College football season is upon us, and that means Year 6 of the Gus Rankings! As we did last year, we've added a new wrinkle to the calculations this year to try and better capture the nature of game outcomes. The goal is to try to build a ranking system from the ground up that is as objective and as transparent as possible.
Last year, we considered the margin of victory when determining the points that teams get from games. Instead of having a last second field goal swing a team from a win (and the points generated from a win) to a loss (and the points subtracted from a loss), we generated an algorithm that translated a final point differential into a fraction of a win and a fraction of a loss. (It was at this point that you fell in love with the inverse hyperbolic sine function.) Does it perfectly capture the dynamics of a game? No; we aren't ever going to be able to capture all of the nuances of a game down to a single number. But it was a step in the right direction. Also, overtime games were also viewed as a 50% win and a 50% loss-- why judge some games by more than sixty minutes and others by sixty minutes exactly? We kept everything the same outside of that from the original rankings-- and so that was Gus+.
This year's adjustment corrects for what I consider to be an oversight of last year's system. At the base of the original Gus Rankings was the idea that you gain points equal to the number of wins the teams you defeated had, and lost points equal to the number of losses of the teams that defeated you. That never changed. What changed was that you now didn't always get the entirety of your opponent's win total if you defeated them; if the score was sufficiently close, then you only got a fraction of the win total or loss total. This year, we're keeping track of the effective wins and effective losses that a team accrues throughout the year. So underneath every team's reported, on-the-stadium-program record is their effective record-- the sum of the partial wins and losses that they accumulate throughout the year. It is from your opponents' effective wins and effective losses by which you now gain or lose points. It's called Gus++. If you like the name, thanks! If you don't, it was Rob Holub's idea.
A note of thanks goes out to my graduate assistant, Kevin Gormley, for wrangling a lot of this together into a workable spreadsheet. We're fairly certain the wrinkles have been ironed out; once we are good on that margin, we'd like to be able to report all three iterations of Gus Rankings for comparison and, you know, because data. Onwards and upwards! Comments, complaints, grievances, etc. are welcomed as always.
Without further ado, here are this week's Gus++ Rankings. Oklahoma nabs the top spot; UMass stakes itself to the early "lead" for the worst team in the country. Interesting results throughout!
Tuesday, September 09, 2014
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment