What exactly is PAA? The easiest answer to that is that it is a discrete, HIGHLY simplified version of Kirk Goldsberry's EPV. Why do I say that? Basically EPV and PAA both try to capture how many points a given player contributes to his team above what a hypothetically average player would have in the same situations. PAA is highly simplified because it only looks at the end state, instead of tracking the motion throughout, and because it can't control for the quality of a player's shot or how good his pass was or what his other options at the time were. If you're still confused (I would be) I'll try to clear it up by showing how PAA would measure a given play in a game. Let's look at maybe the most famous play of the last few years:
15 Comments
Quick update in my methodology.
I have now made an effort to include steals in my model. Because SportVu does not track the amount of time a player spends guarding a ball handler, I first made the general assumption that a player guards the ball handler roughly the same amount of the time as he himself has the ball. Obviously this is not the case for defensive specialists, or for offensive specialists, so to counterbalance that I added on to that one forth of a players minutes, which represents one half of the time a player spends on defense. This also counterbalances the fact that players can almost randomly get a steal with a bad pass or double team - indeed, before this, big men actually dominated the top ten with respect to steals. Using this time frame, I could then calculate the average number of steals per minute and see who was above and below average. I am still fiddling around with this one - if you have any suggestions feel free to let me know. THE THEORY BEHIND IT ALL EDIT: as of 12/29 I will be using the rough approximation of contested rebounds above average. |
AuthorA struggling Celtic's fan Archives
April 2015
Categories
All
|