Crunchy Numbers

This is a little dated -- this is the NBA, things get old in a hurry -- so bear with me. Last week the Evil Numbers Maestro John Hollinger posted his ultimate prediction of the 2009-2010 Mavericks. Using statistical modeling he takes the raw data for the season and gives his opinion regarding possible postseason success.

His opinion? Boil it down and he considers Dallas a second-tier team at best.

I won't argue any specific point. I'd get shellacked if I did that, because Hollinger is smarter than I am and has a better grasp of statistics.

However.

The bulk of his data -- roughly eighty percent of it -- is flawed. The after-trade Mavericks are a substantially different beast than the before-trade Mavericks. I don't see this version giving up an at-home blowout to the Utah Jazz, do you? And not just production, it's attitude. Granting that the Bobcats lost that game just as much as the Mavericks won it, the Mavs did win with an eighteen hour turnaround time between games and with a starting lineup that had had to play heavy minutes the night before. There would have been disappointment but no real surprise if the Mavs had simply waved the white and crawled back to the plane to get some sleep, and if this was the team still trying to figure out how to get blood out of a stone that might've happened. It didn't.

I also think Hollinger is enough of a statistician to know that. It's hard to imagine he could get a feel for where the v2.0 Mavericks are going with a sample size of only eight games. So in the end, I think he's based his prediction on what he expects to see out of the Mavericks -- an excellent team, but lacking the extraordinary factors that separate excellence from Best In The World -- and backed it up with bad numbers.

Starting an experiment with a theory already formed and ignoring results that don't support your theory . . . it's called bad science. Is there peer review for sports statisticians? Might be a question to pose at Dorkapalooza (Sloan Sports Analytics Conference).
-BJ