Our Defense, Their Offense - Judgment Call
Iowa Post-mortem
Three weeks ago we predicted the MSU yardage exactly and modeled the score very closely for what would have been in a turnover neutral game. Two weeks ago vs Iowa we had this:
Dear god, help us.
It's about time for our defense to turn in a performance that exceeds expectations, and when these are your expectations one can only pray that they are exceeded.
If they are, Michigan might have a shot in this game - but only if our offense actually puts together a real game when it counts in the Big Ten.
Michigan 28
Iowa 38
Iowa did gain fewer yards than predicted, thank god. I feel like this was mostly due to conservative play-calling on Iowa's part. It was clear that when they desired greatly to move the ball, they were able to pull out the stops and march down with little resistance. But their coaches saw a path to victory that involved simply out-executing and took it.
Notes:
- Turnovers killed.
- If turnovers didn't kill, special teams and penalties killed.
- The defense held Iowa to less yardage than expected but that could possibly be due to Iowa's average field position (40 yard line anyone?), our turnovers, and conservative play-calling from Iowa once they had a huge lead.
Enough emotional prognosticating, here be the numbers:
Chart of Expectations (through 7 games)
Rank | Opponent | N-PPG | N-YPG | SoS |
1 | osu | 36.2 | 399 | 67.18 |
2 | Michigan | 32.4 | 479 | 68.21 |
3 | Wisconsin | 32.3 | 384 | 68.76 |
4 | Iowa | 31.5 | 386 | 70.89 |
5 | MSU | 30.9 | 402 | 67.90 |
6 | Indiana | 25.1 | 336 | 61.13 |
7 | Notre Dame | 25.0 | 381 | 73.14 |
8 | Illinois | 24.4 | 330 | 75.74 |
9 | Connecticut | 24.2 | 298 | 65.22 |
10 | BGSU | 21.2 | 255 | 65.65 |
11 | Penn State | 18.8 | 313 | 70.12 |
12 | UMass | 18.1 | 312 | 56.04 |
13 | Purdue | 17.1 | 290 | 67.10 |
Metrics
N-PPG or Normalized Points-per-game is taken from the teams average PPG with a SoS multiplier factored in to deflate numbers from playing bad competition and inflate numbers based on playing good competition.
N-YPG or Normalized Yards-per-game is calculated using the same SoS multiplier as N-PPG but using this metric will help us determine a less variant guess as to how offenses will perform (PPG is subject to wild variance based on turnovers and special teams).
Strength of Schedule is taken from Sagarin rankings.
Usage: The chart doesn't predict that #3 would beat #5. Instead it tries to predict with the most accuracy how many points/yards on average each of these teams would score against a common opponent.
Results
Surprisingly boring results after a bye week.
- We have yet to play the best two N-PPG offenses we will face this year.
- Our other three games are against three of the bottom half, including arguably the two worst.
- I really hope that at some point we start converting the yardage gained disparity into the points column.
- But it could mean that we just have awful special teams and killer mistakes that make us have to move the ball a lot farther for every point we get.
- But at least we're moving the ball!
- All of our wins are looking cheaper by the week:
- Indiana is tied for last in the Big Ten.
- BGSU is last place in the MAC.
- UConn is last place in the Big East.
- UMass got rolled twice by some terrible Colonial teams.
- Notre Dame got blown out by Navy.
Conclusions Based on Saddening Data
What a big, big opportunity for the Maize and Blue this weekend. Penn State took a gracious Minnesota team to bring up it's averages (Minnesota fwiw would be a 22.1 NPPG team and 351 NYPG on the chart - better than BGSU, worse than UConn).
The yardage doesn't lie, Michigan has a potent - albeit highly volatile - offense. And the fact that the team has an great opportunity to improve from last years mark with a defense that's the worst we've ever seen is very encouraging if you believe that defense will eventually get better.
Best Case:
Michigan's new best-case scenario (Michigan offense operates at or near 100% N-PPG and opponents score 125%) in the Big Ten is 4-4. That means the next three games are must win to stay at .500 in Big Ten Play.
Worst Case
In a worst-case scenario our only remaining win is Purdue.
The Truth!
Our new outlook ranges between 6-6 and 8-4!
Prediction for Penn St:
Based strictly on the numbers:
Team | PPG vs Mich | YPG vs Mich |
Penn St | 23.5 | 391 |
Michigan 28
Penn State 24
GO BLUE!
October 29th, 2010 at 10:07 AM ^
...and congrats(?) on the Iowa score prediction. I'll take the PSU prediction (obviously), but it's still a little too close for comfort. The negative effect of Michigan's swiss cheese defense is clearly on display here.
In the end, the game will hinge on winning the turnover battle. I don't see how PSU can stand up to Michigan's defense unless the visitors shoot themselves in the foot.
October 29th, 2010 at 10:11 AM ^
Does your analysis take into account the relative strengths of rushing and passing offenses and defenses? Here's the reason why I think that 28 is too low. Penn State's rushing defense, statistically, is similar to UConn's and ND's. This is probably due to all of the injuries that they have had (and some that they still have). Regardless of the reason, if Scheelhausse and co ran for 280 yards on them and scored 33 points, I've got a feeling that we'll pass that easily.
October 29th, 2010 at 11:29 AM ^
Agreed. I would very much like to hang 50+ on PSU just to blow them out and make them feel bad. That of course is contingent upon our defense being able to stop them from scoring nearly an equal amount, but, if there' s any team and any week I have a shred of confidence that our D will not be absolutely terrible, it's against this team this week.
October 29th, 2010 at 11:25 AM ^
Just wondering, but what's with the 150% bonus you gave Iowa and MSU vs. the 125% you gave Penn St? I realize it's because they're obviously better, higher scoring offenses, but isn't that number (150, 125) a bit arbitrary? Or is there some statistical precendent you're basing that off of?
Otherwise, great analysis. I always enjoy reading people's individual approach to predicting these matchups.
October 29th, 2010 at 12:13 PM ^
It's because of the way that I calculate N-PPG, it requires me to put in an additional calculation (multiplier) to take into account the caliber of our competition's offense vs the "normalized" offense.
I've arrived at the 150%, 125%, and 100% modifiers after calculating what Michigan traditionally yields vs N-PPG over the course of the past two seasons.
Comments