Big 10 (Results-Based) Power Rankings: Non-Conference Edition

Submitted by Eye of the Tiger on

Hi, I'm Eye of the Tiger. You may remember me from such defunct diary series as "Reading the Tea Leaves," "Zone Blocking Zealot" and "Yet Another CC Roundup!" With the new era starting here in Ann Arbor, I'd decided to start a new diary series--I just couldn't figure out what it should be. Then it hit me: power rankings! Only, I wanted to see what happened if you tried to eliminate biases based on past performances (i.e. last year) or preseason hype. And I wanted to take my own biases out of the equation as much as possible. So I came up with a methodology for a results-based ranking system:

 

Methodology

1. All teams are scored on their results this year, and nothing else.

2. A baseline score is derived by associating results with numerical values, as follows:

  • 2 – win over “good” team
  • 1 – win over “solid” team
  • 0 – win over “not good” team/loss to “good” team
  • -1 – loss to “solid” team
  • -2 – loss to “not good” team

[Note: "good," solid" and "not good" are also performance based, using the same methdology. That means I looked at every team that any Big 10 school has played and tried to score their performances across 4 games. Yes, it took a long time. And was kind of boring.

Notes on categories: "Good" means scoring highly on this scale, and is highly correlated with top 25 deisgnations in major polls and advanced statistical rankings (S&P, FEI, FPI, etc.). On this scoring system, it generally means any team scoring 1.5 or higher. "Not good" generally means any team scoring -1.5 or lower. "Solid" are the teams between 1.5 and -1.5. There aren't many of those in the Big 10 right now, but there are across all of the FBS. As the season progresses, at least a couple Big 10 teams now classified as "good" will end up reclassified as "solid." Possibly one or two of those now classified as "not good" will as well.

Also of note: designations may also change week-to-week. Oregon, for example, is classified as "solid" due to its losses to MSU and Utah; Utah is classified as "good" due to its wins over Michigan and Oregon. If Oregon were to finish the season 10-2, while Utah slid to 8-4, that would likely reverse--and affect Michigan and Michigan State's scores ex post facto.]

3. A place of venue weght is added:

  • 0.5 - road win
  • 0.0 - home win/road loss
  • -0.5 - home loss

4. A margin of victory weight is added:

  • 1.0 – large win over “good” team
  • 0.5 – large win over “solid” team/small win over “good” team
  • 0.0 - small win/loss over "solid" team; large win/loss over "not good" team"
  • -0.5 – large loss to “solid” team/small win/loss to “not good” team 
  • -1.0 – large loss to “not good” team

[Note: "large" win is defined as a win by 20 points or more. This is a fairly arbitrary number, but one that seems--to me, at least--to be a reasonable indicator of a blowout.]

What's Being Measured

The purpose here is to rank teams according to what they've accomplished and not according to residual biases from preseason and/or the previous year. However, it does output some results that buck the conventional wisdom, and are largely a function of differences in schedule. That is to say, if two teams are both 4-0 and team X played better opposition than team X, it will be ranked higher--regardless of whether team Y is considered to be a more capable squad. These rankings thus do not predict who will end the season "best," just who has done the "most" in the tiemframe covered.

 

Planned Posting Schedule

I plan to update these power rankings either every 2 or 4 weeks--depending on how much time I have (which, between commuting to two jobs, marriage, parenting, actually watching football and writing freelance on the side isn't very much). But enough about that...

 

4 Week Rankings

 

1. Northwestern (4-0, AP #16): 3.5

Well there's one surprise! But if you think about it, it shouldn't be that surprising. After all, Northwestern may not be the most talented or heralded team in the Big 10 right now, but they've accomplished the most: beating a good Stanford team at home and a solid Duke on the road (a close win against Ball State is the only blemish). If you take away preseason hype and previous year bias, they have the best resume in the conference. 

2. Ohio State (4-0, AP #1): 2.5

Last year's champion has taken care of business, with two victories over solid teams (VT and NIU), but doesn't have a victory on par with Northwestern beating Stanford. It won't get that opportunity until November 21, when it travels to East Lansing.

