tag:blogger.com,1999:blog-38600807.post1369915628500173864..comments2023-11-05T04:16:44.937-05:00Comments on Advanced Football Analytics (formerly Advanced NFL Stats): The Best FG KickersUnknownnoreply@blogger.comBlogger7125tag:blogger.com,1999:blog-38600807.post-61946073524717417252007-09-20T22:27:00.000-04:002007-09-20T22:27:00.000-04:00I just redid the regression. I probably don't have...I just redid the regression. I probably don't have time to post everything tonight. But here are a few observations.<BR/><BR/>1. The perfect Vanderjagt year really threw off the data. His zero miss distance skewed the entire model.<BR/><BR/>2. By using a natural log of the accuracy as the dependent variable, the relationship between kick accuracy and attempt distance is more linear.<BR/><BR/>3. But by excluding Vanderjagt's perfect season and by using the ln(accuracy) model, the results are very different. The difference between the best and worst kickers may be several times larger than I previously estimated.<BR/><BR/>4. Surprisingly, neither warm weather or indoor kickers were significant factors.Brian Burkehttps://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-87540755421849451362007-09-20T18:22:00.000-04:002007-09-20T18:22:00.000-04:00Yeah, I get the same thing.best ~ +8% adj. accurac...Yeah, I get the same thing.<BR/><BR/>best ~ +8% adj. accuracy<BR/>worst ~ -12% adj. accuracy<BR/>difference = 20% adj. accuracy<BR/><BR/>average FG attempts = 24 per yr<BR/><BR/>24 att * 20% diff * 3 pts = 15 pts<BR/><BR/>A couple observations: FG kicker accuracy is fairly steady from year to year. The variance mostly comes from situational variables--attempt distances. Even including variance from luck (small # of reps), the difference among kickers is very small. This is probably why they are underpaid relative to their contribution to game outcomes. They're replaceable.Brian Burkehttps://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-74898530373295860132007-09-20T16:29:00.000-04:002007-09-20T16:29:00.000-04:00Doesn't this analysis then address your earlier qu...Doesn't this analysis then address your earlier question of whether Kickers are paid enough? If the difference between the best and worst kickers is about 15 points a year vs. average (my math may be wrong there), then pay is supressed because the differential between players is low. <BR/><BR/>Isn't high pay associated with high differential impact above average. Thus why QBs, RBs, WRs, and Pass Rushers are the highest paid players as the best have a higher differential over the average player?Rob Cullinhttps://www.blogger.com/profile/09388965581761000447noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-67253597435850564332007-09-20T11:20:00.000-04:002007-09-20T11:20:00.000-04:00Good point about the indoor/warm weather kickers. ...Good point about the indoor/warm weather kickers. I'll see if I can redo the regression with dummy variables for domes/warm climate soon.<BR/><BR/>Not sure about blocks, but I agree there are too few to effect the data in a meaningful way. I'm not sure kickers can really control that much of most types of blocks anyway. My bet would be that if a kicker puts his foot to the ball and it doesn't go through the posts for any reason, it's considered a miss.Brian Burkehttps://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-9562982700642753622007-09-19T13:55:00.000-04:002007-09-19T13:55:00.000-04:00Two comments:- The top 5 kickers all play for warm...Two comments:<BR/><BR/>- The top 5 kickers all play for warm weather or dome teams. I would suggest including a dome and temperature variable in the regression, or at least testing to see if these have significant effects on accuracy.<BR/><BR/>- Are blocks included in the data? My suspicion is that blocked kicks are not strongly dependent on the kicker. Of course there's not enough blocked kick data to really test this, but my guess is that blocks should be thrown out.Tarrhttps://www.blogger.com/profile/14368810359650066790noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-18799870389741079162007-09-19T11:22:00.000-04:002007-09-19T11:22:00.000-04:00Ed-I agree on both points. But I don't have indivi...Ed-<BR/><BR/>I agree on both points. But I don't have individual kick-by-kick data for the logistic regression. That would be interesting though.<BR/><BR/>I tried distance^2 variables, along with other operations to make the relationship between distance and accuracy as linear as possible. Actually using ln(FG%) as the dependent variable worked best, but it did not change the results much at all. Within the narrow range of average distances, the curve is already approximately linear.Brian Burkehttps://www.blogger.com/profile/12371470711365236987noreply@blogger.comtag:blogger.com,1999:blog-38600807.post-29418985478862955502007-09-19T11:05:00.000-04:002007-09-19T11:05:00.000-04:00You could also perform a logistic regression, usin...You could also perform a logistic regression, using an individual made/missed FG as the response regressed on distance + player, where the player coefficient would be the number of interest. I also think having a distance^2 variable makes sense.Anonymousnoreply@blogger.com