Michigan Football, Team 138, 2017 Season.

[iam416]

Fancystats have their value too, but they have to be looked at with some scrutiny, and the stats that have little to no predictive value have to be thrown out. Too often people just accept them because smart people come up with them and, therefore, they must be right.

Correct. As mentioned on numerous occasions, fancystats lends itself to high sample size sports. That's not football. Fancystats lends itself to high sample size; fairly competitive sports. That's not CFB. Fancystats also lends itself to sports with very discrete matchup and players of interchangeable importance. Again, not CFB. They're perfect for baseball, though!

Efficiency numbers and all that crap that are, in part, built on performances against Air Force or Rutgers or Nebraska aren't all that useful. To use Ohio State -- I'm still not sure about OSU's offense. Fancystat-wise, the OSU offense is probably approaching juggernaut status after bludgeoning shit teams. But those games don't matter. What matters is whether the offense can move the ball against PSU, Sparty and M and even at Iowa at night. My starting point for each of those assessments isn't fucking efficiency numbers. It's personnel matchups. Can OSU's WRs get separation against M's DBs? Can OSU handle Hurst w/o an f'n triple team? Can they put Dobbins in space against them? Those aren't questions answered by stats.

The fancystats analysis for the M-PSU game say it's a blowout. But the matchups suggest that PSU is going to have difficulty moving the ball and that M's defense ought to do enough to give them a chance to win if they get some favorable breaks (ST or TO).

[Jeff Buchanan]

Correct. As mentioned on numerous occasions, fancystats lends itself to high sample size sports. That's not football. .......

I'm pretty sure you know this but "FancyStats" actually has a purveyor and that is the WaPo. Fantasy Sports players use this source a lot. Fancy stats, OTH, is kind of light hearted term used to describe the various statistics based analytic tools that are out there purportedly to predict outcomes and measure performance.

I'm not sure I agree that football is not amendable to statistical analysis if that is what you are saying. I'm also not sure that analytics do not have some measure of predictive power > than a coin flip.

The data bases for metrics like Connelly's Advanced Stats (S&P+ and FEI) collate and can compare millions of plays across 130 FBS teams, thousands of plays across conferences and hundreds of plays by teams. These aren't small sample sizes. I'm going to have to study this a bit more as I'm not familiar with the benefit that accrues to other sports (baseball, e.g.) that does not accrue to CFB. There are others statistical measures and analytic tools. This web site explains most of them and goes into more detail about why analytics has viable predictive and comparative (ranking) power:

https://thepowerrank.com/guide-cfb-rankings/

Efficiency numbers and all that crap that are, in part, built on performances against Air Force or Rutgers or Nebraska aren't all that useful.

If you'rer speaking of FEI (Fremeau Efficiency Index) it does account for SOC. I'm not a math or stats expert I can't tell you how or to what impact but it does. S&P+ does too as well as multiple other rating systems.

........My starting point for each of those assessments isn't fucking efficiency numbers. It's personnel matchups. Can OSU's WRs get separation against M's DBs? Can OSU handle Hurst w/o an f'n triple team? Can they put Dobbins in space against them? Those aren't questions answered by stats.

You know this. There are ways to use both observational data and analytics to good effect. Saying predicting match-ups "isn't (about using) efficiency numbers" appears to me to be dismissive of of any analytics and seems to me to show you really don't know about them. There is not one "efficiency number." There is the FEI, among about 5 other analytic measures in CFB, and it is, IMO, fairly robust as a predictive tool.

http://www.footballoutsiders.com/stats/fei

The fancystats analysis for the M-PSU game say it's a blowout. But the matchups suggest that PSU is going to have difficulty moving the ball and that M's defense ought to do enough to give them a chance to win if they get some favorable breaks (ST or TO).

The one I chose to use (ESPN Power Ranking) did show a 21 point difference between the power rankings of PSU and M. But the rest of them show a spread much closer to the 10.5 points that is (one of) the opening line for the game ...... i.e., most do not show a blow out unless you consider an 11 point win to be a blow out.

The first time I really dug into the Advanced Stats Match-ups posts at mgoblog was the M/IU game...... and this is for you as well, Hack ..... The thing that stood out in this analysis and resonated with me was the benefit to M of passing on standard downs. I made the comment that during the first half JOK did throw the ball on standard downs and, IMO, it was this approach that made the offence click, i.e., improved the offense in both the run and pass game. Unfortunately, deserved penalties and bad refereeing got M off schedule during more than one series.

Admittedly, I'm making that an eyeball observations. I've not gone back and done a play by play b/c Brian will do that. If I'm wrong about this, I'll acknowledge that here. I do know that there were 20 passing attempts and I think most if not all of them were in the first half.

Hack, it was my perception that both in your post itself where I mentioned this and the tone of it, you dismissed the analytics that pointed to the benefit of passing on standard downs.

Let's be clear. I've never stated or implied that the best way to make pre or post game comments is to base them only on objective (analytic) data. While that is my choice in making predictions or post game comments, and I've voiced that, I've also been clear that this forum is here for posters to post anything they want about a game; stand alone observations and observations based clearly on the poster's emotions are fine (Hanni's post applies).

