ESPN Team Efficiencies Update, 2017 Week 2

Submitted by Ecky Pting on

The updated ESPN team efficiency ratings have been posted, and as you might expect after managing to not lose to Cincinnati, Michigan drops 12 spots to #31 overall. So, this makes even more room for improvement than M had after beating Florida. The U-M splits are offense, #68 (down from #59); defense, #12 (down from #3); and special teams, #65 (down from #37). Mediocre in all be defense, which itself is looking markedly less elite than it had initially.

Of U-M's other 2017 opponents, MARYLAND now looks the strongest at #6 overall, followed right behind by Penn State at #7. Next week's matchup, Air Force, is holding up at #15, which constitute M's strongest opponent to date by a wide margin, since the Gators pull up at a perfectly mediocre #69. After its loss, OSU dropped as well, just not as much as M, and so the edge M once had on the Buckeyes is no more. Oh yea, Sparty climbs 18 spots to #34, just 3 behind M to reclaim its long lost status of Best Little Brother Team in the State of Michigan. Don't look now, but Sparty has both a more efficient offense and defense than Michigan. Meanwhile, closer still is Indiana at #33.

Overall, 2 of Michigan's opponents are in the top ten, 4 are in the upper quartile of the ratings; all but 3 are in the upper half; and only one (Rutgers) is in the lowest quartile.

Here's a brief synopsis of the ratings, or click the link above to see the whole list &/or sort by category:

RK
TEAM
OFF
DEF
S/T
O'ALL
1
Oklahoma, Big 12
97.0
88.1
18.9
96.2
2
Miss St, SEC
72.2
98.5
72.4
95.8
3
Alabama, SEC
82.2
90.3
91.6
95.2
4
94.7
89.6
28.0
95.1
5
LSU, SEC
91.5
88.1
39.5
93.9
6
Maryland, Big Ten
96.0
77.8
66.8
93.4
7
Penn State, Big Ten
74.1
88.4
88.6
92.4
8
Duke, ACC
83.5
94.7
19.9
91.0
9
85.5
84.1
58.4
90.8
10
TCU, Big 12
74.5
85.0
88.7
89.7
15
88.4
72.4
34.5
85.8
23
OSU, Big Ten
70.4
71.3
73.8
79.1
31
Michigan, Big Ten
48.5
90.0
53.2
75.8
33
Indiana, Big Ten
63.3
64.1
76.7
72.2
34
Michigan State, Big Ten
50.9
90.9
23.2
72.1
38 Notre Dame, Indep. 60.1 72.7 52.6 70.0
40
Purdue, Big Ten
73.7
69.6
17.4
69.4
41
Wisconsin, Big Ten
55.6
80.0
39.0
68.2
45
Minnesota, Big Ten
51.0
71.0
65.8
66.2
66
56.8
40.9
76.2
54.6
69
Florida, SEC
12.2
84.3
35.1
53.4
74
34.2
74.8
30.7
51.1
92
38.5
41.8
44.5
36.8
96
Cincinnati, American
27.2
45.3
57.1
35.0
102
Rutgers, Big Ten
23.2
56.0
31.0
30.9

 

Ecky Pting

September 11th, 2017 at 3:44 PM ^

per the ESPN references:

Team efficiencies are based on the point contributions of each unit to the team's scoring margin, on a per-play basis. The values are adjusted for strength of schedule and down-weighted for "garbage time" (based on win probability). The scale goes from 0 to 100; higher numbers are better and the average is roughly 50 for all categories.

The Football Power Index (FPI) is a measure of team strength that is meant to be the best predictor of a team's performance going forward for the rest of the season. FPI represents how many points above or below average a team is. Projected results are based on 10,000 simulations of the rest of the season using FPI, results to date, and the remaining schedule.

My take regarding the Efficiency Metrics is that in general, the longer it takes a team to reach garbage time, the lower the efficiency score. That said, in the case of M's game with Cincy, it's damn near impossible for M to raise its efficiency aggregates since it's already starting out at a win likelihood in the high 90's. However, it's really easy to bring efficiency aggregates down, by screwing around in the 2nd and 3rd quarters and generally doing things that end up looking like this:

Mv_Cincy_winpct

NittanyFan

September 11th, 2017 at 4:28 PM ^

however - looking at your chart Michigan has an (approximately) 96% chance of winning from the game's opening snap.  Which, sure, Michigan is simply a better football team than Cincinnati.  

