Thanks for the post. It's nice to get some insight into your work and methods. I'd like to hear what you think of these two potential issues:
1- "Recursivity" of the opponent strenght adjustment: You adjust the value of each play by team A based on how it performed relative to the average success of Team B's opponents against B over the season. But this average value is itself conditioned by B's strenght of schedule. So if A plays against B, a good SEC team with a brutal schedule that forced it to drop lots of points over the season, PAN will weigh down A's success against B based on B's performance in all these other tough games. PAN will be pessimistic about A's success. Opponents with cupcake schedules will induce the reverse bias in PAN.
2- Taking league conversion averages to assign points to the different plays/positions seems to implicitly trash a lot of interesting team-specific information. I'm pretty sure there's considerable variation in goal line conversion rates for example, so getting that goal line position may not be as beneficial for all teams. I actually don't like this approach (also taken by Football Outsiders) of assigning a priori weights to different plays or statistics. It seems to me like this strategy can't help but be ad hoc. I would advocate mushing together all the relevant stats into something like a factor analysis, where the value given to each type of play is essentially recovered inductively from the covariance structure (weights will be the factor loadings). Would you consider doing this?
Thanks again for the great work. I enjoy reading it very much!
You're right that this quote is wrong, but it doesn't mean the article is not probabilistic. This is a statistical model so you can only draw probabilistic infnce from it anyway. The author uses a lot of "on average" language elsewhere in the post, and it seems nit picky to bash him from that one lazy language slip.
But yeah, you're right, the quote gives the wrong impression.
Ver y cool stuff. I love it! Do you have enough instances of coaching changes in the early years to look at 2 or 3-year effects? I réalise this is going to crush your sample size, but it might be intesting given delays for the implementation of new schemes.
His argument was probabilistic, not deterministic. ON AVERAGE replacing a coordinator doesn't make a difference. One outlying case never falsifies a probabilistic argument.
Also, the whole point of the second part of his piece is to highlight teams that don't fit this average pattern. You should read the rest.
Pardon my ignorance, but is that the golden M that's right in front of Hatcher in the middle of the diag? If so, I had no idea you weren't supposed to walk on that thing... Anyone know the origin of this?
Tim, Love the profiles. Thanks for your hard work! Quick suggestion: It would be nice, for added convenience to the reader, if you could include links to the rating agencies' profiles of the players you discuss. Just a suggestion, do with it what you want
I don't dispute the observation; these kids really look like they want to play as freshmen. But why would a kid who would benefit from a year bulking up and learning the system want to burn the redshirt? Is it just the fun of playing? Seems to me that if you're a highly touted recruit who could get better with a year of S&C, you improve your odds of high performance, and thus improve how NFL scouts will look at you if you wait a bit.
I'm not saying they are making a wrong choice. I just honestly don't understand the reasoning. What do you think?
The slope of the least squares line is about 1. I don't think you can say that "first-year coaches tend to do as well as or better than their predecessors". Their success rate seems to be neither better nor worse. This is also pretty clear from the top graph which looks like a nice spread-out Bell centered around 0.
The observation about taking over 12-win teams is accurate, but there's obviously not much place else to go but down when your team wins 12 games in a season...
Thanks for the contribution! This is somewhat OT, but not really. Given the recent discussion about commits visiting other schools, I was wondering how safe was the assumption you make here, i.e. that the current commits are going to stay the course till signing day.
You know these guys more than any of us, so I'd be curious to know what you think.
I wrote this short document a few weeks ago. It's not great but it has a short critique of proprietary stats that is quite along those lines. Tell me what you think.
http://db.tt/A4a32YC
Maybe we can split the work. I am typically able to extract data from webpages and to format it in a nice database if it all looks the same. But I really don't have time to figure out what the HTML naming convention is. Tell me the sets of HTML pages I have to loop over, and how these URL addressee match teams/seasons/opponents, and I should be able download them (well, I'll try anyway).
Recent Comments
Ver y cool stuff. I love it! Do you have enough instances of coaching changes in the early years to look at 2 or 3-year effects? I réalise this is going to crush your sample size, but it might be intesting given delays for the implementation of new schemes.
His argument was probabilistic, not deterministic. ON AVERAGE replacing a coordinator doesn't make a difference. One outlying case never falsifies a probabilistic argument.
Also, the whole point of the second part of his piece is to highlight teams that don't fit this average pattern. You should read the rest.
This is a pretty nice article on him: http://www.nytimes.com/2011/05/22/sports/soccer/lionel-messi-boy-genius.html?_r=1&ref=sports
Pardon my ignorance, but is that the golden M that's right in front of Hatcher in the middle of the diag? If so, I had no idea you weren't supposed to walk on that thing... Anyone know the origin of this?
This is really cool stuff. The critics should chill a bit. I really appreciate this fun insider's view. Thanks dude!
Plural form is easy.
Tim, Love the profiles. Thanks for your hard work! Quick suggestion: It would be nice, for added convenience to the reader, if you could include links to the rating agencies' profiles of the players you discuss. Just a suggestion, do with it what you want
yeah man. This was *very* nice
I don't dispute the observation; these kids really look like they want to play as freshmen. But why would a kid who would benefit from a year bulking up and learning the system want to burn the redshirt? Is it just the fun of playing? Seems to me that if you're a highly touted recruit who could get better with a year of S&C, you improve your odds of high performance, and thus improve how NFL scouts will look at you if you wait a bit.
I'm not saying they are making a wrong choice. I just honestly don't understand the reasoning. What do you think?
may get you informative answers from people not named "Tom"
The slope of the least squares line is about 1. I don't think you can say that "first-year coaches tend to do as well as or better than their predecessors". Their success rate seems to be neither better nor worse. This is also pretty clear from the top graph which looks like a nice spread-out Bell centered around 0.
The observation about taking over 12-win teams is accurate, but there's obviously not much place else to go but down when your team wins 12 games in a season...
Makes sense. Thanks for the reassuring reply!
Hi Tom,
Thanks for the contribution! This is somewhat OT, but not really. Given the recent discussion about commits visiting other schools, I was wondering how safe was the assumption you make here, i.e. that the current commits are going to stay the course till signing day.
You know these guys more than any of us, so I'd be curious to know what you think.
Thanks!