Win probability

[Sef Dresslar]

I've attached a modified version of the play-by-play cfbstats data.  Notes:

1. "Win" and "Loss" columns added based on the game result for the team listed as the offensive team.  Also a "Time Remaining" column for total time remaining in the game.

2. Only field goal attempts, punts, PAT attempts, 1st and 10s, and 1st and Goal situations are included.  As I mentioned in the other topic, I'm looking first at tackling win probability (and more specifically Win Probability Added) on a unit basis.

3. Only plays with a value in cfbstats' Clock column are included.

There are 246,192 entries in the file for the 2005-2011 seasons.

[Bill Connelly]

And four days later, I'm finally getting a chance to take a look at this. A couple of thoughts/questions:

1. I know Brett T. uses win probability at MGoBlog a lot. Brett, are you doing that because of information you've gleamed from work like this, or are you using the NFL calculations I've seen used elsewhere?

2. Honestly, Win Probability was one of the reasons I wanted to create this group, as I'm pretty sure a lot of us are either doing a lot of the same work or intending to do a lot of the same work. It would be so cool to have something of a unified tool from which to work.

[Brett Thiessen]

For the last couple of years I have added a win probability calculation with my numbers. I started from scratch in building it, no NFL or other models were built in.