[hack]

I am without question dismissing anything that suggests JOK should pass. For obvious reasons. Talent brought up some good points. Small sample size in football, smaller still because of varied quality of opponents. Sure you can correct for things, but the more corrections you have, the more room for over/undercorrecting. Suggesting that it's a good think for JOK to pass because numbers say it is, to me, a wonderful example of how an overreliance on stats leads to narrow, tunnelized understandings of something. It's fucking nuts to suggest that JOK passing is a good option in most scenarios. We had a guy run for 200 on 25 carries. How the heck do you conclude that the offense was improved by a passing game in which we averages 2.9 ypa on a low completion rate? You can't even make the case that it was a useful changeup -- everyone is overplaying the run and they aren't capitalizing on that.

[froot loops]

It is a small sample size compared to the other sports.

[iam416]

Yeah. IMO, they lack sufficient predictive value. That's not to say they don't have some and should be dismissed. But they aren't the same for football, and especially CFB, as it is for MLB.

Re weaker teams, I think it's difficult to take into account blowout games because coaches start to do so many different things. Some coaches will keep the foot on the gas; others will change play calls based on the margin.

As I've said before, I enjoy advanced stats for baseball and basketball. I don't think they're all that useful for football.

[whodean]

It would be foolish to dismiss all gameday observations as some completely irrational emotional thing.

[Jeff Buchanan]

I am without question dismissing anything that suggests JOK should pass. ....... How the heck do you conclude that the offense was improved by a passing game in which we averages 2.9 ypa on a low completion rate? You can't even make the case that it was a useful changeup -- everyone is overplaying the run and they aren't capitalizing on that.

I think I know where our differences on this are. Did you look at the piece Advance Stats match-up that was posted on Friday at mgoblog? I've posted the link below.

Generally, the kind of stats that go into Pting's pieces are player independent. The data points are all plays. Matchup analysis draws upon the Advanced Stats Profiles published weekly by Bill Connelly on Football Study Hall. The profiles feature Connelly?s well-known Five Factors, and also include the more detailed groups of S&P+ metrics that break down elements of the game such as Rushing and Passing, as well as the down-and-distance scenarios known as Standard Downs and Passing Downs.

If you study these posts, you will see which match-ups, if any, favor one of the opponents. When there is an advantage for one team, in the case of the match-up we are discussing here, it is player independent. You are discussing the match-up as if it was player dependent. It's not. More on that later.

What the analytics do in this particular case, favored to pass on standard downs, is to inform the coaching staff. If the plays are executed, you will have an advantage if you pass on 1st and 2nd down. The analytics fail if the play is not successful.

Clearly, and this is your take, O'Korn can't execute the first down pass so, I can see where you are coming from ..... why in hell with O'Korn at QB would you call a pass on first down. With O'Korn a dependent variable the answer to that question is, don't call a pass.

I can tell you from reading around the web the last two days, this is a hotly debated subject ..... pass on non-standard downs or not. Advance Stats says yes on a player independent basis. So, a poster at mgoblog did an analysis of this comparing Speight and O'Korn (link below):

Speight: 65%/35% run/pass on first down. Pass plays had a 70% Success Rate (SR). Run plays 62% SR.

O'Korn: 61%/39% run/pass on first down. Pass 62% SR, Run 68% SR.

It was obvious to me that when O'Korn went in for Speight in the Purdue game and in all subsequent starts (MSU, IU), there was a lot more passing on first down. He was mostly successful at Purdue, horrible at MSU and in the first half at IU plainly not good. By late second quarter at IU and for the entire second half, the staff adjusted to his low success rate and shifted to the run.

So, yes, when you consider a QB's performance as a dependent variable to play selection if the QB is not making the plays, don't call them. This obviously highlights the role of observational data, in this case the QB can't complete passes on first down, in judging the value of the analytics.

This does not invalidate the analytics going into the game. i.e., pass on non standard downs when playing IU.

http://mgoblog.com/diaries/advanced-...chigan-indiana.

Here's the Cliff Notes of the post at the link belwo:

- With Speight, Michigan threw the ball 35% of the time on first down

- With O'Korn, Michigan has thrown the ball 39% of the time, and the numbers were even higher before the coaching staff adjusted after it was clear O'Korn could not complete the passes that were being called in (44%!!! against MSU--not including the last desparation series)

- With Speight, the off-schedule first down throws were relatively successful and led to several explosive plays, and seemed like a major component of the overall offensive philosophy

- With O'Korn, Michigan is missing both those effective and explosive first down passing plays

- With both Speight AND O'Korn, down series that ended with FG attempts vs. TDs were more likely to start with a pass on first down..

http://mgoblog.com/mgoboard/analysis-first-down-plays-speight-vs-okorn

[hack]

when you consider a QB's performance as a dependent variable to play selection if the QB is not making the plays, don't call them.

When shouldn't one consider that?

[Detroit Dan]

I'm with Jeff on this one. Clear, cogent analysis.

