One of the early surprises of the 2007-08 season is the Orlando Magic. After 12 games the Magic have won ten times (the first team to get to double digits in wins) and currently hold the second best record in the Eastern Conference. And by handing the Celtics their only loss of the season, there is now talk the Magic might be able to contend for a title in 2008.
How the Magic Improved
To see why all this is surprising, let’s project how many wins the Magic could have expected had each player on this team maintained his productivity from 2006-07 this season.
Table One: Projecting the Orlando Magic in 2007-08
As Table One reveals, even with the addition of Rashard Lewis, the Magic should only expect to be on pace to win about 44 games this season. Instead, as the team’s efficiency differential of 7.35 indicates, this team is actually on pace to win about 60 games. How do we explain this leap?
The answer can be seen in Table Two.
Table Two: The Orlando Magic After 12 Games
The improvement in this team can be traced to three players – Jameer Nelson, Dwight Howard, and Keith Bogans. Howard has always been an above average player, so although he’s playing better than he ever has before, we are not surprised to see him produce wins. Nelson and Bogans, though, have career marks that are below average. After three seasons, Nelson has produced 11.1 wins and posted a WP48 [Wins Produced per 48 minutes] mark of only 0.093 (average is 0.100). Bogans, after four seasons, had produced 4.1 wins with a 0.031 WP48. How did each of these players improve so much?
For an answer, let’s turn to Table Three.
Table Three: Nelson, Bogans, and Howard
This table reports the individual statistics for each player in 2006-07 and after 12 games in 2007-08. We can see that although there are improvements in multiple areas, where all three improved is with respect to rebounds and blocked shots. This suggests a team that is playing better defense (although we also see improvements on offense).
Now why did these players improve? Was it the addition of Rashard Lewis (who has not improved much at all)? Is it better coaching? Is it weak competition? Is this just a fluke from a small sample of 12 games?
At this point, it’s hard to say. But we are able to see in the data which players are responsible for the leap. So I guess that’s something (or maybe not).
The Trevor Ariza Trade
Apparently the leap forward is not enough for this franchise. Now it has sent Trevor Ariza to the Lakers for Brian Cook and Maurice Evans.
Ariza’s WP48 stands at 0.153 for his career. When we look at the 260 players who played at least 1,000 minutes last year, less than 30% posted a mark that’s this high. In sum, Ariza is a very good player.
Cook and Evans, though, are not. In four seasons Cook has produced 1.8 wins and posted a 0.021 WP48. Evans, who has also played four seasons, has a career WP48 of 0.048. In sum, the Magic traded a somewhat rare talent for two players who are decidedly below average.
The one redeeming feature of this trade is that Cook might take the minutes of Pat Garrity. In nine seasons, Garrity has produced -14.6 wins. Yes, he is one of the worst players to every log 10,000 minutes in an NBA career (maybe an entire post on this topic might be worthwhile someday).
What does this trade mean for the Lakers? That is going to have to wait for another day. For now, though, it does look like the trade makes the West a bit stronger. And the chances of the Magic truly contending with Boston in the East just a bit weaker.
– DJ
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.
Mike
November 21, 2007
DB – you ignore the fact that Ariza could never get on the floor with Lewis and Turkoglu playing essentially the same position as him. Cook is going to allow the Magic to really space the floor (not unlike that happens in San Antonio around Tim Duncan). I think what your analysis repeatedly assumes is that the sum is equal to the parts. However, this is not the case – Cook will not need to rebound or block shots b/c Howard is there to do just that. What Howard cannot do and what Cook can do with remarkable proficiency is knock down the open 3 pointer. The sum in this case is going to be greater than the parts. While I appreciate Wins Produced, your equivalent of the PER, you cannot compute wins for a team in the NBA simply by adding up the Wins Produced for each player constituting the team – there is a lot to be said about synergies and how particular players complement each other’s skills. For example I am sure that a team comprising of Howard, Duncan, Garnett, David Lee and Shaq (of 2 years back) would by your model produce 70+ wins. But do you really think that this team would produce 70+ wins? Who will guard the PG, SG or SF? Who will ever knock down a 3 pointer for this team, who will distribute the ball etc?
The trade was an excellent one for the Magic – only time will tell which one of us is right…
Rasta
November 21, 2007
As a Magic season ticket holder, I was surprised and disappointed by the trade. Last season, before he got hurt, Ariza looked like a future star. Most Magic fans believed he was more crucial to the team than Nelson or Turkoglu.
