APBRmetrics

[supersub15]

Trying to wrap my head around applying this SRS explanation to an Excel spreadsheet.

Has anyone done it before? Thanks for any help.

[DSMok1]

[supersub15]

[DSMok1]

[supersub15]

Thanks. When you have a minute, can you over some of the main fields?

[DSMok1]

[The Specialist]

No offense, but that is much more complicated than it should be. Go by the library and check out Mathletics, there is a section on creating Power Rankings that lays out the easiest way to do this in Excel.

[DSMok1]

[The Specialist]

Hey DSMok1-

I have SRS set up "my way" in Excel for NBA and NFL, but I actually tried to set up using your NCAA FB numbers and ran into an issue. I use the SOLVER, and it can only handle 200 variables. However, with there being 121 teams in NCAA FB, it's not possible using my methods.

[DSMok1]

[Rhuidean]

Never implemented this before myself before, but I coded it up in Octave

EDIT: Fixed a few typos in the comments, here is the new version (http://pastebin.com/hSu5D5bd)

Here is a sample run:


octave:33> clear;srs
R =

0.94568
1.79588
0.10213
1.96496
2.06186

octave:34> M
M =

0 0 1 0 1
0 0 0 1 0
1 0 0 0 0
0 1 0 0 0
1 0 0 0 0

octave:35> V
V =

0.22436
1.14089
-0.21310
1.36633
1.74663

octave:36> N
N = 5
octave:37> G
G = 3
octave:39> Matchups_vec
Matchups_vec =

2 4
1 5
3 1



So I have an N=5 team league, with 3 games played thus far. The vector V gives the point margin for the 5 teams. The Matchups_vec and the matrix M tell the same thing; we've seen a game with team 2 versus team 4, team 1 versus team 5, and team 3 versus team 1.

Running the code then gives the ratings specified by the vector R for each team.

If you don't use Matlab/Octave, you can probably convert the code I wrote into your preferred language.

Feel free to download the above and play around with it. Let me know if you catch any bugs (or have any questions.)

And of course, as you can see from the code I wrote, you can try any values of N (number of teams), G (number of games), Matchups_vec, etc you like.

[DSMok1]

[Rhuidean]

Is that just a preprocessing step? So rather than using the team's actual point margin (which is a sum of the point margins over the games it played), you transform each of the game point margins in some way (for example, capping it somehow, transforming it to +1 for win, -1 for loss, 0 for tie?)

Or is it something a bit different from preprocessing?

If it is just preprocessing, then you'd probably want to write a bit of code that handles that step first, before you hand it off.

[JoshEngleman]

[DSMok1]