3. (tie) Michigan State (4-0, AP #2): 1.5

Michigan State's win over Oregon looks much less impressive after their dismantling by Utah, and the polls are likely overrating the Spartans. Meanwhile, injuries are starting to take their toll. But this is still a good team by Big 10 standards--they should end with a winning record in conference play.

3. (tie) Michigan (3-1, AP #22): 1.5

The loss against Utah is no longer working against the Wolverines, while the shellacking of BYU accounts for all of Michigan's points. Upcoming games against Northwestern and Michigan State will show us how far we've come in Jim Harbaugh's first year.

3. (tie) Minnesota (3-1, AP NR): 1.5

Another surprise for me--I don't personally find Minnesota all that impressive, but the loss to TCU doesn't really hurt them, while the road win over Colorado State scores well.

6. Iowa (4-0, AP NR): 0.5

A lot of people are wondering how good Iowa is. This system isn't terribly impressed so far. If they beat Wisconsin, though, that's another story...

7. Wisconsin (3-1, AP #19): 0.0

Wisconsin looked respectably enough in the loss to Alabama, but has only played cupcakes since--a fact that keeps its score low.

 

8. Indiana (4-0, AP NR): -1.0

For the record, this is as bad a score as a 4-0 team can get. Fall back to Earth imminent.

9. Maryland (2-2, AP NR): -2.0

I expected Maryland to grade out worse than this, frankly. But they are clearly trending downward, and with a hurricane and "weather-proof" Michigan team both coming to town this weekend, the death spiral may come fast.

10. Illinois (2-2, AP NR): -2.0

I'd say the same thing about Illinois, but a shockingly easy schedule could help the Illini stay in fighting distance of bowl eligibility, if not quite make it.

11. Penn State (3-1, AP NR): -2.0

The system doesn't think much of the Nittany Lions, and neither do I.

12. Nebraska (2-2, AP NR): -2.5

It's more surprising to see how badly Nebraska fares. I mean, they are not exactly good, but I'd still expect them to beat Maryland, Illinois or Penn State. Losing to BYU and failing to blow out Southern Miss at home hurt though...

13. Purdue (1-3, AP NR): -4.5

Purdue is bad. If they don't pull off a major upset or two, Hazel will be out of a job before the season ends. Apparently they think they have a plan for Michigan State. Somehow I doubt it, but okay.

14. Rutgers (2-2, AP NR): -5.0

LOL.

 

Summary Statistics

  • The mean score for all 14 Big 10 teams is -0.6 and the median is -0.5.
  • The range is 3.5 - -5.0.
  • There are 6 teams with positive scores, one with a score of 0 and 7 with negative scores.

 

 

Comments

wolfman81

September 30th, 2015 at 9:11 PM ^

So how do you define good/solid/not good?  My feeling is that it is purely subjective.  Could you comment on this definition a bit?

An idea for how I might try it within the spirit of your model:  Develop an iterative algorithm.  For the first 4 weeks use an "all teams are solid" definition.  At this point, you have (some) data which you can base future rankings on.  Start by defining all 4-0 teams as "good", 3-1 teams as "solid", and all other teams as "not good".  Then use these definitions for the 5th week rankings.  From that point, you can use your rankings to set good/solid/not good definitions for the next week.

Eye of the Tiger

September 30th, 2015 at 9:35 PM ^

Based on the same methodology--who they've played, what the results were, etc. When in doubt (about, say, lower level teams), used advanced stats to make judgement calls. There's still an element of subjectivity, but it's minimized.

One thing I probably should have noted is that, according to this methodology, only Northwestern, Ohio State, Michigan State and Michigan would qualify as "good" right now. Minnesota too, I guess, though I really don't think they are actually good. 

Also, I didn't want to just judge by record, as not all 4-0 records are the same. Michigan's 3-1 record, I think, is much more impressive than Indiana's 3-1. Similarly, this system "prefers" 2-2 Maryland and Illinois over 3-1 Penn State.

Broader point: as I'm sure you are aware, it isn't possible to be fully objective in analysis of any social phenomenon, even when utilizing statistics (there are always "judgement calls" involved). The exercise here was to *reduce* feelingsball type subjectivity and produce a power ranking system that isn't dependent on preseason/last season biases.




