Football Outsiders FEI Ratings

Akinji07 said:
I have the data and can do it for any school you want. When you put all of the schools together it's kinda messy and the r2 drops to .65 at best.
Click to expand...

I would say to start by putting a few of the big football programs in, Bama, UGA, etc as individual sets of data and see how each of them trend.
 
coit said:
I would say to start by putting a few of the big football programs in, Bama, UGA, etc as individual sets of data and see how each of them trend.
Click to expand...

Here's the best model I can fit for uga.

FEI3.jpg
 
ee8384 said:
So this year is the low win percentage data point? If so, then the regression would say Tech should end up at about 0.500 if the average efficiency remains unchanged.

Three game winning streak inbound.
Click to expand...

That was my deduction as well. Good analysis.
 
What do linear fits of the data look like? Since it is illogical to expect win percentage to go up with declining efficiency ratings...
 
coit said:
What do linear fits of the data look like? Since it is illogical to expect win percentage to go up with declining efficiency ratings...
Click to expand...

FEI4.jpg


That's predicting 6.34 wins for UGA and 6.35 wins for Tech.
 
gt2690b said:
big difference is turnovers... makes the D look worse cause theyre not getting any plus they are getting put behind the 8 ball when our O commits them
Click to expand...

Turnovers are taken into account, and so is starting field position.

gt2690b said:
I would like to see our defensive efficiency vs. similar opponents. Last year we had a moderate road to the orange bowl with our opponent rankings/skill levels. This year we got some really tough ones in Notre Dame, Clemson, etc. We may even have a better defense this year and a worse performance.
Click to expand...

The efficiency rating is adjusted based on opponent offensive strength, so it shouldn't matter.


Fact is, our defense is not any better this year than it has been. The bright side for me is that a TON of young guys have played meaningful minutes, so perhaps the future will be brighter. Hinges on Gotsis' replacement, IMO.
 
Akinji07 said:
That's predicting 6.34 wins for UGA and 6.35 wins for Tech.
Click to expand...


That looks visually much better. Now, if you combine these and a few others onto a single graph, it will be interesting to see what is being shown: We achieve a higher win percentage for a given efficiency!
 
More perspective:

Opponent DFEI Ratings
Clemson - 3
Notre Dame - 29
Duke - 36
FSU - 46
Pitt - 48
Virginia - 52
UNC - 75

Avg - 41


Opponent OFEI Ratings
Notre Dame - 6
FSU - 13
Clemson - 14
UNC - 19
Pitt - 60
Duke - 81
Virginia - 95

Avg - 41
 
I want to see specific secondary numbers comparisons between 2009 and this year to make sure I'm not crazy. Marcus Wright was more likely to disrupt a route or a pass that season than any of our corners.
 
coit said:
That looks visually much better. Now, if you combine these and a few others onto a single graph, it will be interesting to see what is being shown: We achieve a higher win percentage for a given efficiency!
Click to expand...

Well if uga was at our efficiency this year they'd win less games is what the model is saying.
 
The Jacket said:
I want to see specific secondary numbers comparisons between 2009 and this year to make sure I'm not crazy. Marcus Wright was more likely to disrupt a route or a pass that season than any of our corners.
Click to expand...

We had Derrick Morgan and Morgan Burnett in 2009. The coaching was worse (IMO), but the players still made some big plays that we don't see a lot of right now.
 
Akinji07 said:
Well if uga was at our efficiency this year they'd win less games is what the model is saying.
Click to expand...

Dude, if you don't know how to put more than one set of data on a chart just say so. It's OK, really. Many men have difficulty performing.
 
coit said:
Dude, if you don't know how to put more than one set of data on a chart just say so. It's OK, really. Many men have difficulty performing.
Click to expand...

I already told you different teams have different correlations and different slopes and it's a mess to put them all together. Also average rating has a much better correlation than simply looking at offense or defense vs win%. Also I like the binomial model better than linear.
 
johncu said:
We had Derrick Morgan and Morgan Burnett in 2009. The coaching was worse (IMO), but the players still made some big plays that we don't see a lot of right now.
Click to expand...


No IMO required, our defensive coaching/coordination that year was horrendous. At the same time, Burnett was a man among boys in our secondary. It was during the Florida State game that year that it became clear to me we had bigger problems in that department than bad coaching. Two seconds after the snap you'd think that the receivers had actually begun the play already behind the man covering them.
 
Akinji07 said:
I already told you different teams have different correlations and different slopes and it's a mess to put them all together. Also average rating has a much better correlation than simply looking at offense or defense vs win%. Also I like the binomial model better than linear.
Click to expand...

Put your spreadsheet on dropbox or something and let me do it then. I'll make that data shine like the sun.
 
coit said:
Put your spreadsheet on dropbox or something and let me do it then. I'll make that data shine like the sun.
Click to expand...

Why don't you go öööö yourself instead, how's that sound?

ööööing idiot.
 
You must log in or register to reply here.
Back
Top