In-game win model

13 views
Skip to first unread message

justin

unread,
Mar 7, 2013, 2:16:02 PM3/7/13
to study-hal...@googlegroups.com
Hi,

I'm Justin, founder of the Tempo-Free Gridiron (http://blog.tempo-free-gridiron.com/). I've been playing around with college football analysis as a side hobby for a few years now, and more recently Eddie (blog co-author) and I finally got around to posting in-game win probabilities for CFB games. I know we're not the first to do this -- I was motivated by some charts I saw over at MGoBlog that I can't seem to locate anymore -- but decided to take a stab at it in our own way. The first cut of our model provided win probabilities for the end of the 2011-12 season, and last year we were able to create graphs for all FBS-FBS games.

We've outlined our methodology in some detail here:


and I'd love to get some feedback from y'all.

Thanks, and thanks again to Bill for setting up this group.
-j

DSMok1

unread,
May 7, 2013, 8:33:03 AM5/7/13
to study-hal...@googlegroups.com
Hey, Justin!

Wow, great work!  I already always went to your site as my first place to get recaps of each Saturday's games.  I'm a chart geek, I guess...  Your work on automating the in-game win probabilities is impressive.

Your methodology for this update appears solid.

One of your "new" charts appears a bit odd to me, though--the chart of the SEC title game.  I can't understand why the probabilities were jumping around so much early in the 1st half, when the teams were simply trading punts.  Shouldn't that have been pretty stable, mirroring your old win probability metric's steady decline?  Any thoughts on that?


-Daniel

Justin Moore

unread,
May 9, 2013, 10:48:15 AM5/9/13
to study-hal...@googlegroups.com
Thanks,

That's a good question. I'm looking over the chart and my first guess is that it's a sampling issue. In order to not overwhelm the Google chart server I only take a data point from every 30 seconds of game clock time. There were some early back-and-forth drives before anyone scored, but there's still the possibility of a bug somewhere in my odds calculation code.

Thanks for pointing that out. I'll have to go check it out.
-j


--
You received this message because you are subscribed to the Google Groups "Football Study Hall: Data Sharing" group.
To unsubscribe from this group and stop receiving emails from it, send an email to study-hall-cfb-...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

Reply all
Reply to author
Forward
0 new messages