One of my favorite lines in the Wages of Wins was actually written by Allen Barra:
“Stat Nerds” they snort contemptuously at me, and probably at you, too, if you’re smart enough to have picked up this book-but the truth is that they depend as much on numbers as anyone else when it comes to making decisions. What else, after all, are you going to rely on? What, in the final analysis, are statistics but a record of what a player does when you’re not watching him? And we don’t have time to watch 99 percent of the players 99 percent of the time.
This quote captures the essence of statistical analysis. Statistical analysis is not a substitute for watching games. Statistical analysis helps you understand what you are watching.
Although proper statistical analysis is important, it’s important to note (as I noted yesterday) that statistical analysis is not a crystal ball. We can’t look at the numbers and know the future with absolute precision. In other words, just because Erich Doerr called the outcome of the NCAA championship game before the tournament started in 2008, we should not think this will happen every year (so please don’t bet your life-savings on next year’s forecast).
Erich based his forecast on a Monte Carlo simulation, which primarily employed each team’s offensive and defensive efficiency. For the NBA playoffs I am going to take the same approach. Well, I’m not going to bother with a simulation. But I am going to offer a forecast utilizing each team’s efficiency measures (and yes, I did do this a few days ago, but now we have the final numbers for the regular season).
First, let’s look at the data.
From Table One we see that the best offensives – in terms of offensive efficiency – are offered by Phoenix, Utah, and the LA Lakers. The best defensives – in terms of defensive efficiency – are in Boston, Houston, and San Antonio.
Efficiency Differential considers both efficiency measures, and hence tells us about the quality of the entire team. When we turn to efficiency differential we see that Boston, Detroit, and the LA Lakers were the best teams this season.
With this information in hand, we can now forecast the playoffs (and yes, I am a couple of days late but the regular season data hasn’t changed in that time).
Eastern Conference Forecast
Boston over Atlanta
Detroit over Philadelphia
Orlando over Toronto
Washington over Cleveland
Boston over Washington
Detroit over Orlando
Boston over Detroit
Western Conference Forecast
LA Lakers over Denver
New Orleans over Dallas
San Antonio over Phoenix
Utah over Houston
LA Lakers over Utah
New Orleans over San Antonio
LA Lakers over New Orleans
And in the NBA Finals
Boston over LA Lakers
The above forecast is based solely on each team’s efficiency differential for the season. There are two additional factors we should consider in looking towards the future.
Several teams made mid-season moves. And given these moves, the players being brought into the playoffs are not the same as the players who generated the observed differential.
For example, the Suns had a 6.2 differential before acquiring Shaq, and a 3.54 afterwards. This leads us to think that San Antonio and Phoenix – despite the first game of the series – is not as close as the differential for the season would suggest.
The Dallas Mavericks also made a major trade. Before the acquisition of Jason Kidd, Dallas had a 4.27 differential. After this trade – despite injuries to Dirk Nowitzki and Josh Howard – Dallas had a 5.99 differential. This tells us that Dallas could be favored to defeat New Orleans, as well as their potential second round opponent (either San Antonio or Phoenix).
Last year I placed third in the True Hoop Stat Geek Smackdown. Here is what I said last summer about this contest:
The similarity between Jason Kubatko’s picks and my own extended throughout the competition. In 11 out of 15 series, Kubatko and I had the same winner in the same number of games. Yes, 73% of the time we had exactly the same forecast.
Here is where we differed:
Rockets vs. Jazz: I had the Rockets in seven. Kubatko had the Rockets in five. Jazz won the series in seven.
Cavaliers vs. Wizards: I had the Cavs in four, Kubatko had the Cavs in five. Cavs won in four.
Cavaliers vs. Nets: I had the Cavs in seven, Kubatko had the Cavs in five. Cavs won in six.
Pistons vs. Bulls. I had the Bulls in six, Kubatko had the Pistons in seven. Pistons won in six.
In the end, the Pistons and Bulls were my un-doing. In discussing his victory, Kubatko revealed his methodology. Kubatko said he picked teams strictly by the numbers, which is also what I did. In other words, we each tried to ignore how it looked like a team was playing in the playoffs. Consequently we each picked the Jazz to defeat the Warriors in five.
But whereas I only considered efficiency differential, Kubatko considered both the quality of the two teams and home court advantage. And when you consider home court advantage we see that the Pistons should have been slight favorites to defeat the Bulls. So I failed because I ignored home-court advantage (which was a bit stupid on my part).
To summarize: Last year I only considered efficiency differential. Kubatko considered team quality and home court advantage. The latter factor appeared to give him the decisive edge.
If we consider home-court advantage, then I think Cleveland should be favored to defeat Washington. And New Orleans – despite the Jason Kidd trade – could be favored to defeat Dallas.
About the Smackdown
Henry Abbott graciously invited me to participate in the Smackdown this year. Unfortunately, I don’t think I am going to have the time to do this (there is also another constraint that I will explain in a few months). Although I don’t have the time to participate in this challenge, I did offer Henry my forecast of the Smackdown.
I predict that Kubatko repeats.
The best predictor of these games is efficiency differential and home court advantage. I think that’s Kubatko’s basic model (I could be wrong about that, but I think that is what he considers), therefore I think he will repeat.
Henry did say he hopes I can participate next year, and hopefully I can take him up on that invitation. Certainly I hope I will have more free time in 2009 (at least, I better have more free time in 2009).
Our research on the NBA was summarized HERE.
Wins Produced, Win Score, and PAWSmin are also discussed in the following posts:
Finally, A Guide to Evaluating Models contains useful hints on how to interpret and evaluate statistical models.