyway, I hate to start with a criticism, but it seems like you’ve probably come so close to reflecting reality that I’d offer something that may be helpful.
John W. Davis
October 29, 2008
nice work!
Erich
October 29, 2008
Mountain, thanks for the contributions, I’m still getting to know the APBR crowd.
Tball, you are right on the Vegas line on OKC. It is possible that Vegas is accounting for the emotional upgrade of a lame duck season to an enthusiastic new home arena. Also, from a numbers standpoint, Durant still has a lot of potential, some of which could be helped by playing less guard and more forward, where he has shown rebounding ability while at college.
Evan, I definitely need criticism, so don’t be shy in bringing it. Criticism will help improve my model by offering new angles and ideas.
The Monte Carlo analysis took the team projections as is, there was no random element applied to player minutes for injuries or rotation changes. This is obviously an area for improvement in the future.
I did want to note that the Wins Produced projections off the Pelton numbers will be skewed upwards for NY and Dallas and downwards for Pheonix. These skews are related to Pelton’s projected pace changes for those respective teams.
For those that are curious, Wins Produced projections for Portland don’t change much at all if Oden’s injury minutes are fully replaced by additional time for Joel Przybilla.
Serhat
October 29, 2008
Congrats on the nice post Erich!
Jason J
October 30, 2008
Great article, Erich.
The Vegas line on Boston is ludicrously low unless they were projecting a major injury.
slowjoe66
October 30, 2008
Can anybody here tell me where I can find the average win score info for each position in the NBA? I am trying to do some PAWS stuff and can’t find the “average”.
Thanks so much in advance.
Owen
October 30, 2008
I think this is what you are looking for…
https://dberri.wordpress.com/2007/06/10/game-two-thoughts/
dberri
October 30, 2008
Joe,
Here are the numbers from 1991-92 to 2007-08.
Win Score per minute is on top, Win Score per 48 minutes is the second number.
Centers
0.2290168
10.9928070
Power Forwards
0.2176455
10.4469834
Small Forwards
0.1550842
7.4440418
Shooting Guards
0.1292688
6.2049012
Point Guards
0.1337853
6.4216943
slowjoe66
October 31, 2008
Thank you. Thank you. Very much appreciated!
Joe