Gauging Team Effectiveness: Part II: San Diego State
The reason I wanted to adjust the results for quality of opponent was to try to account for SDSU’s #88-ranked schedule versus Michigan’s #23-ranked schedule. The best I could come up with was: average the offensive PPG with the defensive PPG, then take that “projected score delta” and subtract it from the actual score delta to get a pseudo-PAN (to borrow the Mathlete’s term). Don't worry, there will be an easy-to-understand number at the end.
Offensively, SDSU averaged 455 YPG and scored 35 PPG. That compares pretty closely to Michigan’s 488 YPG and 32 PPG. So right off the bat the offense looks to be a little more “efficient.” More points + less yards = good. Defense is a whole different animal. SDSU allowed 355 YPG and 22 PPG, compared to Michigan’s 455 YPG and 35 PPG. They have a defense, we did not. SDSU is still capable of scoring lots of points, scoring over 40 5 times, and over 30 in 9 games. Again, number of drives is pretty variable, from 9 to 18. For the season, SDSU actually averaged more drives per game than Michigan, at 12.3 to 12.0.
Since we established that the raw drive data is pretty pointless yesterday, I’ll skip it and get right to YPD. Basic data and Chart:
| opponent | yards | drives | points | ypd | ppd | d-yards | d-drives | d-points | d-ypd | d-ppd | net ypd | net ppd |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| NSU | 531 | 10 | 47 | 53.10 | 4.70 | 171 | 11 | 0 | 15.55 | 0.00 | 37.55 | 4.70 |
| NMSt | 563 | 16 | 41 | 35.19 | 2.56 | 328 | 14 | 21 | 23.43 | 1.50 | 11.76 | 1.06 |
| Mizzou | 440 | 18 | 24 | 24.44 | 1.33 | 440 | 15 | 27 | 29.33 | 1.80 | -4.89 | -0.47 |
| UtahSt | 504 | 11 | 41 | 45.82 | 3.73 | 245 | 12 | 7 | 20.42 | 0.58 | 25.40 | 3.14 |
| BYU | 273 | 11 | 21 | 24.82 | 1.91 | 413 | 11 | 24 | 37.55 | 2.18 | -12.73 | -0.27 |
| AF | 452 | 9 | 27 | 50.22 | 3.00 | 487 | 9 | 25 | 54.11 | 2.78 | -3.89 | 0.22 |
| NM | 384 | 13 | 30 | 29.54 | 2.31 | 302 | 13 | 20 | 23.23 | 1.54 | 6.31 | 0.77 |
| Wyom | 441 | 13 | 48 | 33.92 | 3.69 | 338 | 14 | 38 | 24.14 | 2.71 | 9.78 | 0.98 |
| CSU | 319 | 12 | 24 | 26.58 | 2.00 | 293 | 11 | 19 | 26.64 | 1.73 | -0.05 | 0.27 |
| TCU | 300 | 15 | 35 | 20.00 | 2.33 | 466 | 17 | 40 | 27.41 | 2.35 | -7.41 | -0.02 |
| Utah | 587 | 12 | 34 | 48.92 | 2.83 | 500 | 10 | 38 | 50.00 | 3.80 | -1.08 | -0.97 |
| UNLV | 588 | 11 | 48 | 53.45 | 4.36 | 234 | 11 | 14 | 21.27 | 1.27 | 32.18 | 3.09 |
| Navy | 555 | 9 | 35 | 61.67 | 3.89 | 382 | 9 | 14 | 42.44 | 1.56 | 19.22 | 2.33 |
| averages | 456.69 | 12.31 | 35.00 | 39.05 | 2.97 | 353.77 | 12.08 | 22.08 | 30.42 | 1.83 | 8.63 | 1.14 |
Looking at YPD, SDSU averaged about the same as Michigan, at 39.0 vs. 41.9 YPD for the season. The anomaly in that data is the TCU game, where SDSU only managed 20 YPD. Michigan’s low for the season was Purdue, at 26.3 YPD. Once again, defense is much improved at 30.4 YPD versus Michigan’s 37.3 YPD. The defense was inconsistent, giving up 50+ YPD twice, a feat even Michigan’s Decimated Defense didn’t match. Overall, I think YPD is a useful indicator, but not as valuable as PPD.
