After the first round of the TrueHoop Stat Geek Smackdown, I am currently holding down third place. Jeff Ma, John Hollinger, and I all correctly called seven of the first eight series. Ma and Hollinger, though, did better at calling the number of games in each series. Consequently, I am a bit off the lead.
The big story in this contest, though, is not the leader-board, but the consistency in the picks. There was complete agreement with respect to five of the first round match-ups. And in only one series – the Dallas-San Antonio match-up – did the majority number less than six. And in this case, the majority was incorrect (I was with the minority).
Across the past few days we have all been submitting our picks as the second round contestants were determined. And now that all of the second round participants have been identified, I want quickly review my picks.
Before I get to these picks, let’s review the basic methodology:
For the most part I am considering two pieces of information:
- Efficiency Differential or Offensive Efficiency minus Defensive Efficiency.
- Home-court advantage.
In general, efficiency differential is all you need to know in choosing the favorite. But if the teams are close, home-court advantage can make a difference.
With my approach outlined again, let me briefly comment on each series in the second round.
Denver Nuggets vs. Dallas Mavericks
Comments: If we focused strictly on the regular season efficiency differentials (Denver’s mark was 3.51 while Dallas finished at 2.09), and the fact Denver has home-court advantage, the Nuggets would be a clear favorite. But as I noted before, Dallas underperformed this year. Specifically, Dirk Nowitzki and Josh Howard underachieved. Had Nowitzki and Howard played as they did last year, the Mavericks would have challenged the Lakers this year. Towards the end of the season – and in the first round against the Spurs – Nowitzki and Howard returned to form. If this continues against the Nuggets, Denver is in trouble.
So this is the dilemma facing someone trying to call the series. Will Nowitzki and Howard keep playing well? And will this be enough to overcome Denver’s home-court advantage? My preference is to not pretend to know the answer to these questions. But the contest calls on me to make a call, so I picked… Denver.
I will note – despite today’s game (and I made my pick before today) – of all my second round picks it’s this series that was the hardest to call. And if Dallas does what I think they can do, I am going to be unhappy.
Pick: Denver over Dallas (4-3)
LA Lakers vs. Houston Rockets
Comments: Not really much to say here. The Lakers are a better team than the Rockets. Of course, I thought the Blazers were better than the Rockets, and that didn’t work out for me. But this time I really, really think the Rockets are not the best team.
Pick: LA Lakers over Houston Rockets (4-1)
Boston Celtics vs. Orlando Magic
Comments: If we focused strictly on the regular season efficiency differentials (Boston with a mark of 8.0 vs. Orlando at 7.1), and the fact Boston has home-court advantage, I think we would say Boston in six. But Kevin Garnett is injured. Without Garnett the Celtics struggled against Chicago. And when we look at some numbers, it’s easy to see why.
Table One: Comparing Glen Davis and Kevin Garnett
As I noted a few days ago, Glen Davis was one of the least productive players in 2008-09. Table One reveals that Davis in 2008-09 was below average with respect to shooting efficiency, rebounds, blocked shots, assists, and personal fouls. When we look at his 2009 post-season performance, Davis is still below average with respect to shooting efficiency, rebounds, and blocked shots.
Of course, the below average conclusion comes from a comparison between Davis and an average power forward. What if we compare Davis to Garnett? As Table One notes, with respect to almost every stat the Celtics suffer when they move from KG to Big Baby. And I think this is why the Celtics will suffer at the hands of the Magic.
Let me add one more observation. Marcin Gortat’s performance against Philadelphia in the decisive Game Six might be just enough to convince the Magic to give Gortat more playing time (for more on Gortat see HERE and HERE). If that happens, Orlando should fare even better in the second round.
At the moment, though, I am going to assume that Dwight Howard will avoid suspensions and the Magic will not be playing Gortat much. So although I still favor Orlando, I think Boston’s home-court advantage will keep it close.
Pick: Orlando over Boston (4-3)
Cleveland Cavaliers vs. Atlanta Hawks
Comments: This appears to be the biggest mismatch of the second round. It’s possible the Hawks can win a game, but it’s very unlikely the Hawks could win the series.
Pick: Cleveland over Atlanta (4-0)
And I am still not sure who I will pick as the eventual champion (although I will have to decide soon).
