No B1G Teams in Power Rank Top Ten

Submitted by Ecky Pting on

Check this out: the Power Rank has no B1G teams in the top ten. Yet, five teams from the state of Florida are in the top 25, including a very curious one at the top spot! It seems the disparity in games played between the FL teams and everyone else is wreaking some havoc with the fancy stats algorithms...

Rk Team Rating
1 UCF 20.67
2 Clemson 19.93
3 Oklahoma 19.52
4 Alabama 18.84
5 Georgia 17.02
6 Washington 16.03
7 TCU 15.70
8 Notre Dame 15.24
9 Virginia Tech 14.92
10 Auburn 14.75
11 Ohio State 14.30
12 Wisconsin 13.26
13 USC 12.96
14 Michigan 12.55
15 Stanford 12.39
16 Miami (FL) 11.91
17 Penn State 11.54
18 Oklahoma State 10.58
19 Texas 9.84
20 Florida State 9.62
21 Georgia Tech 8.96
22 Louisville 8.68
23 Florida 8.60
24 Duke 8.59
25 South Florida 8.54

 

ColeIsCorky

September 26th, 2017 at 10:01 PM ^

That "argument" was Ed and Brian differing on language more than concept. Brian offended Ed by saying a statistic had no purpose, and Ed, being a statistics guy, argued that it did and took offense to how he worded it. But in conclusion, they both agreed that this particular statistic had no real meaning in and of itself. They argued for 5 minutes with each other about something they knowingly agreed upon in context.

It was quite hilarious. Brian basically enjoys irritating Ed, and Ed has a hard time not becoming defensive. Brian's sense of humor basically tortures Ed's sense of being.

MGoViso

September 27th, 2017 at 10:08 AM ^

I listened, and you're right about how they "resolved" the dispute. However, it seemed like they actually still had a fundamental disagreement. It didn't quite get worked out because (1) Brian used hyperbole to emphasize his point that Ed misinterpreted, and (2) Brian didn't explain his point fully.

RE: (1), Brian claimed that "red zone offense" doesn't exist, meaning that a team runs its offense regardless of field position. He admitted that this was wrong because short yardage "heavy" plays and short fades play a disproportional role in red zone offense, so a correct statement would be "red zone offense is an overblown idea."

RE: (2), Brian's evidence consisted of a study on NFL teams that found early-season red zone success had almost no predictive power for later-season red zone success. That is, unlike many other metrics (like completion percentage or yards allowed per attempt) where you can mostly expect what has happened to continue happening, red zone offense is more like turnovers, where you shouldn't necessarily expect what has happened to continue happening.

Ed didn't grasp either (1) or (2), but I don't blame him because Brian's (good) points weren't communicated well. Ed was focused on the fact that if you can measure it, it matters.

Mike Damone

September 26th, 2017 at 5:36 PM ^

at #1 is ignorant.  But putting Notre Dame in the Top 10 is stupid AND offensive. Who is Power Rank?

My name is Ed Feng, founder of The Power Rank.  I use my Stanford Ph.D. in applied math and my own algorithms to make football and March Madness predictions. To get a sample of my best predictions, sign up for my free email newsletter. 

I ran my own analysis and algorithms.  The data clearly shows that Scott Frost and Brian Kelly are loudmouth blowhards, and that Notre Dame will lose at least 4 more games, 2 in choking fashion.  Subscribe to my newletter - Ed Feng and Power Rank are big piles of dogshit...

Profwoot

September 26th, 2017 at 5:23 PM ^

It's really easy to create a ranking algorithm. It makes no sense to draw attention to one of them unless there is any reason to think it has some special utility, particularly at this early stage when even the good ones are completely unreliable.

If it's due to a more general fascination with failed analyses, You'd love the endless blocks of poorly-conceived and buggy-ass code I've accumulated over the years.

bronxblue

September 26th, 2017 at 5:53 PM ^

Yeah, UCF at the top should be a bit of a giveaway that the numbers aren't quite representative for these models to make much sense.  Oh, and Notre Dame in the top-10 is also a dead giveaway.

Again, it's the same problem S&P has early in the year.  FSU is 0-2 and #3 because of pre-season expectations and scheduling quirks due to Irma.  I won't slag on Ed because he's just showing you what the model spits out, but it's why I don't take him as seriously when a couple weeks into the season he's touting his models.

Mr. Yost

September 26th, 2017 at 6:02 PM ^

This makes no sense whatsoever.

