Wednesday, April 25, 2018

New Power Rankings System (?)

As much as I love some of MFN's window dressing, the Power Rankings are not among my favorites. I have always thought the rankings didn't match up well with how the teams were faring, and the fluctuations from week-to-week were maddening.

So, at least for our league, I decided to do something about it. That is how MOHR (Modified Ohio Harbin Ratings) was born.

MOHR is based on the Harbin ratings system used by the Ohio High School Athletic Association to determine playoff qualifiers in that state. I just adapted the formulas for our league's use. The real Harbin system was developed in the 1960s by Jack Harbin, a cash register repairman and volunteer coach in the Cleveland suburb of Willowick. The OHSAA formally adopted it in 1972.

The short story is this: you get first-level points for games you win. You get second-level points for games your defeated opponents wins.

There are no points for losses. There are no margin-of-victory points.

The OHSAA's explanation of the system is a bit more complex, because they need to deal with bye weeks, canceled games, out-of-state opponents, etc. For the MFN purposes, everybody is a Division II school (6 first-level points) and everybody's multiplier is 100 because every team is playing the same amount of games.

Joe Eitel's site tracks the Harbin ratings during the season, if you want to see how this thing works in real life.

Because of small sample size, the early ratings and their variables can be a bit wonky. The OHSAA doesn't release them until after the fourth game (40 percent of the season). In our 16-game schedule, that would be after Week 6.

First, here is a screen shot of the MOHR from the 1977 USFL/WFL regular season (click to enlarge all screen shots).


Here is a screen shot of the MOHR top-to-bottom, with the teams' W-L records:



Finally, here is a screen shot of side-by-side comparison with the MFN power rankings at the conclusion of the 1977 season vs. the MOHR counterpart:


I think all of the above offer some evidence as to MOHR's comparative accuracy. But, you tell me.