On to PPD, we now get to see that SDSU was indeed more efficient in putting points on the board compared to Michigan, at 2.9 versus 2.7 PPD average for the season. Defense shows similar improvement, at 1.8 versus Michigan’s 2.9-PPD average. To put that in perspective, SDSU only had one game (Utah) where they allowed more PPD than Michigan’s average. So what the chart shows is that while SDSU’s offense doesn’t have the firepower of Michigan’s former spread (maxed out at around 5-6 PPD), the low isn’t so low, either. SDSU’s bottom end looks like it’s around 2 PPD, where Michigan could get down to 1.5 on occasion. Even in the win against ND, Michigan was below 2 PPD. On defense, we all know the story. Michigan flirted with 3 PPD for the season, giving up more than 3 PPD on 7 occasions. SDSU was more consistent, only giving up more than 3 PPD once.
On SOS
I think there are a number of valid comparisons between SDSU’s schedule and Michigan’s. Their #1 game was TCU, against the #4 offense and #1 defense. They had 300 yards of offense and scored 35 points, and gave up 466 yards and 40 points in a loss. That game is comparable to the OSU game, against the #11 offense and #5-ranked defense. Michigan had 351 yards and only 7 points, while giving up 478 yards and 37 points. Advantage: SDSU. Overall SOS differences are obvious (Michigan’s opponents scored 28.6 and gave up 23.4, compared to SDSU’s 24.5/28.4), but “score delta” should let us normalize those results.
Looking at the “score delta.” Michigan averaged 1.4 points above expectation for the season, compared to SDSU’s 4.5. It is totally open for debate as to whether this stat has any meaning, but I think that it does. If you’re supposed to blow out your cupcakes, and don’t, it will be reflected. Conversely, if you play well against a better opponent, like say ND or TCU, it is also reflected. I think it does a good job of showing overall team performance versus expectations.That big -32 by the Bowl Game shows that We Got Blown Out.
Michigan
| opponent | o-ppg | d-ppg | opp o-ppg | opp d-ppg | o vs d | d vs o | actual score delta | projected score | score delta |
|---|---|---|---|---|---|---|---|---|---|
| UConn | 32.77 | 35.22 | 26.38 | 22.0 | 27.385 | 30.80 | 20 | -3.42 | 23.42 |
| ND | 32.77 | 35.22 | 26.31 | 20.2 | 26.5 | 30.77 | 4 | -4.27 | 8.27 |
| Umass | 32.77 | 35.22 | 26.55 | 25.2 | 28.975 | 30.89 | 5 | -1.91 | 6.91 |
| BG | 32.77 | 35.22 | 21.25 | 33.6 | 33.175 | 28.24 | 44 | 4.94 | 39.06 |
| IU | 32.77 | 35.22 | 27.17 | 34.0 | 33.385 | 31.20 | 7 | 2.19 | 4.81 |
| MSU | 32.77 | 35.22 | 29.46 | 22.3 | 27.54 | 32.34 | -17 | -4.80 | -12.20 |
| Iowa | 32.77 | 35.22 | 28.92 | 17.0 | 24.885 | 32.07 | -10 | -7.19 | -2.82 |
| PSU | 32.77 | 35.22 | 24.54 | 23.7 | 28.23 | 29.88 | -10 | -1.65 | -8.35 |
| Illinois*** | 32.77 | 35.22 | 32.54 | 23.5 | 28.115 | 33.88 | 2 | -5.77 | 7.77 |
| Purdue | 32.77 | 35.22 | 19.67 | 28.8 | 30.76 | 27.45 | 11 | 3.32 | 7.69 |
| Wisc | 32.77 | 35.22 | 41.46 | 20.5 | 26.655 | 38.34 | -20 | -11.69 | -8.32 |
| OSU | 32.77 | 35.22 | 38.77 | 14.3 | 23.54 | 37.00 | -30 | -13.46 | -16.55 |
| MSU(SEC) | 32.77 | 35.22 | 29.00 | 19.9 | 26.31 | 32.11 | -38 | -5.80 | -32.20 |
| averages | 28.62 | 23.45 | -2.46 | 1.35 |
SD State
| opponent | o-rank | d-rank | opp o-rank | opp d-rank | o vs d | d vs o | actual score delta | projected score | score delta |
|---|---|---|---|---|---|---|---|---|---|
| NSU | 35 | 22.08 | 21.00 | 33.6 | 34.275 | 21.54 | 47 | 12.74 | 34.27 |
| NMSt | 35 | 22.08 | 15.67 | 39.5 | 37.25 | 18.88 | 20 | 18.38 | 1.63 |
