APBRmetrics

Ryan J. Parker · Joined: 23 Mar 2007 Posts: 708 Location: Raleigh, NC

davis21wylie2121's post on the NCAA championship game win probability that is based on Ed's win expectancy inspired me to get to work on understanding this aspect of the game better.

My first implementation is based on a Brownian motion model by Hal Stern (more details can be found in this blog post).

There are some obvious limitations, such as who has possession and the probability of overtime are not taken into account, but I hope it's a good illustration to start with.

Here is a sample graph from game 1 of the Jazz/Rockets series:

The rest of the graphs can be found here.

Any feedback is appreciated!

Neil Paine · Joined: 13 Oct 2005 Posts: 774 Location: Atlanta, GA

Very cool, Ryan. I read the Stern a few years ago but never put together a model; I'm glad somebody finally did. The next step, obviously, is to incorporate ball possession...

Ryan J. Parker · Joined: 23 Mar 2007 Posts: 708 Location: Raleigh, NC

Yeah I don't think this is useful for what I'm interested in, specifically how strategy (such as lineup changes, etc.) affect a team's probability of winning, but going through the process has helped raise some important questions.