Any discussion of O'Korn's performance to date should throw out the 2nd half of the MSU game (weather) as well as the 2nd half of the Indiana game (too few pass atempts). The offensive line was horrible in the 1st half of the MSU game (Ulizio benched as a result), and he was under big time pressure.

O'Korn has been bad, and I'd like to see Peters or McCafferty. But I think Jeff is right that overly conservative play calling in the 2nd half against Indiana was counterproductive. Michigan had 4 drives in the first half for 13 points (not including the last drive where Michigan let the clock run out after 2 plays). In the 2nd half, Michigan had 9 drives for 7 points. Half the points in twice the number of drives = success??

[whodean]

When you are trying to nurse a lead with a terrible QB without turning the ball over, yes, that is a success.

[hack]

Also throw out last year's Indiana game? So that leaves you with, basically, seven quarters to compare to eight for which there's an asterisk? Or 11 to compare to 4? Either way, looks pretty damn bad.

[Jeff Buchanan]

I'm with Jeff on this one. Clear, cogent analysis.

Any discussion of O'Korn's performance to date should throw out the 2nd half of the MSU game (weather) as well as the 2nd half of the Indiana game (too few pass atempts). ...... But I think Jeff is right that overly conservative play calling in the 2nd half against Indiana was counterproductive. Michigan had 4 drives in the first half for 13 points (not including the last drive where Michigan let the clock run out after 2 plays). In the 2nd half, Michigan had 9 drives for 7 points. Half the points in twice the number of drives = success??

The analytics referenced in my posts and applied to the specific question of should Michigan pass or run on standard downs against Indiana would tell the coaching staff to pass on standard downs. It does not tell them anything about O'Korn's ability or inability to do that.

So, I'd offer these conclusions: It appears that the coaches have been cognizant of the value of passing on standard downs independent of who was under center. The numbers of first down passes thrown in M's first 21 quarters (through the first half of the IU game) supports that.

It is also clear that when it became obvious that both QBs, to a greater degree O'Korn, could not enhance the success rate of the down or a drive's success rate, the number of these calls plummeted (see Analysis of first down plays with Speight vs. O'Korn in the link above). It was dramatically obvious v. IU in O'Korn's case.

There's been a lot key strokes made to attempt to conclude that analytics are useful or they are not. Similarly, there's a fair amount of debate about the value or lack thereof of hot takes/emotional reactions to a game outcome.

I've never claimed analytics should be the primary driving force in pre-game game planning or in game play calling. What I have said is that they have the potential to inform the coaching staff in both these areas with real time, direct observation of the result of acting on the analytics equally as important. Same thinking applies to post game conclusions the coaches might make.

Post game hot takes, are IMO, not that useful however, I'll defer to Hannible's excellent post on this ......."Emotions have value. Anger can blind you to facts in the heat of the moment but it can also provide clarity in coming to a conclusion once you know those facts."

You'll never see Harbaugh give insightful hot takes or apply feelings ball post game during an interview. He always says, I'll have to look at the tape when he gets questions on the performance of a specific player. Smart.

I'd like to think that post game comments that I make here provide some objective data to back them up. That is not to say others who post emotional comments about the outcome without such data are FOS. To me, some, not all, of these types of hot takes/emotional comments have limited value, Hanni's comments on this not withstanding, and are frequently but not always found to be inaccurate on further review. That's just me.

[Detroit Dan]

Well, Jeff, after so many years of lengthy posting, you've won me over. In this case, I think your moderation and even-handedness serve you well. Also, I've got to hand it to Hannibal for some perspicacity (look it up).

Well done!

[Detroit Dan]

Harbaugh's a great recruiter, but what's with this:

In 2016, Michigan held late leads against Iowa, Ohio State and Florida State, but was unable to close out those games when they had opportunities to get first downs on offense or make big stops on defense. Against Florida earlier this season, the Wolverines settled for six field goal attempts and let the Gators hang around for virtually the entire game until (finally) a key defensive touchdown in the fourth quarter put the game out of reach.

On Saturday against IU, the Maize and Blue were up 10 when Lavert Hill picked off a pass with six minutes left in the game. The home crowd stood up and began filing out of the stands. It felt like the Wolverines had the game in hand and would run out the clock. But three plays and minus-seven yards later, Michigan punted to Indiana, who ran the ball back to the Michigan 20 yard-line, leading to a Hoosier touchdown.

After narrowly giving up an onside recovery to Indiana, the Wolverines got the ball back with 3:27 remaining. Alas, another opportunity to put the Hoosiers away. Nope. Again, three plays later, Michigan punted to IU, who used its final drive to add a field goal and send the game to overtime.

The Wolverines very much escaped the Hoosiers (again). They got lucky. This conservative play-calling in the fourth quarter, in future contests, will only serve to keep opponents in the game — as it’s done time and time again.

https://www.maizenbrew.com/football/2017/10/16/16480718/michigan-wolverines-indiana-hoosiers-lessons-learned-2017

I think this is the definition of chicken-shit.

Anyway, time to change tactics and win some Rock-Paper-Scissors when the game is on the line.

[iam416]

Dan really ups his game with perspicacious.