tag:blogger.com,1999:blog-38600807.post2307788454000063394..comments2018-06-02T14:19:34.554-04:00Comments on Advanced Football Analytics (formerly Advanced NFL Stats): Year-to-Year Team WinsUnknownnoreply@blogger.comBlogger2125tag:blogger.com,1999:blog-38600807.post-35340793572785775572008-05-20T15:09:00.000-04:002008-05-20T15:09:00.000-04:00Turnovers are partially random, but definitely not...Turnovers are partially random, but definitely not entirely. Some types appear more random than others, such as defensive interceptions (see my March post on Signal vs. Noise.)<BR/><BR/>I haven't looked at time spans longer than 1-year. But a longer longitudinal study of turnover stats might be interesting.<BR/><BR/>Justin-great article, but be careful when looking at large data sets for individual instances of significance. I don't doubt that some teams are systematically and organizationally flawed over the long-term--Cardinals, Lions, Bengals perhaps. But if you look at enough teams over enough possible periods of time you'll find significance in 1 out of 20 cases even when there really is no systematic cause (a Type I error).<BR/><BR/>A quick clarification about "regression to the mean:" As Phil points out about a .260 hitter, we wouldn't always predict that he will regress to the MLB mean, but to his own true intrinsic mean, whatever that happens to be. The same goes for football teams. Regression to the mean doesn't always mean 8-8, but rather regression toward a team's own mean based on it's fundamental talent and ability levels.Brian Burkehttps://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-52701734530214619422008-05-20T14:14:00.000-04:002008-05-20T14:14:00.000-04:00So are we assuming that turnovers are random? I lo...So are we assuming that turnovers are random? <BR/><A HREF="http://justintapp.blogspot.com/2008/04/how-will-your-favorite-nfl-team-do-next_21.html" REL="nofollow">I looked at</A> how many seasons it takes for certain teams to get to .500. Some franchises took a statistically significant long time to get there (either from above or below .500). I wonder how that correlates with turnovers. Have you looked at team progress over multiple seasons, including a dummy variable for time?JTapphttps://www.blogger.com/profile/17812390329612009756noreply@blogger.com