Unverified Voracity Invents Clocks Comment Count

Brian July 9th, 2018 at 4:40 PM

don't fade unless you have Jeremy Gallon [Eric Upchurch]

Sponsor note. If you've got a business that needs creating or safeguarding or shepherding, Richard Hoeg will make you contracts and get you registered and generally set you on the path to being the world's first atomic sled tycoon. What's an atomic sled? I don't know! It's your idea. Need to work on your elevator speech, because I got nothin'.


Hoeg Law: when you figure out what an atomic sled is. We are good at ads.

The goal line fade! Don't do it. Data from NFL two-point conversions:

Also intriguing that every color commentator's favorite option—the rollout pass—is the second-worst decision. Probably because everyone in the world thinks it's a good decision.

Slot corner. The Athletic engages Mike Renner of PFF to detail Michigan's 2019 NFL draft prospects. Many of the same stats you've seen on PFF's tweets—David Long's silly numbers, Chase Winovich's general relentlessness—feature but the most interesting new bit is a negative one on Tyree Kinnel. Not a surprising one, really:

Kinnel was a full-time safety for Michigan who also covered the slot at times. In the NFL and in college, he projects much better to the former. On 91 snaps covering the slot, he allowed 208 yards and a passer rating of 110.7. As a deep safety, he looked much more comfortable.

It is my contention that opposition WRs caught more than their fair share of heavily contested balls against Kinnel a year ago and that even if he's the same player that should be less of an issue this year. But if they've got three really good corners they should probably put one on the slot whenever the opposition has a passing down.

Some good news from the article is that PFF doesn't think Hudson has an NFL position right now and Lavert Hill isn't a slam dunk early entry guy, so Michigan could get them back next year.

[After the JUMP: lies, damned lies, and rosters.]

Mike McCray: not a DE. ESPN article on roster fictions features Mike McCray as a particularly egregious example:

Halfway through the Senior Bowl weigh-in, the same kind of pained look unfurls across the face of Michigan linebacker Mike McCray. Listed at 6-4 his entire college career, McCray does a double take when Boni announces his actual height as barely 6-1. It's the biggest fib by a linebacker anyone has heard since, well, since the year before, when Michigan linebacker Jabrill Peppers turned out to be 5-10, more than 2 inches shorter than the Wolverines claimed.

... he grew to enjoy being thought of as 6-4 and even convinced himself it was an acceptable embellishment because in his heart he knew he was taller than his dad, Mike McCray Sr., a linebacker and captain on the 1984 Ohio State team who was listed on the Buckeyes' roster at 6-3. The only problem? Now an assistant high school principal in Ohio, Mike Sr. admits that he's -- wait for it -- only 6-1 in real life. "When I have my cleats and helmet on and I'm feeling good, I do feel taller than I am," Mike Jr. says after his first Senior Bowl practice. "What else can I say? I honestly don't know what I'm going to say, but I'm sure teams will have questions about it."

Less than five weeks later, a duly chastened McCray found out exactly how much the NFL worries about this kind of blatant heightening when he reported to the NFL combine in Indianapolis. Terrified that he had messed up his draft status, the first thing McCray did after checking in was look up his new official draft bio.

After a few anxious clicks, there it was, in big digits right at the top of the page, his new NFL-approved authentic height: 6-4.

The NBA does a good job of measuring in shoes and out, and usually all but an inch of any discrepancies is explained by the "in shoes" number. I'm guessing McCray is 6'1" in bare feet and then Michigan rounded up. For a given definition of "rounded up" that includes adding one and then rounding up.

RIP NCAA: the game. This is the five year anniversary of the last NCAA Football, and a reminder that Denard was—for now—the last cover gent. People like the game so much that even five years later they're updating Iowa's offensive coordinator:

Seven gamers have formed an “editing team” that’s done the lion’s share of OS’ work to keep NCAA 14 as fresh as possible.

“But there are dozens more contributing to the roster thread with relevant info,” Chris Sanner, the forum’s executive editor, writes in an email.

