Soliciting your takes on Net Success Rate vs. PSU

My honest assessment is that when Michigan is at the top of this chart it's a Very Good Metric and when they're not it's for Nerds who Don't Know Ball.

We went against the best rushing defense in the entire country and didn't throw the ball for the ENTIRE SECOND HALF.  National pundits / stats guys who don't recognize this for it's absolute awsomeness, are morons.

OSU should be fucking terrified.

Thanks for the information!  It is amazing how all kinds of questions can arise from one game (especially one that we won in a pretty dominant fashion).  My take is that most every non-Michigan fan will use the lack of Michigan offense in the PSU game to hype up OSU as the favorite in The Game.  This would be completely ignoring OSU's inept offense (outside of M. Harrison) all year long against good defenses.  Michigan has been prepping for this OSU game all year long and has about the most balanced offense they have had in years.  I believe the OSU game is going to come down to whether or not OSU can run the football.  Henderson has looked a lot better in the 2nd half of the season and I believe we are going to run a pretty conservative defense to not give up the big plays.  Keys of the game....our D-Line vs stopping the run.  If we can do that while having our secondary stay back in zone looks, we will win this game.  I believe it will be close no matter who wins though so get your heart meds ready.  Go Blue!

I've spent way too much time worrying/thinking/ruminating about success rate - especially net success rate - over the years.

First things first:  Fleming is an EPA guy, which means he measures success rate (as I understand it) not based on percentage of yardage gained towards first down per down, but instead by whether or not a play generates more (a success) or less (a failure) than the expected amount of baseline EPA added at that down and distance.  These are not the same measure, although they'd be roughly close.  Suffice it to say, however, Fleming's success rate measure slightly favors teams that gain consistent yardage down-over-down versus the "percent to gain" approach.  This, I think, means EPA-measured success rates will slightly penalize Michigan versus percent-gain approaches because it's obvious that we're more than happy to get 2 yards if it's 3rd and 1.  That's a success for a percent-gain measure, but may well be a failure for an EPA measure if team averages tend to overshoot that measure on similarly-situated 3rd downs.

Second:  there is no garbage time in the Michigan PSU game, which Fleming confirmed on Twitter.  This means we really took it hard for the purposes of net success rate during PSU's last few drives, because we were trying to incentivize small, consistent gains from PSU (preferentially by running the ball to keep the clock moving) in order to protect against big explosives.

Nerds just read spreadsheets and ball knowers understand game state.

JJ was 7/8 with successful scrambles and a PI drawn on his one downfield throw. We could have passed the ball. However, passing just allowed PSU to have the only clear advantage they had in the matchup -- their edge rush against our tackles with a loud crowd helping them. Moore took away that advantage from them, by simply not passing the ball.

Could OSU give our tackles problems? Certainly. But, PSU (and Purdue to some degree) did us the "favor" of showing us our weakness. We have multiple weeks of practice to address the issue now. I'm guessing some TE help and/or max protect is worked in if needed.

On the flip side, OSU people had been focusing on Michigan not being the same level of running attack that mowed them down the last two years. I'm guessing that plowing the #1 D rattled some cages over there. Michigan has now shown they can do everything. 

We're not going to pass the ball only 8 times against OSU.  This game should be taken as a case sui generis and not much of a preview of that one.  The only big takeaway I have is that it demonstrates that we're capable of running on a strong defensive front if necessary. 

It's pretty easy to think of ways that a team can control the game on both sides while coming out even on a metric like this:

• Offensively: You throw relatively few passes, meaning you miss out on a lot of high-percentage 5-to-10-yard plays that count as a Success here. And you grind out yards and scores on the ground through a lot of 2/3-yard runs, a few really big chunk runs, and few negative plays. The end result is a winning number of long drives and scores, but relatively few plays that count as a Success.
• Defensively: You give up a lot of modest gains on the ground, but you rarely bust, and you make a lot of drive-wrecking negative plays. Again, meaning that your defense gives up a good number of Successful plays, but still gets a winning number of stops.

This sounds a lot like how Michigan played against Penn State, and how they're built to play against good teams generally.

I think this chart is correct. I did not have the happy cozy feeling that we were running the ball at will during the game. I felt like PSU was stopping the vast majority of our stuff and we broke a few when they made mistakes. The rosy view of that (see earlier front-page post) is that we planned on that and took advantage because that's what PSU gave us. 

We had 3 TDs, 1 FG, 5 punts, and ended both halves with the ball. It wasn't the domination some were predicting, that's for sure. If those two fumbles go the other way (or if Frames isn't coaching), things may have gone a lot worse.

Anyway, if that was the plan, it worked and we won. But let's not pretend we averaged 5 yds/carry because we got 5 yards each carry. 

As much as I love fancystats, there is no one fancystat that works better than SCOREBOARD at predicting victory.  SCOREBOARD says that we were up 24-9 on Penn State with 4 minutes left and then gave up a touchdown in garbage time due to prevent defense, mistakes, and penalties.

Running 32 times in a row is not going to work vs OSU.  But I think we showed in last year's game that we can adjust to whatever their defense is trying to do.  We bombed them through the air in the first three quarters and on the ground in the fourth.

To me, this year's game is going to come down to which team can pressure and hit the other team's QB most consistently.  Either offense can score 50 if they have the time to execute their plays.  Barnhart is going to have to have a great game for him

That's actually not the formula the Success Rate that Parker uses. He considers any play with a positive EPA to be a successful play for an offense. (You see EPA on Seth's defensive UFRs; in that case you want to see negative EPAs as a defense.)

This chart shows the *net* difference between Michigan's success rate and Penn St's success rate. For example, if Michigan's offense had a positive EPA on 50% of its plays and PSU's offense had a positive EPA on 45%, then Michigan has a 5% Net Success Rate. (I don't know what the real numbers were, but it's different than what Game on Paper uses because it shows Michigan at 41% and PSU at 42%.)

No, we didn't really get that bad.

We just won in a way that doesn't look good to this particular fancystat.

This chart doesn't account for HOW successful the play was, just IF the play was successful.

So, Corum and Edwards' explosives are just "yes" instead of "no" here.

PSU got LOTS of "yes" plays on first down...then got stuck on 2nd or 3rd down.  They were probably getting a fair number of drives with more "yes" than "no" plays.

Michigan got very few "yes" plays on 1st down and often found a way to convert 3rd down.  We probably had several successful drives that were barely even (if even at all) on "yes" vs. "no."

As others have noted, this looked even worse when we trounced OSU last year.

We didn't take any risk on offense, just did what we needed to do. That's not how we'll play Ohio State, so it doesn't make much sense to extrapolate how that game will go based on net success rate with a different game plan