But that means the "a priori" difference between teams is baked into the win probability metric.  If I'm reading things right, their algorithm could be construing the entire first half of the game of the UC game as "garbage time" for Michigan, given their very high win probabilities?

A ratings algorithm should discount "garbage time" drives, of course.  But the first half (and 3rd quarter) of the UC game was far from "garbage time."  

If that's what FPI is doing, that's a very fundamental flaw.

Ecky Pting

September 12th, 2017 at 1:15 PM ^

Yea, I hear ya. It's unclear exactly how the ESPN Effciency scheme applies the concept of garbage time. What is known is how the concept is applied in equivalent points schemes such as S&P+, which defines the garbage time constraint as:

...a score differential within 28 points in the first quarter, 24 in the second, 21 in the third, or 16 in the fourth -- either when a team is winning or losing.

By this definition, the MvCincy game did not enter garbage time until the last pick-6 with 4:21 remaining, which is less than 10% of the game time left.

Getting back to the win probability basis, since the initial probability has UM as a heavy favorite, the win probability definition of garbage time would seem to start at some likelihood even higher than that, and increase still further over the 60 minute game time (so we're talking about probabilities ranging from 99% to 99.99% to reach garbage time given the intial conditions). With that fuzzy notion in mind, and overlaying it onto the chart above, UM didn't reach garbage time until ...sometime late in the game? In the end it means that all that putzing around on offense by Michigan actually counted in terms of fancy stats, and counted in a deleterious way. This is what I was trying to get at by suggesting it's easier to do harm to a team's efficiency score in games like this than it is to improve it.

 

814 East U

September 11th, 2017 at 3:25 PM ^

Here are the things I learned in week 2 in college football:

1) Rutgers is still Rutgers;

2) It is hard to plant flags on artificial surfaces;

3) Brian Kelly is still an asshole. However, I did not learn if asshole is Misogynistic.

DrMantisToboggan

September 11th, 2017 at 3:25 PM ^

Two weeks is an entirely insufficient amount of data. I guarantee this team finishes top 5 in defensive efficiency and top 20 in overall efficiency. 

 

Also, all models are flawed, and S&P+ metrics are more telling than ESPN's.

FauxMo

September 11th, 2017 at 3:30 PM ^

Based on my data analysis, I predict Michigan will score 34.5 points per game this season. My model suggests that there is only a tiny error margin to this prediction, so I predict that this prediction will be virtually perfect.  

#BadStatsSmallSamplesFuckYeah!

Sambojangles

September 11th, 2017 at 4:09 PM ^

As long as we're discussing too-early ratings systems, Michigan dropped from #3 to #4 (jumped by Oklahoma) in the Massey Ratings system. We are a favorite in all remaining games, with the closest being PSU (51% likelihood of winning). 80%+ in the four games between now and then.

Massey

Ty Butterfield

September 11th, 2017 at 4:21 PM ^

Curious if anyone has watched both Staee games so far this season? They don't seem to be the train wreck we thought they might be, but their opponents haven't been that great either.

Blueblood2991

September 11th, 2017 at 4:24 PM ^

I'll wait until they play some better competition, but they do seem to be much improved compared to the team that barely squeaked by a 3-8 FCS team in Furman last year. Lewerke seems like he has some high potential. I think depth is still a huge issue for them though. Injuries are unavoidable.

I'm still crossing my fingers for the collapse though.  Long live the 3-9 jokes!

corundum

September 11th, 2017 at 5:01 PM ^

I watched both games. Bowling Green would have been a dogfight if their QB wouldn't have overthrown at least three blow coverages, and they just lost to South Dakota. MSU is going to be slightly improved because Lewerke makes them two dimensional and is a legitimate run threat. He reminds me of Drew Stanton with less agility or a Great Value Jake Locker. They certainly aren't good and have no chance to compete for the division. I'm thinking 6-6 is likely but 7-5 is the ceiling. They will probably get to a 13th game and play a MAC team in a shit-tier bowl.

ST3

September 11th, 2017 at 5:56 PM ^

I tried watching them this weekend, but something was wrong with my BTN HD feed. It was only right before half that I remembered I have a BTN SD channel. Being forced to watch MSU in SD seems about right. I think last year was an aberration and they will at least be bowl eligible this year. I'm not wandering any further out on that limb.

AmayzNblue

September 11th, 2017 at 4:58 PM ^

At 2....lulz.



Totally insufficient data considering teams like Miss St have played only cream puffs.



Also...Michigan has only a 90.0 rating on defense. Schwaaaaat??