It takes a few weeks each offseason to get the rosters done. A trip through the forums shows fans of different teams chiming in over several weeks to offer help and feedback.

“I would love to do Georgia Techs roster I could 100% have it done by Monday,” one fan wrote in February 2017. “I can contact Muck on twitter about it.”

A couple folks have taken the opportunity to remind us that the problem wasn't with EA, which would happily pay for image rights to the players if they could make a game, but the NCAA, which thinks that players getting money over the table is a grave affront:

EA knew that consumers wanted those improvements to the videogames, and increased sales would result. But the NCAA remained steadfast in its prohibition against sharing revenues with players and worried that any additional similarities between the athletes and the videogames would further expose the fiction that the videogame "avatars" did not represent real people.

Of all the bizarre consequences of amateurism, the NCAA refusing money because it would result in other people also getting money is way up there.

Veteran backup quarterbacks: nope. Paul Myerberg quantifies a trend you've probably noticed:

...transfers are up across the board, and not just at quarterback. There were 211 graduate transfers playing in the FBS in 2017, a drastic increase from the 117 such transfers in 2016 and the just 17 in 2011. But no other position embodies the transfer craze quite like quarterback, perhaps due to the overwhelming attention always paid to the position in the first place.

It’s led to the near extinction of a certain type of college player: the career backup. Once a roster staple, senior quarterbacks who begin their careers on scholarship and stay with the same program through four years of eligibility without ascending to the starting role are now the sport’s rarest breed. ...

There are only seven quarterbacks in the Football Bowl Subdivision who fit the criteria. Just two, California’s Chase Forrest and TCU’s Grayson Muehlstein, play in a Power Five conference.

Michigan benefited from this when Jake Rudock didn't take an offseason benching kindly; then they lost Wilton Speight this offseason. The last senior non-starting QB Michigan had was David Cone in 2009, who probably didn't have a shot to win a job at another D-1 school. The last guy who probably could have found playing time elsewhere was Scott Dreisbach all the way back in 1998. So... yep. Quarterback who doesn't win the job is going to transfer, especially in the grad-transfer era.

Haven't you invented clocks yet? 538 tackles one of my biggest pet peeves in sports: stoppage time. Namely, that it absurdly under-counts actual stoppage time:

Actual stoppage time is a wildly inaccurate measure of how long the game was actually stopped. The average added time flashed on the board for these 32 games was 6:59, which includes both halves. By our calculations — which adhered to FIFA’s rules on the matter — the time that should have been added to each game was 13:10. This means stoppage time was roughly half of what it should have been for most games.

This is probably generous since 538 didn't start counting balls out of play as stoppage time for at least 20 seconds (on a throw-in) and up to a minute for a free kick or penalty kick. "Arguing with the referee" even gets a 30 second allowance.

People know this, and thus you get soccer's epidemic of flopping about and substitutions with two minutes left and all manner of other junk designed to avoid playing soccer when someone's trying to hold on to a result. The game tells them to do it, so they do it.

And while I'm complaining about soccer rules, yellow cards need to mean something more than "don't do that again." Proposal: the free kick after a yellow card foul outside the box can be moved anywhere you want on the field except the penalty box. Let's turn tactical fouls at mid-field into dangerous set pieces instead of nothing.

On the other hand. 538 calculates that video reviews take about 30 seconds in the average game. 30 seconds! I spend five times that much watching ref butt in every basketball game so they can reset one shot clock. How did soccer get replay so right on the first try when every American sport is getting it as wrong as possible? Every efficient VAR usage in the World Cup sends me into helpless conniption fits as I remember the Charles Matthews Zapruder Call against Purdue, which took literally seven minutes.

Etc.: Jack Johnson signs with the Pens. Little bit of 2020 D commit Ethan Szmagaj scouting. My goal is to be able to spell "Szmagaj" without looking it up by the time he enrolls. Money: there is a lot of it. Bush, Gary, Hudson on the Sporting News's preseason AA teams. Sounding slightly grim for Quinn Hughes's return. Heisman snub lists that don't include Ron Dayne over Mike Vick can get bent! That didn't take long: AFC's Stanley Okumu is trialing with a USL team.