However, three things happened:
1. Ariza injured his knee, and has not regained anywhere near his prior explosiveness. Ariza had a fastbreak dunk against Boston on Sunday in which he barely cleared the rim. Compare that to the posterization of Okafor last year, and you’ll see two completely different players.
2. Stan Van Gundy replaced Brian Hill. Van Gundy’s offensive strategy is pretty simple: surround Howard with shooters, and punish the other team whenever they double-team him. Lewis, Turkoglu and Bogans have provided this. Ariza, whose range is about 15 feet, did not make a single 3-point shot in 89 games with the Magic. He’s a bad fit for Van Gundy’s offense.
3. When Battie was lost for the season, it left the Magic with very few options at PF. Giving major minutes to an inept Pat Garrity, aging Bo Outlaw or inexperienced James Augustine is a bad idea. Instead, Van Gundy has gone with Lewis and Turkoglu as his two forwards. The results have been outstanding so far.
With all that said, I guess Otis Smith was looking for another PF who can spread the floor. Cook may not be very good, but if he can keep Garrity off the floor the team should improve.
Finally, another factor with the Ariza deal might be his contract status. He can opt out next summer, and will be looking for more than his current $3 million salary.
Rasta
November 21, 2007
On another note, since somebody has to start talking about rebounds, I noticed that Rashard Lewis’s WP48 is lower than either of the past two seasons in Seattle.
WP48:
Sea 05-06 0.140
Sea 06-07 0.171
Orl 07-08 0.128
For the most part, Lewis’ stats have been consistent over this time frame. The only significant difference is shown in his vastly improved shooting percentage this season.
Effective FG%:
Sea 05-06 0.529
Sea 06-07 0.536
Orl 07-08 0.626
That’s a huge increase, offset only slightly by lower rebounding average.
Rebounds Per Game:
Sea 05-06 5.0
Sea 06-07 6.6
Orl 07-08 4.8
Part of the reason for his decline, I believe, is that Dwight Howard is grabbing every rebound in sight.
Still, we’re left with the conclusion that a small decline in rebounding is more detrimental than a large improvement in shooting efficiency.
I’ve run my “sCore” numbers for the Magic’s starting five, as shown below. (Don’t worry, as a Magic season ticket-holder, this is the only team I track this closely. I won’t clog up this site with alternative models)
Projected 07-08 (through 12 games)
Player, sCore Wins, WoW Wins
Howard , 22.8 , 26.7
Lewis , 15.6 , 8.4
Turkoglu , 7.0 , 7.4
Bogans , 8.0 , 5.6
Nelson , 5.7 , 12.0
SUBTOTAL 59.1 , 60.1
Rest of team 2.0 , 0.7
TOTAL 61.9 , 60.8
Jason
November 21, 2007
Mike, it doesn’t sound like you’ve read Dave’s book else you’d likely have couched your pronouncements differently. He discusses the synergies that many people talk about in basketball and says that he has found far, far less evidence of it than people seem to think. Player performance is not so terribly dependent on teammates as many would like to believe.
He also specifically addresses the issue of position and wouldn’t *EVER* make the claim that fielding a 5 man unit of PFs and Cs would be an acceptable lineup nor would he pronounce any potential win total drawn from their individuals scores as a consequence.
One thing to remember about models is that they are evaluated within a range of the actual variables that we see. Their performance in hypothetical situations that are unlikely to actually occur (e.g. putting 5 centers on the floor at a time) tends to be at best unpredictable, and at worst totally off the mark. When you push beyond any of the parameters seen when the model was formulated and tested, you enter a space where the results drawn may not hold.
Mike H
November 21, 2007
“Still, we’re left with the conclusion that a small decline in rebounding is more detrimental than a large improvement in shooting efficiency.”
Rasta, aren’t you comparing apples and oranges here? Effective FG% is shots made/attempted and Rebounds Per Game is just that. How can you reasonably compare the magnitude of the changes when they aren’t in the same units?
Also, what is the variability (with respect to his averages) in these measures for the given player over any n-game span? Are these changes within his typical n-game variation?
Rasta
November 21, 2007
“aren’t you comparing apples and oranges here?”
Fair enough. Let me try to make it apples to apples. Another way to look at eFG% is points from the field (or Pts – FTs) and FGs attempted.
Sea 05-06: 15.6 points on 14.8 shots
Sea 06-07: 18.0 points on 16.8 shots
Orl 07-08: 18.3 points on 14.6 shots
Just compare 05-06 and 07-08. Lewis scored 2.7 more points on 0.2 less shots attempts. On the other hand, he pulled down 0.2 less rebounds. (His other stats are pretty much unchanged).
