*BuzzerBeater.com Simulation Prediction *
Boston Celtics @ Los Angeles Lakers December 25, 2008
BOS 90 – LAL 94
By Forrest Collman and Charles Steinhardt
This will be the fourth article in a series of articles in which we attempt to use Buzzerbeater.com's advanced game engine to predict the results of a real NBA game. First a brief review of how our last prediction fared. We predicted a San Antonio win, 102-82, although we also thought the result might be closer to 102-87 given how San Antonio tends to underperform defensively early in the season. The actual final score was 107-97.
For San Antonio, we thought Tony Parker would have a big night and he did (25 points, 10 assists, 8 boards). We thought Duncan would score just under 20 points, with solid rebounding numbers. He ended up with 19 points and 7 boards. For the Raptors we pegged Jose Calderon as the likely high scorer. Calderon ended up with 16 points on 7-12 shooting, which leaves one wondering why he didn’t shoot more. Jermaine O’Neal was actually the top scorer, with 24 points on 12-20 shooting. We did not anticipate him taking this many shots. There are some rumors Toronto is interested in increasing his trade value and thus might be featuring him more offensively. We thought Bosh would score well under his average of 23 and he ended up with 17 on 6-14 shooting. We also thought the rebounding edge would favor San Antonio heavily, by an average of 10 boards. This turned out to be right on as the Spurs outworked the Raptors on the boards by a 48-38 margin. So in summary, we got many things correct, though certainly not everything.
For tonight’s matchup we tackle the classic rivalry between the Boston Celtics and the Los Angeles Lakers. This is an exciting game for us to cover for a number of reasons. First because we grew up watching Larry Bird battle Magic Johnson and listening to chants of BEAT LA! Second, Boston is a team that our standard set of assumptions does not seem to work well with, even though these same assumptions capture the basic team dynamics of most teams. In particular, our standard model fails to predict the correct scoring distribution for Boston as a team. In the standard model Ray Allen averages 27, Garnett 20, Pierce 11 and Rondo 19, when in reality the numbers are 19, 16, 18, and 12 respectively. The interpretation is that our standard coaching model believes that Ray Allen and Rondo are the best shooters on the team and should be taking more shots, where as Pierce and many of the substitutes shoot worse from the field and should therefore limit their shots.
This difference between our model and reality reflects that Boston’s offensive philosophy is very different from most teams. They look to pass first and so tend to more evenly distribute their shots amongst the players on the floor. If you look at shot attempts per 48 minutes played, the stars for Boston (Allen, Pierce, Garnett) average 18 shot attempts each. On the other hand, role players like Eddie House, Leon Powe and Tony Allen average 16 shot attempts each. In comparison, Bryant alone averages 29 shots per 48 minutes for LA and role players like Sasha Vujacic, Lamar Odom and Derek Fisher average 14 shots each.
We can retune our model to more closely match the distribution of shots Boston displays in reality, so we thought it natural to ask whether Boston performed better or worse when they distributed their shots more evenly.
Simulating the Christmas day matchup in Los Angeles with our standard model led to a 97-82 victory, with 30+ nights for Kobe Bryant and Ray Allen likely, and LA winning 90% of the time. Pierce and Garnett finished in the mid-teens, and there was little consistent pattern to Los Angeles’s supporting role in scoring.