- DJ
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.
DB, in picking the series using home-court advantage you seem to use a static home-court, but isn’t it possible that home court advantage really is variable among teams? I’m mainly thinking about Atlanta. At home they were a 62-20 team this year, and on the road Cleveland as a 54-28 team this year. That seems to me to say Atlanta has a pretty damn good shot of winning at least half their home games this series, and almost no shot of winning in Cleveland. So 4-1 or maybe 4-2 seems to make more sense to me.
This is obviously very simplified analysis, but the Atlanta Home/Road differential has been very large for the past two years.
Doing well so far!
One question though, given the supposed superior predictive power of wp48 why do you use efficiency differentials to make these predictions?
People always bang on about the predictive power of the wp model, yet when it actually comes to making predictions, you don’t use it?!
JAW
using a different homecourt advantage for certain teams worsens the predictive power of point differential (at least in the regular season). It sure seems counterintuitive but that’s how it seems to be.
@berri
I think you’re forgetting to take series format into account.
Things like “I think we would say Boston in six”-sounds weird in my head, it would mean Boston would have to win game 6 in Orlando.
Then, picking Orlando in 7 means
a) Orlando was behind 3-2, wins two games in a row, including a game 7 in Boston or
b) Orlando was 3-2 in front, Boston wins in Orlando and Orlando then wins in Boston
Scenario a) is probable i guess(b), not so much), but I think if you go with picking the non homecourt team in a 2-2-1-1-1 format you have to go with “Orlando in 6″
(just run a simultion 1000 playoff series where you give the hometeam a 60%-win% with a 2-2-1-1-1 format)
Michael,
Please read how WP is created. At the team level (which is what matters here), efficiency differential and WP are the same. This is because WP is derived from efficiency differential.
123,
I think you are correct. In making these picks (which I don’t spend much time on), I just go off what the simple model I have suggests. So I am ignoring sequencing. But I think you are correct that Orlando in 6 makes more sense. Unfortunately, it is too late to change my mind.
Michael,
If you recall an article on this blog earlier this year, and I think you commented on it,the relationship between team stats and wins in basketball is very strong. The efficiency differential will explain most of the wins in basketball, even more so than the run differential in baseball does. And this is something people can agree on. WP is basically a “slice” of that team efficiency differential pie attributed to a player.
I would love to see an article on who beat the Trailblazers. I have a couple theories just by looking at the playoff number over a bbreference.
What I say from the 5-10 minute look I took with the numbers were very high t.o percentages by most of Portland’s players(Roy, Aldridge, and Outlaw were o.k). And two of those guys(Aldridge, outlaw) were much worse than there regular season efficiency(offensively). Outlaw was a terrible at 79(39% shooting was the culprit) while Aldridge was 11 points lower than his regular season number at 104(only had 20 fta on 104 shot attempts and below average for big man 49% shooting).
Basically, Portland got killed by their own turnovers because most of the guys had efg% above 50 with the two low guys being Roy who got the ft line so much he made up for it and Aldridge at 49. And there was Outlaw with 34%. Everyone else who played major minutes put the ball in the basket when they got shots off.
123,
I couldn’t find point-differential home/away last night when I was looking, but supposing we looked at that and then regressed to the mean a bit, I think we might find some meaningful differences in the extent of a home-court advantage.
Dave, what’s your take on LeBron being awarded MVP?
You can derive caramel from corn but that doesn’t mean you should eat caramel cakes for roughage!
My point is, if wp48 (the player level) is the same as efficiency differential, and you already have wp48 numbers for every player, then you should be able to accurately predict team wins using the wp scores for each player!?
Michael, I think it’s just a lot easier to use point differential.
What would be interesting, however, would be to look at how minutes played in the playoffs differs from regular season minutes played, and the concentration of those minutes, which may give WP48 more predictive power than team differential in the regular season. This would be especially true when teams coast at the end of a season having locked up a spot.
A team that has a short bench may look much better in the playoffs if we extrapolate from wp48 than from point differential. Houston, for example, relied much more on their starters in the first round than they had in the regular season, whereas Portland did not shorten their bench. This could be hard to predict, ex ante, but I knew my roommate said before the series that he expected Portland to struggle because much of their value was in their bench.