| Mizzou | 35 | 22.08 | 29.85 | 16.1 | 25.54 | 25.97 | -3 | -0.43 | -2.58 |
| UtahSt | 35 | 22.08 | 22.00 | 33.8 | 34.375 | 22.04 | 34 | 12.34 | 21.67 |
| BYU | 35 | 22.08 | 26.15 | 21.6 | 28.31 | 24.12 | -3 | 4.20 | -7.20 |
| AF | 35 | 22.08 | 30.85 | 21.1 | 28.04 | 26.47 | 2 | 1.58 | 0.43 |
| NM | 35 | 22.08 | 15.83 | 44.3 | 39.665 | 18.96 | 10 | 20.71 | -10.71 |
| Wyom | 35 | 22.08 | 19.17 | 30.3 | 32.665 | 20.63 | 10 | 12.04 | -2.04 |
| CSU | 35 | 22.08 | 16.50 | 34.7 | 34.835 | 19.29 | 5 | 15.55 | -10.55 |
| TCU | 35 | 22.08 | 41.62 | 12.0 | 23.5 | 31.85 | -5 | -8.35 | 3.35 |
| Utah | 35 | 22.08 | 33.08 | 20.3 | 27.655 | 27.58 | -4 | 0.08 | -4.08 |
| UNLV | 35 | 22.08 | 18.38 | 39.7 | 37.345 | 20.23 | 34 | 17.12 | 16.89 |
| Navy | 35 | 22.08 | 29.69 | 23.3 | 29.155 | 25.89 | 21 | 3.27 | 17.73 |
| averages | 24.60 | 28.5 | 12.92 | 4.52 |
So What?
Well basically I think that the defense would have improved regardless of what happened, but I feel that the improvement with the new staff will be greater than the improvement with the old staff. I’m also hoping the MANBALL worries will be unfounded. After all, SDSU scored more points than Michigan, had more drives than Michigan, and darn near had as many yards as Michigan. I’ll trade that for a defense that gives up 100 fewer yards and almost 2 fewer touchdowns per game. I realize that most of the defensive improvement is speculation, since Mattison wasn’t Hoke’s DC at SDSU, but here’s hoping for Defensive Mediocrity in 2011, and a return to Kicking Competency!
"I realize that most of the defensive improvement is speculation, since Borges wasn’t Hoke’s DC at SDSU, but here’s hoping for Defensive Mediocrity in 2011, and a return to Kicking Competency!"
You mean Mattison.
Don't know what I was thinking with that one.
You post the same SDSU ypd chart twice, and then don't post Michigan's ypd and ppd charts at all.
Can you go back and fix this? It's an interesting way of looking at things, but I can't really use it the way I want to in the present form.
Removed the duplicate chart, the Michigan data is in Part I. This would be War And Peace if I put them both together.
April 26th, 2011 at 12:47 AM ^
You're missing a big opportunity by not putting the UM and SDSU ppd charts next to each other.
April 25th, 2011 at 10:07 PM ^
Seems like a much easier way of doing this would be to simply look at FEI and S&P rankings. The grunt work here has already been done for you.
SDSU FEI Offensive Ranking is 12th, Defense 45th. Those are SoS adjusted.
For comparison, Mich was 2 and 108.
Jeez those defensive numbers are always tough to take.
It relies on too many subjective decisions of what's considered "garbage time" or "clock killing" drives. So teams like Oregon, which blew out virtually every opponent they faced, get penalized for the simple reason that they were so dominant. It also overrated us by a long-shot. Would anyone, other than perhaps UCONN, be seriously more afraid of our 2010 offense than Oregon's or Stanford's? How about Wisconsin's?
Points per drive isn't quite a "pure" offensive metric, as it doesn't take into account starting field position, but it directly indicates something we were demonstrably poor at: points scored per offensive possession.
If we look at how we stacked up against Wisconsin, who played a lot of the same teams we did, it's downright depressing: 2.77 ppd (UM) vs. 5.88 ppd (UW).
Think that says more about our offensive output over the course of 2010 than our #2 FEI ranking...
