I Put ESPN's Position Rankings in a Spreadsheet

  • You are viewing Orangepower as a Guest. To start new threads, reply to posts, or participate in polls or contests - you must register. Registration is free and easy. Click Here to register.
Nov 17, 2012
32
29
568
40
#3
Just the offense averaged looks like this:

OSU 3.25
OU 3.5
TCU 3.75
KSU 4.5
TX 4.75
WVU 4.75
BU 6.5
ISU 7.5
TTU 8
KU 8.5

The defense shakes out this way:

TX 2.33
TCU 3.33
OSU 3.67
KSU 3.67
OU 3.67
WVU 6.33
BU 7.00
ISU 7.67
KU 8.33
TTU 9.00
 
Last edited:

Lab Rat

Hold on while I make a chart
A/V Subscriber
Jan 5, 2012
6,952
9,906
743
#10
Obviously. But how much to each position group? @Lab Rat , help please.
Your suggestion is good, but it isn't obvious to me how to assign weights to each position group. One way to would be to retrospectively rank each team's position groups from past seasons, compare those rankings to their team's conference records, and then find weighting parameters by fitting a regression. This is probably the best way to do it, but is pretty complicated.

Another less accurate (but perhaps easier) way would be to calculate the average NFL salary for each position and then use weighting parameters that are proportional to those salaries. This assumes that the NFL pays players based on their relative importance and that NFL and college positions perfectly scale. This isn't exactly true (RBs are considered more valuable to college teams than to NFL teams), but are still pretty good assumptions, IMO.

If we use the latter method, the salaries can be found here for different NFL positions (warning: Sports Illustrated link).
Screen Shot 2017-05-18 at 6.12.19 PM.png


Here's the resulting spreadsheet. Instead of using the rankings, I computed a point value for each ranking, where higher numbers are better (11 - conference ranking = point value per position). A ranking of #1 equals a point value of 10. A ranking of #10 equals a point value of 1. Then, I multiplied this point value by the corresponding position. The weighted parameter for the DL was the average of DTs and DEs. The weighted parameter for the DB was the average of S and DBs.
Screen Shot 2017-05-18 at 6.15.39 PM.png


Here are the charts:
upload_2017-5-18_18-16-6.png

upload_2017-5-18_18-16-18.png


upload_2017-5-18_18-18-44.png

upload_2017-5-18_18-18-57.png
 

NYC Poke

The Veil of Ignorance
Sep 24, 2007
36,863
44,775
1,743
#13
That assumes equal weighting to each category. I think there should be various adjustments. For example, having the top quarterback is probably more important that having the top special teams. Does anyone have any idea how this could be done?
My guess is that you're right on the weighting. It would take some serious number crunching to weight it, though.

My hunch is that special teams tend to be underrated. A TD on special teams counts as much as a TD made by the offense. A punt that lands inside the 20, followed by a drive that stalls at the 30 instead of the 20, followed by a missed FG, can swing a close game.

And magical seasons are made or broken in close games. See 2011.

I haven't seen a team with no QB that could control the game through special teams. Conversely, I have seen excellent teams lose through poor special teams play.

I'd like to see where the numbers fall.
 

CPTNQUIRK

I'm Your Captain!
A/V Subscriber
Nov 20, 2006
28,494
19,297
1,743
66
In the Boondocks in Creek County, Oklahoma
#14
Your suggestion is good, but it isn't obvious to me how to assign weights to each position group. One way to would be to retrospectively rank each team's position groups from past seasons, compare those rankings to their team's conference records, and then find weighting parameters by fitting a regression. This is probably the best way to do it, but is pretty complicated.

Another less accurate (but perhaps easier) way would be to calculate the average NFL salary for each position and then use weighting parameters that are proportional to those salaries. This assumes that the NFL pays players based on their relative importance and that NFL and college positions perfectly scale. This isn't exactly true (RBs are considered more valuable to college teams than to NFL teams), but are still pretty good assumptions, IMO.

