Calling All Statistics Experts

[HoosierCat]

This make my head hurt...maybe DC_Bengal (if I remember our resident stats expert correctly) can help.

http://www.bbnflstats.com/2007/06/rating-g...ad-coaches.html

Looks like Marvin comes out pretty good as a gameday coach tho, contrary to most (incliuing mine) fan opinions...?

[HairOnFire]

Looks like Marvin comes out pretty good as a gameday coach tho, contrary to most (incliuing mine) fan opinions...?

Pretty good? The 8th highest ranked of all NFL head coaches since 1983 nets a pretty good?

It is what it is. The stats were compiled before the 2007 season began so his most disappointing season isn't reflected, but I've always felt he was a good gameday coach, and still do. My biggest complaint is that he's occasionaly too conservative...too concerned with playing the field position game.

While we're at it, I've never understood all of the criticism about Marvin's clock management either. Just this season I watched Mike Holmgren, Mike Shanahan, and Joe Gibbs lose games few other coaches would have simply because they couldn't manage the clock. When has Lewis ever done this?

[HoosierCat]

Looks like Marvin comes out pretty good as a gameday coach tho, contrary to most (incliuing mine) fan opinions...?

Pretty good? The 8th highest ranked of all NFL head coaches since 1983 nets a pretty good?

It might rate better, I don't know. That's why I titled the thread as I did.

[sparky151]

This make my head hurt...maybe DC_Bengal (if I remember our resident stats expert correctly) can help.

http://www.bbnflstats.com/2007/06/rating-g...ad-coaches.html

Looks like Marvin comes out pretty good as a gameday coach tho, contrary to most (incliuing mine) fan opinions...?

His method calculates an expected number of wins based on offensive and defensive efficiency, modified by turnovers and penalties. He then assigns the difference between the expected value and the actual value to the coach.

I'm dubious. Look how he has Bill Walsh and George Seifert assigned negative values, meaning they cost the 49ers games during the Montana and Young era. The guys at the bottom of his list are certainly bad coaches and the guys at the top are good coaches but I think he attributes too much of the variance to the head coach. The timeliness of a penalty or turnover can change the outcome a lot more than an untimely one. There's a big difference between an intercepted Hail Mary at the end of the first half and a pick run back for a touchdown but they count the same in turnover and efficiency stats.

[HoosierCat]

Thanks sparky, that helps.

[walshfan]

Dave Wannstat better than Cowher??? Im not feelin it sorry.

[DC_Bengals_Fan]

I can't for the life of me find where on the site he actually develops the model and gives specifics. He could be using either a neural net or simple multiple linear regression to weight the contributions of the different variables he employs to predict wins.

I'm a little leery of some of the conclusions. First, if I'm reading it right, he uses penalties as a variable. For me, that's part of the job as head coach - to get penalties down. So I'd like to see what happens when he drops that variable - I expect it would result in a lower correlation but more informative statistic. Additionally, there's going to be a lot of skew to these numbers due to the coarseness of NFL records - 16 games is a very short sample, so you'd need a number of years.

But some of the conclusions seem pretty reasonable, actually. Especially for the Patsies, Belichick consistently eeks out a couple extra wins out of those guys, it looks like.

I would like to see results from his model to see how the accuracy holds at the high and low ends. Often, a model like this will be less accurate near the tails, making conclusions inaccurate. It might also be less accurate in extreme cases - say, a team with an exceptional offense but poor defense, or vice versa. But to me, the work looks pretty good at first glance. Interesting stuff.