FEI may have some place, but UM scored tons of points against either 1) bad teams, or 2) when trailing hugely and RR's run slanted spread was thrown out the door. I never felt FEI was doing anything other than hiding the problems the team was having. (See getting shut down during normal game time against any decent competetion)
Michigan had almost no ability to score from the red zone. It seemed most scores were from long plays from outside the redzone. While that is exciting (all those long TD plays) it blows to march down the field and turn the ball over on downs at the 12 because you can't get 4 yards a play on a short field and can't kick anything.
those weren't inherent difficulties with the SYSTEM, but with a stage of its development. Have a less than complete crap D and you have 2-3 more wins with that sometimes stuttering O, no one is trying to figure out how Hoke's stacks against it.
FEI accounts for your criticisms... you clearly have no idea how it works.
Some attempt at normalization is better than nothing, and FEI is orders of magnitude better than this attempt.
The concept of normalizing for "garbage time" and "clock killing drives" is really problematic.
Why should "garbage time" be eliminated in a league where teams purposefully go full-throttle, even when wins are assured, in order to do well in the polls? This is basically saying that every point in our 67-64 win over Illinois should count, but the last, what, 17 pts Oregon scored against Stanford don't? That makes no sense, given that Stanford was better than Illinois and Oregon won more convincingly than we did.
...and how about "clock-killing drives?" Why should those be eliminated from contention if a defense can't stop them?
This is over-normalization, in a way that renders the metric problematic.
Like I said, would any team, aside from UCONN and Illinois, find our offense more frightening than Oregon's, Stanford's or Wisconsin's? That 2.77 ppd looks like a better gauge of our season than the #2 FEI from where I'm sitting...
Any statistical ranking that had Michigan's offense ranked 13 places higher than Oregon (among others) is very flawed. You don't have to understand FEI to realize that; just understand football. You're comparing offenses that did the same thing, and didn't just have better results due to other factors, but actually did those things better (due to experience, talent, whatever). To think otherwise is just believing in a measure because you like what the outcome says; and that's bad statistics.
April 28th, 2011 at 10:59 PM ^
FEI is attractive, in a moneyball way, because of a sense that traditional metrics don't really capture offensive production well in an era of spread offenses. But it's trying to do way, way too much, and commits one of the worst errors possible in statistics--skewing the distribution.
Oregon's likely ranked #15 because so many of its points fall under "garbage time." But why should this be eliminated from consideration? Oregon played the same teams Stanford did, for the most part; they just outscored them by more. Yet Stanford ranked ahead of Oregon by 9 places.
Points per drive is a limited metric that doesn't tell you everything you need to know about an offense, and doesn't try. It does tell you one useful thing, though: given possession of the ball, what is the predicted # of points scored (p-hat)?
Ours, 2.77, was not very good. Wisconsin's, by contrast, was very good. The great thing about this metric is that it works for both slow-plodding pro-style offenses and quick-flashy spread ones. If you score 5 ppd and possess the ball 8 times, then the other team isn't likely possessing the ball much either, due to long drives. If you score 5 ppd and possess the ball 20 times, the other team likely gets a lot of possessions too. Ergo, if our ppd increases, it likely means our ratio of points scored to points allowed will too.
played 6 of the 20 worst Ds in D1 football. In other words, six of their games were against Ds which were worst than ours last year. Add in a seventh game against a 4-7 FCS school and their offensive numbers don't mean squat. We faced 7 of the top 30 Ds in the country and would have averaged IMHO 60 points per game with SDSU's schedule.
You talk about the TCU game, but SDSU put up 199 of their 300 yards on the last drive of the 3rd and the fourth quarter after they were down by 24 points. They had 101 yards of O for almost three quarters. They had six straight three and outs after being up 14-0 on a flea flicker and defensive fumble recovery in the end zone. During that time TCU scored on 5 of 6 drives with only a missed FG keeping them from being 6 of 6. Halftime 35-14 TCU.
had just 100 yards of offense thru almost three periods. The games you mention we scored quickly starting the second half our D just couldn't stop the other team so they are not similar.
Did I ever mention RR? No, just making points about the comparison which is like comparing apples to oranges. Hoke is our coach now and I will cheer for the team the same way I always have. What is on the helmets is all that matters.
April 26th, 2011 at 11:50 AM ^
SDSU had the ball in the fourth quarter against TCU with the chance to take the lead.
