Well, I'm a bit of a numbers nerd and I was bored at work yesterday, so I threw together a little application that scrapes league data from GLB and computes a few different ratings formulas that measure various aspects of each team's performance.
Enjoy.
http://grimland.uta.edu/glb/SouthAmericaA7-Zeta-Week5.html
http://grimland.uta.edu/glb/SouthAmericaA7-Alpha-Week5.html
All columns are sortable.
Formula explanations:
AIA:
Just looks at wins and losses. Rewards wins over good teams.
JLG:
Just looks at wins and losses. Rewards wins over good teams, and penalizes losses to bad teams.
OMF:
An attempt to predict what a team will score in a given game based on their previous game performances. It takes average score into account, but also recent extreme high or low scores, and winning or losing trends. This one especially suffers from a small sample size, so it should get more accurate as the season goes on.
DMF:
Much like OMF, but this one attempts to predict points allowed rather than points scored (thus lower is better). Same limitations apply regarding sample size.
RPI:
The same formula found on the Wikipedia page for RPI (Ratings Percentage Index), minus the bit about home games counting less and road games counting more. RPI is just measuring the strength of the teams you've already played. The more cupcakes you've had, the lower your RPI. The tougher the teams you've faced, the higher your RPI. Your winning percentage is a small component, but in general it doesn't really say a whole lot about your own team's strength.
NOTES:
I call these ratings rather than rankings, because not all of them are useful for direct comparisons between teams. AIA and JLG are the only ones that could even possibly be used to compare teams against each other, with JLG being the more accurate one.
All of the formulas except OMF/DMF give more weight (in an indirect way) to conference games, and less importance to the Week 1 non-conference game.
These are formulas that I mostly came up with on my own (except RPI and parts of OMF/DMF). I'm no statistician so they could be flawed or useless or misleading or just plain wrong. But it's fun to play with numbers.
Enjoy.
http://grimland.uta.edu/glb/SouthAmericaA7-Zeta-Week5.html
http://grimland.uta.edu/glb/SouthAmericaA7-Alpha-Week5.html
All columns are sortable.
Formula explanations:
AIA:
Just looks at wins and losses. Rewards wins over good teams.
JLG:
Just looks at wins and losses. Rewards wins over good teams, and penalizes losses to bad teams.
OMF:
An attempt to predict what a team will score in a given game based on their previous game performances. It takes average score into account, but also recent extreme high or low scores, and winning or losing trends. This one especially suffers from a small sample size, so it should get more accurate as the season goes on.
DMF:
Much like OMF, but this one attempts to predict points allowed rather than points scored (thus lower is better). Same limitations apply regarding sample size.
RPI:
The same formula found on the Wikipedia page for RPI (Ratings Percentage Index), minus the bit about home games counting less and road games counting more. RPI is just measuring the strength of the teams you've already played. The more cupcakes you've had, the lower your RPI. The tougher the teams you've faced, the higher your RPI. Your winning percentage is a small component, but in general it doesn't really say a whole lot about your own team's strength.
NOTES:
I call these ratings rather than rankings, because not all of them are useful for direct comparisons between teams. AIA and JLG are the only ones that could even possibly be used to compare teams against each other, with JLG being the more accurate one.
All of the formulas except OMF/DMF give more weight (in an indirect way) to conference games, and less importance to the Week 1 non-conference game.
These are formulas that I mostly came up with on my own (except RPI and parts of OMF/DMF). I'm no statistician so they could be flawed or useless or misleading or just plain wrong. But it's fun to play with numbers.
Last edited Mar 4, 2009 19:16:03






























