A few weeks ago Stop-n-Pop linked to a piece I wrote for Hickory-High, trying to find some statistical measures to evaluate coaches. That first piece looked at the correlation between the effectiveness of each five-man unit and how many minutes they were allotted by their coach. In that analysis Adelman came out looking great, Rambis not so much.
In the comments here at CanisHoopus, there were some concerns about the volatility of the the statistic season to season, and also how well it accounted for injuries and roster changes. There were also some questions about how much information was really offerred about the abilities of coaches like Doc Rivers and Phil Jackson. They have the best players, so of course it's much easier for them to make roster decisions seemed to be part of the sentiment.
To try and answer some of those concerns I put together another simple technique for looking at how coaches distribute minutes. Here's the link to the entire piece.
For each team, I looked at all the five-man units that played at least five minutes together and calculated what percentage of those units finished the season with a positive Net Rating. I then calculated what percentage of total minutes were played by those positive units. If a coach distributed minutes completely evenly, with no regard for a lineup’s effectiveness, those two percentages would be equal. Unsurprisingly, that was not the case in any of the seasons I looked at. So to restate, we’re comparing the percentage of lineups which outscored the opposition, with the percentage of time those lineups were on the floor. For our purposes here, we’ll consider the difference between those two percentages, positive or negative, as a representation of a coach’s ability to manipulate their roster.
Since we are just looking at total percentages in the positive range, I believe this minimizes the influence of injuries and roster changes in the stat. If Kevin Love goes down with an injury it would be Rambis' job to replace those Love units with other positive units. Since the units aren't being compared to each other in effectiveness, the statistical output is only affected by a coach's ability to find some other positive replacement. The comparison here is between the amount of positive units available and the amount of minutes those positive units played. Injuries could still affect these numbers, but I think in a much smaller way.
Don't worry there are plenty of other limitations. Here are a few that I identified, feel free to point out others.
Now this method is not perfect. I think it covers some of the holes in my first analysis, while creating plenty of its own. Obviously, a coach has more responsibility for a unit’s success or failure than just deciding what players make up that unit. Not all positive lineups are created equal, and neither are all negative lineups. Also a coach’s ability to create effective lineups can always be limited by injuries and roster changes. Despite the flaws, I think there is valuable information to be gleaned.
The volatility issue hasn't been resolved, and is likely a combination of injury issues and some natural fluctuations in a coach's ability to manipulate his talent each season. Rick Adelman's Houston seasons in 2008, 2009, and 2011 all ranked in the top 23. His 2010 season was the second worst. As you would expect, Rambis' numbers are a disaster.
If anyone has them, I'd love to hear thoughts on this work, and if it seems any more valid than the first piece I wrote.