Tuesday, October 30, 2012

Win Predictions 2012-13 Preview

Without further ado, using my simple model here are my results for the projected win totals for the season with a few surprising results:



Western conference projected standings:
Team
2012 Wins/games
2013 Wins/games
2013 Wins
Offensive efficiency
Defensive efficiency
LAL
0.621
0.720
59
4.69
2.87
SAS
0.758
0.683
56
5.36
0.77
MEM
0.621
0.646
53
1.27
3.71
LAC
0.606
0.646
53
4.89
0.32
OKC
0.712
0.634
52
2.96
1.43
DEN
0.576
0.561
46
1.49
0.46
DAL
0.545
0.537
44
-1.53
2.67
NOH
0.318
0.476
39
-0.01
-0.29
MIN
0.394
0.463
38
0.13
-1.06
UTA
0.545
0.463
38
0.26
-1.12
PHO
0.500
0.451
37
-1.23
-0.22
HOU
0.515
0.439
36
-1.43
-0.17
POR
0.424
0.366
30
-1.79
-2.03
GSW
0.348
0.354
29
-1.85
-2.49
SAC
0.333
0.341
28
-0.31
-4.41



Eastern conference projected standings:
Team
2012 Wins/games
2013 Wins/games
2013 Wins
Offensive efficiency
Defensive efficiency
MIA
0.697
0.732
60
4.53
3.44
ATL
0.606
0.610
50
2.27
1.40
CHI
0.758
0.598
49
1.92
1.13
BOS
0.591
0.585
48
0.60
1.92
IND
0.636
0.585
48
1.31
1.28
NYK
0.545
0.573
47
0.98
1.44
NJN
0.333
0.512
42
2.13
-1.77
PHI
0.530
0.500
41
-0.77
0.87
MIL
0.470
0.500
41
-0.07
0.35
TOR
0.348
0.488
40
-0.33
0.37
WAS
0.303
0.402
33
-1.67
-1.17
ORL
0.561
0.402
33
-0.21
-2.35
CLE
0.318
0.280
23
-4.96
-2.01
DET
0.379
0.256
21
-2.64
-5.18
CHA
0.106
0.195
16
-5.58
-4.37


Notes:
-Utah is where I have the most disagreement. A large part of this is because of Mo Williams' defense, but it's also because of the logjam they have in the frontcourt.
-The west now has no truly terrible teams and lots of decent but middling ones from the Nash-less suns to the poor Blazers.
-Davis' Hornets are difficult to project. Ryan Anderson reaped the benefits of playing next to Howard; it shows in his numbers. The question is, how good is he?
-Miami is head and shoulders above anyone in the east, but who's surprised?
-The Bulls will do well without much of Rose this year because of all their defensive talent ... as long as Deng and Noah are intact.
-Atlanta sans Johnson is akin to Atlanta sans Horford. You lop off one body part, and they regrow a limb.
-I don't see how losing both Brand and Iguodala can be replaced by Bynum, who will probably get injured.
-Brooklyn, I'm sorry that your defensive lapses are so severe.
-Congratulations, Charlotte, you nearly doubled  your winning percentage. But you're still last.

Championship finals:
Miami Heat in seven games over the Los Angeles Lakers

9 comments:

  1. Miami at #1 for the Western Conference!?!?!?

    Could you please update with a fix? I'm very interested to know how your model projects the Rockets with Harden!

    ReplyDelete
    Replies
    1. Sorry, I had some formatting problems. It's all fixed now.

      Delete
    2. Is Harden on the Rockets in your prediction?

      Delete
    3. Harden is. The trade helped them by about six wins, I believe.

      Delete
  2. Well, this screws up your beef with Wins Produced to their favor. The way things are going, the Wins Produced pre-season predictions of team records are pretty close to how they're actually panning out as compared to yours. Yes, I am an occasional visitor of the Wins Produced site and frankly speaking, their side of the story still seems more valid (albeit, not perfect) than most. However, I appreciate your work here, nice to read some fresh perspective.

    ReplyDelete
    Replies
    1. My issue with them is on the individual level, not the team level. Also, I didn't have much time to do this, and I screwed around with some stats to see how accurate they'd be and I'll note the issues at the end of the season.

      One big problem is that the model overvalued Nash and Howard because of their unique offensive and defensive contributions. However, injuries have left that projection looking stupid, and Pau is now a corpse. The problem with modern NBA forecasting is about injuries, not production.

      So if this method equaled the Wages of Wins method in terms of predicting team wins, the joke's on them because this site is a hobby, as I'm still a student in engineering.

      Delete
    2. Yes, I understand your point regarding individual and team evaluation. Here's a "brief" follow-up:

      In terms of team wins, box score statistics generally do provide a good fit for why a team wins or loses. Generally, a team wins/loses because: the team made shots efficiently a.k.a offense and didn't allow the opponent to make their shots efficiently a.k.a defense, the team created more shots out of their possessions a.k.a offense and allowed less shots for their opponents per possession a.k.a defense. These winning combinations are reflected in team statistics found in the box score. And with a claimed 95% correlation with TEAM wins, the Wins Produced method seems pretty valid. Like I said, on the team level, WP looks logically sound because it is based on the concept of OffRtg, DefRtg and Possessions (statistics which to me are most appropriate on the team level, not on individual players).

      However, I'm a bit skeptical on whether the weights and values derived from team statistics necessarily translate to weights and values for player statistics as well. WP may be able to calculate that, say, a certain team has 6.7 wins, but can we truly know how much of the 6.7 is "produced" by a certain player on the team? This is where I find a problem with WP as a player evaluation metric -- assigning team wins to players. On the team level, it makes sense to reward a defensive rebound with a win-value because that defensive rebound was a result of good defense on the possession. But on the player level, should the corresponding win-value of that rebound (i.e. credit for the DRbd) go solely to the player who grabbed it? The same line of questioning goes for statistics like assists and steals (maybe more), as well as shot creation (which you mention). This causes some players being inappropriately valued ("liked" or "disliked") by the metric (though I believe the number of such players is a minority).

      Regarding your predictions, I don't quite understand your method, but like I said, as it stands, WP-based team projections are much closer to current standings (which to me, is pretty much expected as I've just explained) than yours. Advanced stats and metrics-wise, I'd still prefer WP over others currently around based on "fundamental soundness" of the concepts from which the metrics are based, but I'm not completely buying it as "perfect".

      I apologize if you find the comment as long as a blog post. I'm always like this when I get excited about something.

      Delete
    3. By the way, Justin, what exactly do you mean by Nash's and Howard's "unique offensive and defensive contributions." All I can think of is Nash's defense, which is probably worse than a sack of potatoes standing in front of an offensive player (Not hating, just saying). I wouldn't call Howard's defense unique.

      Delete