Yet his WP48 is lower in 2007-08 than 2005-06. Why is that?
The answer, I believe, is that Lewis is treated as a PF in 2007-08, while as a SF in prior years. Same player, better stats, but lower WP48 because he’s now considered a PF.
Jason
November 21, 2007
I think the difference in Lewis’s WP has to do with the position correction. Lewis is presumed to be playing PF in Orlando. His performance, relative to the average PF is not as good as his performance relative to the average SF. Whether or not this is fair has been a subject of debate here, but the general reasoning for such a move is that if he’s playing the 4, it’s more likely that Orlando employs another SF in their lineup who will not be as statistically productive as another PF is likely to be.
Sorting out how much of the rebounding decline is due to Howard’s presence is tough. They are almost always on the floor together, so it’s tough to see if Lewis without Howard sees his rebounding rate go up. The team is a better reboounding team with Lewis in the lineup, but the counteris that there’s not a large sample with him out and almost all of that time also includes Foyle instead of Howard in the middle.
From the actual minutes played in Orlando, I think a more fair correction is something in between. It looks like Lewis is paired with a SF type almost all the time. Should Turkoglu or now that he’s arrived, Ariza (both of whom are about the same size as Rashard) get classified as SF and Lewis as the PF? In terms of the team results, it doesn’t matter. In terms of how WP regards them as individuals, it does. If we figure that when two SFs (Lewis’s “natural” position are on the court together they’re really splitting the PF assignment as it relates to replacing one of them in the lineup–there isn’t really a PF on the team to speak of), then Lewis is about what he was last year in terms of WP and also means that Turkoglu is performing at about the level he was.
None of this changes how Orlando projects and may actually mean that the projection is a bit more solid as it indicates that two of the key players are more or less on target for what they did in the past.
Lewis’s stats aren’t really better. His pure per48 production (no adjustment) last year was nearly identical (.383 in SEA, .389 this year in ORL). His increase in FG% has been offset by the decline in rebounds. You might expect him to grab more rebounds were Howard not around, but he might also face tougher defense and shoot less efficiently. That’s speculation. What’s not speculation is that his statistical contribution towards win probability has been nearly identical.
Ted Nelson
November 22, 2007
I’ve never posted here before, but I’m somewhat familiar with the site and WoW based on what I’ve seen here over the last couple years and learned from Owen at Knickerblogger.net.
Again, without knowing a lot about WoW, I would contend that Mike, above, might have a point. Individual players stats might not vary significantly based on who’s on the court, but this doesn’t necessarily mean that team play doesn’t vary based on who’s on the floor together. For example, Rashard Lewis might play the same in Orlando as in Seattle, but playing on a more well rounded team rather than one with a couple perimeter shooters some young guys and nothing else his contribution to the team may increase. Or, conversely, you could say it decreases because in Seattle he had a bigger role in what will almost definitely have been a smaller number of wins. I don’t know if that’s a good or bad example to illustrate my point, but basically, with nothing other than my personal observations to back me up, I’m saying that a well run team with players with talents suited for well defined roles will usually win more games than a team were similar players cause a sort of diminishing returns or players are just so different that they can’t coexist. I would think that on a team level you can see this statistically, but it would be hard to measure individually.
Not sure if I’ve been at all clear, but I’ve been interseted in learning more about WoW for a while now and my question is basically whether WoW feels it needs to account for players’ “roles?”
Owen
November 22, 2007
Ted – I don’t think it accounts for roles in the sense you are thinking of. It does accounts for position. Through the position adjustment, guards are compared to the historical benchmark for performance at their position, as are samll forwards and big men.
oren
November 22, 2007
“Still, we’re left with the conclusion that a small decline in rebounding is more detrimental than a large improvement in shooting efficiency.”
Actually, it’s the opposite. I think you realized this yourself, but in case you didn’t, if you look at Chart 2, you’ll see what Lewis’s Win Score for 2006-2007 would have been if he was a PF. You’ll notice that if Lewis was a PF in 2006-2007, that his Win Score would have increased this year from last year by a good 0.25 or so. As a PF, Lewis’s WS was about .103.
Despite the fact that Lewis is getting fewer rebounds, he’s still much more productive because he’s shooting better.
This would also seem to indicate that this model does control somewhat for rebounds. We know that it’s more likely that Centers and PFs get more rebounds then say a PG and SG. And therefore the position adjustment is larger for those positions.