Predicting the Finals With SPM
Posted by Neil Paine on June 1, 2009
In the final installment of the season, here's your official Statistical +/- prediction for the 2009 Finals:
West #1 Los Angeles Lakers vs. East #3 Orlando Magic
|Lakers in 4||10.6%|
|Magic in 4||2.8%|
|Lakers in 5||15.9%|
|Magic in 5||8.8%|
|Lakers in 6||23.5%|
|Magic in 6||8.8%|
|Lakers in 7||20.0%|
|Magic in 7||9.6%|
|Lakers win series||70.1%|
|Magic win series||29.9%|
One caveat is that SPM did not get a good read on the Magic in the East Finals -- while it fairly accurately estimated their odds at 53-54% against the Celtics, it greatly underestimated their chances vs. Cleveland, pegging their probability of beating the Cavs at just 18%. Then again, the 2009 Eastern Conference Finals may go down as one of the shining examples of point differential's failure to effectively predict the future under certain specific circumstances.
As John Hollinger wrote after Game 1 of the Magic-Cavs series, and as Orlando proved time and again over the subsequent 2 weeks, the Magic posed a number of significant matchup issues for the Cavaliers. What was, in retrospect, obvious during the regular season proved incontrovertible during the conference finals: the Magic were basically tailor-made to exploit Cleveland's deficiencies, with the deadly combination of Dwight Howard's inside presence and a host of expert 3-point marksmen on the perimeter. That LeBron James' supporting cast had a difficult go of things against Orlando's top-ranked D was an added bonus for the Magic, who may very well have won the series even if, say, Mo Williams hadn't shot an anemic 37% from the floor. Their advantages on offense were just too much for the Cavs to deal with, especially down the stretch in a number of games. But point differential/pythagoras saw none of this coming. Cleveland had the league's best differential, its best SRS, and yet the matchups posed by the supposedly inferior Magic spelled their doom, numbers be damned.
All of which is to say, it's not an exact science. Point differential (specifically one adjusted for strength of schedule like SRS) remains the best predictor of future performance, but it's not bulletproof -- or else we'd have all made a killing in Vegas a long time ago. Sometimes you can identify a matchup that seems significant before a series begins, but it turns out to be a non-factor. And sometimes the stats are overwhelmed by matchups. This recent Cavs-Magic series happened to be a very glaring example of the latter. I'm still going with the numbers over the next few weeks, though: Lakers prevail in 6.
June 1st, 2009 at 9:10 am
I wouldn't be so quick to dump the data. I mean, they had an 18% chance! Sure, Orlando looked pretty dominant, but what if mo williams shot 50%, and Orlando missed a bunch of 3s? Different series!
Anyway, I think it's perfectly plausible that we just saw an unlikely outcome not that the numbers were wrong in the first place.
June 1st, 2009 at 12:23 pm
"Anyway, I think it’s perfectly plausible that we just saw an unlikely outcome not that the numbers were wrong in the first place."
I don't. Not when Orlando has now won 12 of the last 17 games against Cleveland. That's not an unlikely outcome. That's a trend. It would seem like sound reasoning to question the consistency of the three-point shot, but there's a huge difference between a contested three and an open one. Orlando wasn't shooting well for no reason. They were shooting well because they were getting good looks. Why? Because Cleveland simply could not guard Howard one-on-one and because Orlando had plenty of guys who could dribble penetrate and break down Cleveland's defense. As for Mo Williams, he has struggled against the Magic all season, so maybe his shooting 50% will only ever be the exception against them.
The numbers once again go against Orlando, but I bet they went against the Houston Rockets in 1995 as well. Just last season everyone picked the Lakers to beat the Celtics, so I'm not surprised that everyone has the Lakers winning this time as well. If the Lakers win, I believe it'll be because of home court, although Orlando already defied history by stomping Boston in Game 7 at the garden. Orlando should give L.A. a harder time than everyone thinks. For one thing, L.A. has done a poor job of defending the three all season, so the 40+ percent mark that some writers have already brought up (which is how well Orlando shot the three against L.A. this season) may not be an aberration, but a result of the Lakers's questionable three-point defense. Let us not forget that Houston went up 30 on the Lakers by hitting open threes in game four. Also, Orlando is a better defensive team than any L.A. has faced thus far, which should at least keep Kobe from averaging 36 ppg like he did against Denver. Anyways, I think that Bynum will be something of an X-factor (he hasn't had much of an impact so far this postseason) and it'll be interesting to see who gets the better of the Lewis/Odom matchup. Those things could swing the series one way or the other.