If we use the latter method, the salaries can be found here for different NFL positions (warning: Sports Illustrated link).
View attachment 54263

Here's the resulting spreadsheet. Instead of using the rankings, I computed a point value for each ranking, where higher numbers are better (11 - conference ranking = point value per position). A ranking of #1 equals a point value of 10. A ranking of #10 equals a point value of 1. Then, I multiplied this point value by the corresponding position. The weighted parameter for the DL was the average of DTs and DEs. The weighted parameter for the DB was the average of S and DBs.
View attachment 54264

Here are the charts:
View attachment 54265
View attachment 54266

View attachment 54267
View attachment 54268
Absolutely awesome. Are you human or robotic?
 

hcebb

Wrangler
A/V Subscriber
Jul 23, 2008
81
47
1,568
GA
#17
Your suggestion is good, but it isn't obvious to me how to assign weights to each position group. One way to would be to retrospectively rank each team's position groups from past seasons, compare those rankings to their team's conference records, and then find weighting parameters by fitting a regression. This is probably the best way to do it, but is pretty complicated.

Another less accurate (but perhaps easier) way would be to calculate the average NFL salary for each position and then use weighting parameters that are proportional to those salaries. This assumes that the NFL pays players based on their relative importance and that NFL and college positions perfectly scale. This isn't exactly true (RBs are considered more valuable to college teams than to NFL teams), but are still pretty good assumptions, IMO.

If we use the latter method, the salaries can be found here for different NFL positions (warning: Sports Illustrated link).
View attachment 54263

Here's the resulting spreadsheet. Instead of using the rankings, I computed a point value for each ranking, where higher numbers are better (11 - conference ranking = point value per position). A ranking of #1 equals a point value of 10. A ranking of #10 equals a point value of 1. Then, I multiplied this point value by the corresponding position. The weighted parameter for the DL was the average of DTs and DEs. The weighted parameter for the DB was the average of S and DBs.
View attachment 54264

Here are the charts:
View attachment 54265
View attachment 54266

View attachment 54267
View attachment 54268
Couple of questions/comments about this that I find interesting since I'm a statistics nerd. First, the NFL salary distribution is truncated due to the salary cap, thus, the true value of a QB could be a lot more than what is shown but the salary cap skews this. Don't think you can correct for this, but it does skew ratings a bit. The second thing, and this is something I think @Lab Rat can account for, can you do this by total POSITION spending. In other words, the average OL salary is 1.76, but since there are 5 OL, the total spending is actually 8.8 making the group as a whole more valuable than the QB. Same thing goes for WR, DB/S, LB, DL. I wonder if this changes the rankings, because having a great OL where you can run all day on the opponents defense can neutralize a sub par QB. This is where the regression might be a lot more accurate.
 

Carneyman14

I wanna be formal, but I'm here to party
A/V Subscriber
Oct 20, 2008
20,046
18,680
743
30
Minneapolis, MN.
#19
Your suggestion is good, but it isn't obvious to me how to assign weights to each position group. One way to would be to retrospectively rank each team's position groups from past seasons, compare those rankings to their team's conference records, and then find weighting parameters by fitting a regression. This is probably the best way to do it, but is pretty complicated.

Another less accurate (but perhaps easier) way would be to calculate the average NFL salary for each position and then use weighting parameters that are proportional to those salaries. This assumes that the NFL pays players based on their relative importance and that NFL and college positions perfectly scale. This isn't exactly true (RBs are considered more valuable to college teams than to NFL teams), but are still pretty good assumptions, IMO.

If we use the latter method, the salaries can be found here for different NFL positions (warning: Sports Illustrated link).
View attachment 54263

Here's the resulting spreadsheet. Instead of using the rankings, I computed a point value for each ranking, where higher numbers are better (11 - conference ranking = point value per position). A ranking of #1 equals a point value of 10. A ranking of #10 equals a point value of 1. Then, I multiplied this point value by the corresponding position. The weighted parameter for the DL was the average of DTs and DEs. The weighted parameter for the DB was the average of S and DBs.
View attachment 54264

Here are the charts:
View attachment 54265
View attachment 54266

View attachment 54267
View attachment 54268