Perhaps there could be a way to predict which teams use their starters significantly more in the playoffs than the regular season by looking at the differential of the WP48 btw the starters and the bench.
Michael,
I think you’re getting confused. Caramel from corn is processed, not cut.
WP is a way to evaluate each player’s contribution toward win in a season, or in other words (darn it, maybe I should stop using this analogy) it’s a way to decide how big each player’s slice from that big pie should be.
So if you add all the slices then you end up with…the whole pie. But why would you want to put the pie back from pieces when you already have the whole thing ready? Here dberri is trying to predict which team will likely win, not worthiness of individual players.
assumption: infinitely sharp&frictionless knife was used to cut the pie.
Ah come on JAW, you and your succinctness makes me look bad. “I think it’s just a lot easier to use point differential.” I should’ve just written that one line too.
Jaw, I think it’s that WP48 does not have the predictive power of points scored per 100 possessions minus points given up per 100 possessions. I think that is why when you predict wp from paws, the total wins produced for each team is not particularly accurate. (At least it wasn’t for the 30 games or so that I did it at the beginning of the season, mostly it overestimated the number of wins each team would have.) On the other hand when wp is derived from ED it is obviously much more accurate since ED is such a strong predictor of team wins (I mean of course the team which scores more points and gives up fewer points than the other team is likely to be better.) The obvious implication of this being that using predictive power to prove that the way in which the stat is weighted at the player level (wp48) is correct is not necessarily a valid approach.
“I think you’re getting confused. Caramel from corn is processed, not cut.”
Simon, I said derived, i.e formed or developed from something else. It was the correct usage not sure where you got ‘cut’ from.
I actually think that if you had time, you could effectively use wp48 instead of differential, because teams definitely use starters much more in the playoffs than in the regular season. A frequently talked about example of this is Lebron James. In a close playoff game, he can play 48 minutes, often at the expense of someone terrible like sasha pavlovic or wally szczerbiak. A team with a great bench and mediocre starters isn’t likely to be much better in the playoffs, if at all, but a team with amazing starters and a so-so bench can do better by playing the good players more minutes. If you could somehow predict playoff minutes and you know everyone’s WP48, you can predict whether a team is likely to be better or worse compared to the regular season.
Not that I think the effect from this would be particularly large, but it’s worth thinking about.
Michael, what I tried to imply was that at the team level they are the same, thus my comment of “cut” vs “process,” since caramel is transformed substance that cannot be reversed back to corn where as slices of pies can be…ok you get my gist. Yes that was a dumb analogy on my part although I still think it kinda makes sense.
I haven’t seen everyone’s end-of -season WP48 numbers yet, but I’ll go out on a limb and guess the Lakers have the deepest team in terms of number of players with positive measures. If so, the ‘rotation gets smaller’ theory could suggest that LA will benefit less from that tendency. However, there are two additional conclusions one might draw: 1) LA is also better constructed to withstand injuries, as more of their replacements are valuable, and 2) if the best players get the bulk of the minutes, LA has a chance to be even better now. (Should we begin to evaluate coaches at least partly on their ability to recognize, and ride, the most productive players? Does anyone have a model for this?) My gut still says Cleveland wins it all, but do the numbers, when all is said and done?
How bout those rockets, Prof Berri?
What a great way to open your series against the West’s No.1 seed.
The Rockets make great use of several “very good” role players. However, they don’t have a high usage reasonably efficient scorer other than Yao. That’s going to catch up with them.
You can count on role players to lock down the other team and score some points very efficiently, but you can’t count on them to score enough to consistently win very high level professional basketball games.
As overrated as McGrady is (and he’s wildly overrated), they need someone like him (preferably a little more efficient) that can score in bunches, open up the floor, and take the heat off less talented and skilled offensive role players.
“As overrated as McGrady is (and he’s wildly overrated)”.
By this I am talking about now, but he was always overrated.
simon said in his comment that WP is a slice of the team efficiency differential.
Is that to be taken exactly?
If that’s true, does it make sense to evaluate WP as a model by comparing team WP-sums to team wins? Wouldn’t that be just comparing team efficiency differential to team wins, a relationship that has already been well established?