In the discussion of my recent Damian Lillard thread it was proposed that I use PER instead of game score as the response variable for regression analysis. I used game score because it is already calculated for individual games and PER is only calculated as the aggregate of all games played in a season. I wondered if my best friend Dave, a software engineer, could write a program to calculate PER by pulling data from Basketball Reference. It turns out that yes, he could. And here it is.
https://drive.google.com/folderview?id=0Bwig50PlEPuPOEFNWUtnbEpfMkU&usp=sharing
After you download and extract the ZIP, run PER Calculator.exe. Next, paste the URL of the Basketball Reference game log page for the player and year you want to calculate PERs for. Select the player's team from the dropdown menu. Enter the team's pace for the year of interest (pace numbers can be found here http://espn.go.com/nba/hollinger/teamstats) and then enter the league average pace for the year of interest. Go to the bottom of the application and click Fetch. A data table will populate and the ePER column is what you'll want to look at. ePER is estimated PER since he's used officially reported PER figures to back calculate the league average PER. This is not exactly the way PER is calculated but for all the players we've looked at, ePER is within rounding error of actual PER.
This is still a work in progress so if you encounter any errors or have any suggestions, please let me know I will pass the information along. Enjoy!