2016-09-06

NCAA Football Power Rankings, Week 1

It's algorithm season for me again. Here's how these work, lightly edited from last year's week 1 post. in turn mostly taken from 2014, and so on:

A few years ago, I thought to myself, "I could do a better job than some of the BCS computers".  As I recall, at the time I felt Mizzou was underrated.  The BCS rules stated margin of victory cannot be used, so I thought I'd stick to that.  I put together a convergence algorithm, in which I initially rank the teams by record (so all the undefeated teams start tied at #1), then calculate each team's strength of schedule based on the current rankings of its opponents, and re-sort based on that strength.  I run through this process until 2 consecutive iterations give the same result, or there is a loop.  So, if iteration 10,002 gives the same rankings as iteration 10,000, then each subsequent iteration would fluctuate between the rankings in 10,000 and 10,001.  I average those rankings to come up with the final rankings.  It is still possible - but generally unlikely - that teams can tie for a final ranking. Now that the BCS is dead, I'll just consider these rankings my advice on #1 to #4 to the College Football Playoff committee.

Due to most teams playing other teams in lower divisions at times, and a lack of desire on my part to delve so deeply that I have to track down NAIA schedules, I've decided to count all games listed on ESPN's college football site, which includes all FBS, FCS, and Division II and III schedules, and any of their lower-tier opponents, but only when those opponents play at least a Division III team. (i.e., an NAIA team will be listed when they play against a Div III team, but the rest of their games won't be listed, resulting in a lot of 0-1 teams mixed into my system).  I then filter the final results to only the FBS schools.

In week 1, there are only 2 possible situations for a team*. First is to be 1-0, having defeated an 0-1 team. Second is, unsurprisingly, to be a 0-1 team, having lost to a 1-0 team. For this reason, in week 1, all 1-0 teams will be tied for #1, and the rest tied at N+1, where N is the number of #1 teams. Did your team choose to beat up on an FCS (formerly known as Division 1-AA) team? Congrats, you're #1 this week!

* - Normally this is true, but this year, Hawai'i decided to have a pre-Week-1 game, as well as one in Week 1. So they're 0-2, and there's just the slightest bit of extra gradation in the rankings. California and Michigan are the worst of the 1-0 teams, having beaten an 0-2 team instead of everyone else who beat a 0-1 team. I was a little concerned when I saw Hawai'i only ranked #127, but that's because Marshall hasn't played yet. You don't get a ranking until you play.

I believe the FBS stayed steady at 128 teams this year, which is just begging for a 7-round tournament, right? Anyway, no teams left or joined, but I still have to modify my filter slightly, to account for ESPN's stylistic differences in listing the teams. For example, Hawaii is now Hawai'i. A few years ago, they waffled between "Mississippi" and "Ole Miss". I try to keep up with any adjustments, but if you notice your FBS team missing, leave me a comment and they'll be there next week. I think I'm good, because I have 127 teams represented and the aforementioned Marshall still to add.

Every year, someone makes me tweak my software a little, because I wrote it awhile ago and it's not very dynamic. The previous long school name was "Southwestern Assemblies of God", but this year there's now "University of Texas of the Permian Basin". Good ol' UTPB.

Now onto the slightly less-meaningless-than-usual Week 1 rankings for 2016.

1Louisville1-0
1Central Michigan1-0
1Wake Forest1-0
1Cincinnati1-0
1Connecticut1-0
1NC State1-0
1Indiana1-0
1South Carolina1-0
1Utah1-0
1Utah State1-0
1Western Kentucky1-0
1Minnesota1-0
1New Mexico1-0
1Idaho1-0
1UNLV1-0
1Stanford1-0
1Michigan State1-0
1Baylor1-0
1Eastern Michigan1-0
1Ball State1-0
1Syracuse1-0
1Army1-0
1Colorado1-0
1Toledo1-0
1Nevada1-0
1Alabama1-0
1Clemson1-0
1Houston1-0
1Wisconsin1-0
1Ohio State1-0
1TCU1-0
1Washington1-0
1Texas A&M1-0
1Iowa1-0
1Georgia1-0
1Oklahoma State1-0
1Oregon1-0
1Florida1-0
1Georgia Tech1-0
1Navy1-0
1Purdue1-0
1Western Michigan1-0
1Boise State1-0
1South Alabama1-0
1West Virginia1-0
1Maryland1-0
1Virginia Tech1-0
1Pittsburgh1-0
1Air Force1-0
1Penn State1-0
1Texas State1-0
1Illinois1-0
1Arkansas1-0
1Florida Atlantic1-0
1Duke1-0
1East Carolina1-0
1Miami1-0
1Georgia Southern1-0
1Troy1-0
1Akron1-0
1Middle Tennessee1-0
1SMU1-0
1Tulsa1-0
1South Florida1-0
1UCF1-0
1Kansas1-0
1Memphis1-0
1Texas San Antonio1-0
1Louisiana Monroe1-0
1Southern Mississippi1-0
1Nebraska1-0
1UTEP1-0
1Texas Tech1-0
1San Diego State1-0
1BYU1-0
1Wyoming1-0
1Arizona State1-0
1Texas1-0
1Old Dominion1-0
1Florida State1-0
1Tennessee1-0
82Michigan1-0
82California1-0
84Northern Illinois0-1
84Boston College0-1
84Notre Dame0-1
84Buffalo0-1
84Ole Miss0-1
84Northwestern0-1
84Louisiana Lafayette0-1
84Rice0-1
84Mississippi State0-1
84Temple0-1
84Missouri0-1
84Charlotte0-1
84Colorado State0-1
84Oregon State0-1
84Arkansas State0-1
84Virginia0-1
84Kent State0-1
84Ohio0-1
84USC0-1
84Tulane0-1
84Louisiana Tech0-1
84Auburn0-1
84Oklahoma0-1
84Florida Intl0-1
84LSU0-1
84Bowling Green0-1
84Appalachian State0-1
84Kansas State0-1
84North Texas0-1
84San Jose State0-1
84Vanderbilt0-1
84Rutgers0-1
84UCLA0-1
84Miami (OH)0-1
84Kentucky0-1
84North Carolina0-1
84Iowa State0-1
84Fresno State0-1
84Washington State0-1
84New Mexico State0-1
84Massachusetts0-1
84Arizona0-1
84Georgia State0-1
127Hawai'i0-2

2 comments:

  1. LOLOLOLOL. This is absolutely wonderful. I work in the department that has been "waffling around with these names". We just changed Nicholls State to Nicholls and McNeese State to McNeese (both non-FBS) yesterday. I would say look out for an accent mark added to San Jose State soon.

    On a side note, if you ever find problems with the schedule, feel free to yell at me. :D We do occasionally have problems with records updating correctly, but those issues should all normally be fixed if you consistently run this on Tuesday/Wednesday.

    Neat idea for an algorithm. I believe the College Football Playoff rankings use something similar to this (with a ton of added mumbo jumbo nonsense), which is why they don't release rankings until week 10. I'm curious how this method ends up stacking up to this http://www.espn.com/college-football/statistics/teamratings which uses last year's results and recruitment data to formulate rankings at the start of the season and then slowly phases those two variables out as the season progresses.

    ReplyDelete
  2. I think San Jose State has had the accent mark for awhile on the schedule page, because I have to manually fix it each week they play because my Ruby script doesn't like it.

    I definitely don't take any previous performance or offseason moves into account, it's only based on performance. The last few years, it's been pretty close to the popular consensus near the top by the end of the season, but it starts differing pretty wildly even down around #20. I might have to put together a +/- chart later this year comparing mine to the ratings you linked, just to see how close I am.

    ReplyDelete