The only way this series goes to the Lakers as easily as the numbers would suggest is if Howard finds himself in foul trouble more often than not. Here's hoping the refs swallow the whistle and let the players play. (and if not, they need to at least be consistent on both ends)
Oh, and Orlando did beat L.A. both times they met this season (something SPM doesn't take into account), even if they were closely contested games. Also, I believe Orlando went up BIG in the game in Orlando before giving up their lead, so I'm sure Orlando has a few matchups to exploit against L.A., even if Nelson doesn't play.
June 1st, 2009 at 1:24 pm
Hmm, it seems like my memory might've been a bit fuzzy (and off) on that first game in Orlando. It looks like Orlando had to pull a bit of a gutsy comeback. Still, Orlando can play with the Lakers. It should be a much more competitive series than the numbers would indicate.
June 1st, 2009 at 10:04 pm
One other data point that might be relevant is record vs good teams.
If you look at record vs top 10 teams the teams that come out the best are LA and Orlando.
It's reasonable to think that how a team performed against high level competition is a more relevant predictor of future success vs high level competition than simply looking at how they performed vs good, average and bad teams.
Another thing to look at is how Lebron's offensive efficiency precipitously declines against elite defenses (which is something you already did). Looking at the Cav's Stat +/- it was all Lebron. Everyone else was either average or below average. If suddenly you take Lebron from about +12 to something more like +7 then maybe the math looks different. Maybe.
June 2nd, 2009 at 12:51 pm
I think Jose is pointing out a factor that is true in basketball in a large sense regarding matchups which makes predictors difficult to rely upon as the field slims down and might really separate the truly dominant teams from the very good teams. Cleveland was built to beat Boston not Orlando, and Boston and Orlando are very different animals. However, that being said, a truly elite team ought to have the pieces to adjust to or overcome just about anything. I'm not sure this is a case of the numbers being unreliable so much as the Cavs being a team of gritty focused role players who don't have the ability to adapt on the fly.
If you look at some of the teams the last major dynasties had to face to win it all -- Chicago was built to beat a physical, defensive-oriented team with great perimeter scoring and rebounding (the 1990 pistons), but over the course of their reign they had to beat center dominant defensive teams like New York and Miami, super versatile scoring machines like the Blazers, Suns, and Lakers, well balanced squads like the Cavs and Pacers, star tandems surrounded by shooters and defenders like Shaq & Penny, Payton & Kemp, and Stockton & Malone,... And they had the versatility to handle them all. Up tempo, pressure d, lockdown, switching man to man d, high post action, drive and kick offense - they were able to change their approach when necessary. They played against opponents' weaknesses and won and won and won.
On the other hand you look at the Lakers 3peat w/ Shaquille, and they beat a lot of very different teams in just one way. Play tight perimeter d. Clog the paint with a big center. Get the ball into the low post. Cut to occupy help defenders. Make open jumpers. Win and win and win.
For the most part I believe (though I don't really recall now) that the SPM predictive scores for those past champs were good indicators - except one year in LA where I think they had some injuries during the season that cleared up in the playoffs (also Shaq had a habit of starting slowly because he'd show up to camp out of shape and get better as the season progressed).
Cleveland was the class of the league this season, but do they have the tools for either of those approaches? Maybe not. LeBron's really the only player you can ask to radically change his style of play and expect him to perform at a high level, so you wouldn't call them especially versatile. They're also not really overpowering at one particular thing unless LeBron personally has it going at warp-factor vitamin water. So while they might have the same dominant type of SPM ranking as great teams, they may just not be built to live up to those numbers, whereas other great regular season teams could do so.