Michigan did not against Michigan State, Iowa, Penn State, Wisconsin, Ohio State or Mississippi State.
So yeah, they are not similar.
SDSU scored with 4:58 left in the game to get within 5 and TCU took the ball and marched it down to the SDSU 19 as time expired.
Scoring Offense is not adjusted for SOS, which is why I tried to make up the "score delta" value, to show when you properly blew out someone you were supposed to.
I certainly don't harbor any ideas that the offense will be better this year than last year, in term of points or yards. And I didn't mean to imply that the TCU game was a good game, by any means. I meant it compared to our OSU game, a loss against a better opponent. In that case, the numbers were similar. One the one hand they fell behind, and on the other hand they kept playing and made a game out of it.
April 28th, 2011 at 12:29 PM ^
Given that we return most of our offensive players - and that our defense will hopefully force more turnovers/punts - I think it's entirely possible that we can score more points that last year. Our PPG was not that remarkable (32.8), especially when you consider that we had two giant outlier performances (67 points against Illinois, 65 against BGSU) that served to inflate it.
April 26th, 2011 at 11:28 AM ^
The stats exist and all but I don't honestly think there were 6 defenses worse than Michigans in 2010.
I like the comparisons. I am not to worried about our offense. I am ready for somewhat of a drop off, due to Denard fitting RR's system, and transitioning, but I think Borges will play a lot of spread.
Our defense will definately improve( better coaching, scheme, and ave age of players), but we'll still need to out score some opponents. Should be an exciting season.
Overall, I think YPD is a useful indicator, but not as valuable as PPD.
I think that is a restatement of the "bend, don't break" philosophy. Plus, in the end it isn't who gains the most, but who scores the most...
April 26th, 2011 at 10:00 AM ^
is it really worth saying "better"/"worse" as you comb through the comparisons? A few yards more here or there, under your adjusted averages, just don't mean much. to my mind.
The takeaway is encouraging, though. Certainly looks as though Hoke's SD O was pretty (comparatively) productive.
The other big problem, of course, is that SD's O won't take the field for us this September, but our kids in the first year of a new system; and that system will be a bit of a Frankenoffense.
April 26th, 2011 at 10:52 AM ^
Turnovers aren't factored in and we were terrible again in 2010. Three years in a row. Holding onto the ball better (ex. the MSU int in the red zone in the first quarter) and turning it into points now and again will make a huge difference.
The gain is in raw points but also the emotional/confidence infusion to your team.
April 26th, 2011 at 12:07 PM ^
Nate you beat me to it but I will post just too emphasize the point. Also a number of Michigan' drives went extra downs=extra yards having to go for it instead of try for a FG. Scheme to scheme I have no concern it's personell for said scheme (can DR make progressions-can this O-Line get push against ND/B1G linemen?) that concerns me.
Also Nate first time I have noticed your signature-AWESOME!
"Any street urchin can shout applause in victory, but it takes character to stand fast in defeat. One is noise --- the other, loyalty." Fielding Yost
April 26th, 2011 at 11:07 PM ^
Yeah, you can't really compare the two perfectly, but it's not simply that "Michigan had it worse." SDSU also has generally poorer talent (even considering our subpar recruiting standards over the past few years). The fact is, we can't predict our success based on SDSU's record in 2010. If that was the case, we would have gone 11-1 in 2008. We could be significantly better, or we could be worse.
That said, I think what Hoke and Borges did with SDSU is promising--not the record, but the fact that they were game against the 3 superior teams they played, which included one that was arguably as good as the two that played in the NC. By contrast, we were wretched against all the teams more talented than us, finishing the season with embarrassing losses against Wisconsin, OSU and Miss St. That stretch were Wisconsin just ran right at us like 25 straight times was humiliating--it was something you do to snot-nosed cousin in Madden or NCAA, just to piss him off.
I HOPE the comparison of offensive efficiency (in terms of PPD), and net efficiency (in terms of net PPD) is some indication of a positive change for Michigan football, but rationally, I know it guarantees nothing. It may, excruciatingly, even take more than a year to find out one way or another.
April 28th, 2011 at 11:40 PM ^
I posted a response to/adaptation of this mini-series over at my blog that takes a look at San Diego State's improvements in year one under Hoke. Check it out if you're interested, I guess.
Comments