Here are my power rankings....please create a thread about them.

1. Michigan

2. Eastern Michigan

3. Central Michigan

4. Western Michigan

5. Northern Michigan

6. Detroit

7. Grand Valley St.

8. Oakland

9. Wayne St.

10. Ferris St.

Others Receiving Votes: Huron High School, Pioneer High School, Little Caesar's Pee Wee

Not Listed: Michigan St.

 

...there, post that. It's just as accurate as this bullshit

Leaders And Best

September 26th, 2017 at 6:11 PM ^

UCF would also be in the top 2 in the S&P+ if preseason projections were removed.

UCF is going to be an outlier because they have only played 2 games due to the hurricane and one of them was against Maryland with their 3rd string QB and the other against FIU. That is a double whammy because Maryland's numbers are going to off too until they play more games with their QB situation.

NittanyFan

September 26th, 2017 at 7:05 PM ^

They aren't connected to many other FBS teams.  But of those they ARE connected to - they are on the top of a "chain" that makes them look pretty good:

(1) UCF walloped Maryland on the road - Maryland has no other losses and won at Texas - Texas nearly beat USC on the road - USC is undefeated including a 2-score win at 1-loss California - California has wins over UNC and 1-loss Ole Miss.

If an algorithm necessitates that UCF must be better than all those teams (and why not, if we're only going by empirical results?) - you're going to get UCF being pretty-high overall, given they'll also need to be higher than all the teams Maryland/Texas/USC/Cal/UNC/Ole Miss are better than (all those teams have a lot more results on the books).

UCF does play Memphis this week, so that should get them much better connected.

Hold This L

September 26th, 2017 at 6:18 PM ^

I remember when BLEACHer Report had a decent comment section and collectively all fans went in on this idiot that thought a single, subjective person can come up with made up equation to determine who the best team is when he has no experience in football at any level or position. Probably uses PFF to help him determine who the best players are, like that illinois wideout who has "the most yards run per route". Groundbreaking analytics. 

gremlin3

September 26th, 2017 at 6:57 PM ^

The math is perfectly sound, but the stats are only as good as the assumptions, and that's where the problems with almost all football analytics lie.  S&P is, in my opinion, the best at those and I find those rankings to match up most with the "eye test."  (Yes, I realize this is somewhat of a confirmation bias.)

Basically, Ed needs to rethink his assumptions.

KTisClutch

September 26th, 2017 at 7:31 PM ^

Kind of harsh on Ed to get post this without context. As another poster said, a lot of efficiency based rankings will have UCF really high right now. When they play some more games and Maryland falls back to earth post-QB injuries they will come down.

UMmasotta

September 26th, 2017 at 8:07 PM ^

I've never looked at the Power Rank before, but it's pretty obvious that Ed doesn't actually think UCF is the best team in the country. Which is why the OP should have also pointed out that the blog post to accompany these rankings explains why the model spits out UCF as #1.

 

Blog post here: https://thepowerrank.com/blog/

 

The same thing happens with S&P+, as anyone who listens to Bill Connelly's podcast, PAPN (Podcast Ain't Played Nobody) would tell you. What the model spits out is only good if you understand the inputs and can interpret the results accordingly. Also, neither S&P+ nor The Power Rank (as far as I know) claim that the current rankings are predictors of final rankings. They are indicators of performance to date, based upon data from small populations (especially in the case of UCF, as they only have 50% of the data that most other teams have). 

 

/end rant.

ColeIsCorky

September 26th, 2017 at 10:07 PM ^

Exactly. The Power Rank is actually a very clever tool used to predict games and so on and so forth. It is a unique way of compiling statistics.

In comparison, you might see this is as something similar to one of the "computer rankings" from the old BCS system. It's not that, but you can think of it as that. Not Ed's personal rankings. It's used a lot for game betting and the like. And is very good at that.

Ecky Pting

September 27th, 2017 at 4:11 PM ^

It was mentioned by the OP that the weather-abbreviated schedules in certain regions of the country had an influence on these results, which is elaborated upon by Ed in the blog post you cited. I thought the obvious anomaly might engender a thoughtful discussion of the advanced analytics, and was not intended to be dismissive of Ed or the whole premise of doing analytics. Some folks see it as a challenging problem to solve, while others see it as a waste of time and throughput. For the latter, I'm not sure why they're reading this in the first place.