July 9th, 2018 at 4:57 PM ^

I've never understood why people think goalline fades are ever a good idea. You literally need a perfect pass into a tight space and tight coverage to make them work


July 9th, 2018 at 5:49 PM ^

3rd and improbable when you're fine settling for a FG makes the goal line fade a low-risk call. If you have a large-bodied high point specialist at WR it makes the call much more enticing.


I think much of our fanbase became frustrated due to perceived fades to tiny McDoom-shaped objects. In reality, I think these were coverage reads that were designed as back-shoulder passes which were not properly executed due to poor QB ability.


July 9th, 2018 at 6:41 PM ^

Human bias towards control.  If you do hit the perfect fade there's nothing the defense can do to stop it.  With a draw or inside run, even if you execute perfectly, a well-timed run blitz will stop you.  Humans are biased towards being in control, even if the odds are objectively worse.

Unfiltered Manball

July 9th, 2018 at 5:08 PM ^

Purely for selfish 2019 reasons only - here's to hoping PFF stand by their preseason assessments of Hudson and Hill and do not watch a single snap for either this entire year.


Also, I've heard California's Chase Forrest is breathtaking in the springtime...

Mr Miggle

July 9th, 2018 at 7:33 PM ^

It's true that Cone didn't return for his 5th season. Hard to imagine RR offering him that choice. I don't believe he ended up at Iowa. Big Ten rules at the time would have meant no scholarship for him and I doubt Iowa level programs would have offered one. You might be thinking of Carr's QB recruit in the 2008 class who flipped to Iowa, John Wienke.


July 9th, 2018 at 6:15 PM ^

I'm not a fan of studies like the two point conversion study referenced here. Let's leave aside that (believe it or not) it's not really a large enough sample once you break it down into micro-categories like the author does. The more salient point is this: in general, trying to draw broad conclusions like this from non-randomized data without a control group is a fool's errand.

For example, are we really to believe that coaches would call for nearly three times the number of passing plays as running plays if that running plays were truly nearly 50% more successful than passing plays? It's possible NFL coaches are this stupid, but most likely they aren't. Most likely the entire reason that running plays are so successful is precisely because they are called so much less frequently. You can't suddenly start converting 60+% of your two point conversions just by running the ball. Opposing teams would adjust. The most likely explanation from the data given is that defenses play the pass, since that's what offenses want to do, and so runs are more successful, but less common.

And that's just the start of the problems with this kind of analysis. Teams' play selection is highly non-random. Teams are going to play to their strengths. The opportunity for selection bias is everywhere. For example, it's entirely possible that bad offensive teams throw more fades because they know that they are relatively safe and their other options are terrible, where as good offensive teams with more balance diversify their play-calling. In that scenario, fades looks bad not because they are inherently bad but because bad teams *correctly* run them more; the causality could run entirely (or just partially) the other way. And again, perhaps other types of passes or runs are effective precisely because other teams have to guard against the fade. Teams could easily choose to sacrifice some efficiency on fades to set up greater efficiency on other plays. It's very, very possible that by drastically cutting the number of fades you throw, you greatly reduce the efficiency of the other plays you run instead.

Imagine an analyst looking at a the performance of a baseball pitcher's repertoire, noticing that he performs a lot better on his slider than on his fastball, and so recommending that the pitcher stop throwing his fastball entirely, and start throwing his slider almost all the time. I think we would correctly recognize why that is nonsense: pitches aren't throw in isolation; it's a dynamic, evolving equilibrium with a lot of feedback and response from the actors involved.

It's so incredibly hard to draw valid inferences from data like this. I'm not saying that NFL play callers necessarily have the right mix, but it's really, really naive to treat numbers like this as any sort of conclusive, or even highly suggestive, evidence. Selection bias is a very, very powerful force.