Sent from MGoBlog HD for iPhone & iPad

NittanyFan

September 30th, 2015 at 9:43 PM ^

I think your method would benefit from SOME sort of more rigorous and systematic way of defining things.  As it stands now, "large/small" and "solid/good/not so good" seem to be defined via the method of "after watching 4 weeks of football, I think this feels right."

If you created a formula that was 1/4 a team's rank in per-game scoring margin, 1/4 a team's rank in winning percentage, 1/4 a team's rank in strength of schedule, and 1/4 your pre-season rankings (this last quarter eventually diminishes away as more weeks are played) ---- and then said "good" are teams that are in the top 20% of this formula, "solid" teams are the next 20%, et cetera ----- that may not be perfect, but at least it's a systematic definition.

ChuckieWoodson

October 1st, 2015 at 9:39 AM ^

I like what you're doing.  My only suggestion would be (if you wanted to...) make things a bit more empirical, instead of using  your own rating system for the wins, just pick a poll (AP, coaches, whatever) and base it off of that at the time of the win.  But, just my .02 - pretty cool even with your relatively subjective rating system of the win or loss.  I like the approach.

beenplumb

October 1st, 2015 at 10:33 AM ^

If you've got MSU, UM, and Minn tied for #3 (scores of 1.5), shouldn't Maryland, Illinois, and PSU be tied for 9th, since they all have the same score (-2.0), or am I missing something?

Realus

October 1st, 2015 at 11:28 AM ^

It's good to get this point of view.  Thanks for working on it.  I hope you have the chance to update this every 2 weeks but even every 4 weeks will be pretty cool.  I can see where this is very time intensive.

Nosce_Te_Ipsum

October 1st, 2015 at 11:29 AM ^

I love this idea. Nice work. My critiques of it would be that you give a more detailed explanation for each team showing how they received their scores. Example: Michigan gets however many points for their dismantling of BYU at home. Listing the games with the point values gives it much more depth. My last point would be that you have Minnesota as 4-0 even though you state their loss to TCU. Great work. Looking forward to seeing the next one.

Wolverinefan84

October 1st, 2015 at 11:34 AM ^

That win over Stanford looks impressive but they looked absolutely terrible in week 1. Don't know if that's a product of having a ton of new starters on D or playing a noon game with a 3 hour time difference. But nonetheless their recent success definitely makes that game look better for NW

I Just Blue Myself

October 1st, 2015 at 2:56 PM ^

Good diary. One suggestion: put how you came to the points for each team. 

For example, you gave MSU 1.5 points, so judging by how you're scoring they would have gotten 0.5 points for a road win at Western, I'm guessing 0.5 points for a small win over a "good" Oregon, and somehow received 0.5 points for a small win over a "good" Air Force OR Western OR Central?

And with Michigan, shouldn't UM have 1 point for a win over a solid BYU, then an additional 1 point for it being a "large" victory?

FWIW, I'm an idiot so I may have misunderstood your post. I generally only skim when reading.

kevin holt

October 2nd, 2015 at 2:31 PM ^

Seems like a small win over a "not good" team is -0.5, so maybe they got a negative or two in there?

Edit: I'm guessing (assuming Oregon is "good"), 2 for a win over Oregon, plus 0.5 for road win at Western, plus 0.5 for small win against "good team" is 3 total. Then -1.5 for small wins over "not good" teams? That doesn't make sense because one of their wins was by 20, which is defined as a blowout in the methodology, not a small win.

More likely he's considering Oregon "solid" but not "good." That would be +1 for a win with 0 for the margin, then +0.5 for the road win against Western. Add that to wins against "not good" teams (all 0 pts) and you have 1.5.

funkywolve

October 2nd, 2015 at 2:58 PM ^

By your scoring system I see:

Stanford is +.5 (small home win over a good team)

Eastern Illinois is 0 (large win at home over a not good team)

Duke is 1.0 (small road win over a good team)

Ball St is -.5 (small home win over a not good team)

 

That adds up to 1.0

Eye of the Tiger

October 4th, 2015 at 3:20 PM ^

Thanks for the constructive criticisms. I'm making a few changes to the methodology for this week's entry--reflecting some of the suggestions made. Nothing too earth-shattering, but with the goal of making the process more transparent, replicable and reducing subjectivity further.