Heavy Lifting - thoughts and web finds by an economist
     I also contribute to Division of Labour Load HL's Front Page
Tuesday, February 06, 2007

The bye week is bad?

I am messing around with NCAA football data from the 2006 season. I have data on all the games played in Division IA and created a dummy variable that takes a value of one if the team wins (absolutely) and zero otherwise.

For giggles, I created a dummy variable that takes a value of one if the previous week was a loss for the team and another that takes a value of one if the previous week a bye.

I also included a continuous variable which measures the relative ranking of the team to its opponent - a value of one indicates parity of teams, values above one suggest the team in focus is worse, and values less than one suggest the team in focus is better.

I have a bit of a problem in that each game is included in the data twice - one for the home team and one for the away team. I am trying to figure out a way to get around this problem in STATA, but for the moment, here are some results.

I threw the data into the probit blender and here's what we get:

Probit regression Number of obs = 1428
LR chi2(3) = 215.79
Prob > chi2 = 0.0000
Log likelihood = -881.91787 Pseudo R2 = 0.1090

------------------------------------------------------------------------------
mwin | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
prevlost | -.4698301 .0749153 -6.27 0.000 -.6166613 -.3229988
prevoff | -.2184209 .1137865 -1.92 0.055 -.4414383 .0045965
relrank | -.1920602 .018773 -10.23 0.000 -.2288545 -.1552658
_cons | .5271694 .0551686 9.56 0.000 .4190409 .635298
------------------------------------------------------------------------------
Note: 11 failures and 0 successes completely determined.

. mfx compute

Marginal effects after probit
y = Pr(mwin) (predict)
= .4460042
------------------------------------------------------------------------------
variable | dy/dx Std. Err. z P>|z| [ 95% C.I. ] X
---------+--------------------------------------------------------------------
prevlost*| -.1823939 .02833 -6.44 0.000 -.237915 -.126873 .380252
prevoff*| -.0848726 .04324 -1.96 0.050 -.16962 -.000125 .109244
relrank | -.075918 .00725 -10.47 0.000 -.090123 -.061713 2.39726
------------------------------------------------------------------------------
(*) dy/dx is for discrete change of dummy variable from 0 to 1

The lower panel reports the marginal effects of the dummy variables on the odds of winning. If the previous week was a loss, the team has an 18% chance of losing the current week's game, everything else equal. Moreover, the bye week reduces the odds of winning by about 8.5%.

I wonder if result will prove robust. I have some work to do on the data.

Comments: Post a Comment



Purchase




Le Chai - galerie du vin



Support



Popularity

Posts that contain Craig Depken per day for the last 90 days.



Contacts

Heavy Lifting's Main Page
Email Me
Atom Feed

WWW
Heavy Lifting




Great Links

Money I Found Today

Heavy Lifting - Firehose style (56k warning)



Recent Posts

- The bye week is bad?




Archives


home


Visitors
Site Meter Blogroll Me!



Credits

Modified maystar design
powered by blogger