July 9th, 2018 at 8:36 PM ^

I agree with everything you said in the second paragraph, it is spot on.  Failing to control for initial conditions like formations and coaches known tendencies is problematic. You are also correct that this dataset is not large enough to integrate those things into your model even if you wanted to, and one should be skeptical of this data. It is very likely that running breaks tendency and as soon as coaches start "running enough" the success rates of running will drop.

That does not make the data useless, it just requires responsible consumption of the data. I do not think it would be a mistake to start running more on 2 point conversions until the defenses do more to stop it.  

Finally, I do disagree with your issue on the lack of controlled randomized experiments. Sure, it would be nice to have that type of data but that is not realistic.  Datasets collected from real world outcomes are employed frequently and can have great outcomes.  It just requires more careful thought put into the model and conclusions from the model.  




July 9th, 2018 at 9:33 PM ^

Thank you for your comments. Like I said, I'm not presuming that NFL coaches do have the right mix. I just think that the data as presented here is so inadequate as to make it near worthless for decision making.

With respect to the availability of data from controlled, randomized experiments, I agree: this data is not common. However, the lack of availability of good data doesn't make the data that remains useful. To engage is some wild analogy, if I require a life raft to escape a sinking ship, but all I have on hand is a blender, it won't do to say "Yes, it would be nice to have a life raft, but that is not realistic, so this blender will be fine." I might just be screwed. Similarly, in real life we aren't guaranteed to ever have useful real world data, even though we might like to. The data that we have are either sufficient or they aren't, on their own merits, just as a blender is either a good flotation device or it isn't. That it is all we have on hand doesn't mean anything.

My issue with the way people use real world data is two-fold:

  1. It usually requires the construction of models that can bake in the author's preconceptions about the problem in subtle ways.
  2. I think people are generally unaware of exactly how subtle problems of selection bias (among other biases) can be.

I will admit: I am more than a bit of a skeptic. I have read good arguments made from real world data, but they are all much more sophisticated than this, of necessity.

So in broad strokes, I agree with you: real world data can be made to be useful (though this is not a guarantee), but it requires a tremendous amount of care to do so, at a level which is rarely present. I think we can both agree that the requisite level of care is not present in this instance. In the end, the proof is in the pudding: if you construct a model from real data that yields actionable information that consistently produces good outcomes, then you have found something useful. That's just a really hard process, and not one conducive to tweets at NFL coaches.


July 10th, 2018 at 8:27 AM ^

He would make a battery out of chewed bubble gum and a hair follicle, and then use the blender as both a flotation device and a boat motor.  Chuck Norris would come along and eat the gum hair battery and use the boat blender as dental floss.

Chuck will always win against MaGruber...


July 10th, 2018 at 12:23 AM ^

Your first complaint of the modeling of real world data is the typical frequentist complaint of bayesian analysis, and if employing bayes theorem it only really carries water when a strong prior is utilized. 

The problem of using this data to model 2-point conversion success rates is that there are a ton of parameters to explain the model (22 players, weather, etc). Therefore all someone can do is compute the fraction of successes over various slices of the data. The author of that tweet thinks he's done an analysis, which is funny, but what you've pointed out is correct. It is a problem that even educated writers like Brian link it and think it means much of anything. Perhaps fades are bad choices, but these data don't really tell us that on their own. 


July 10th, 2018 at 8:31 AM ^

A statistical debate over simple division from a second rate, half-wit reporter...

You will not see this argument at RCMB or whatever OSU shack is operating these days.  And it is why I love MGoBlog - come here for a little football talk, learn about statistics.  Time well spent!


July 10th, 2018 at 9:53 AM ^

Interesting. I admit to not being super familiar with the frequentist interpretation of probability. I certainly would have described my viewpoint as generally Bayesian though; I have no beef with priors (everyone has them), I just prefer that they be called out explicitly.

As far as my complaint about preconceptions being baked into models based on real world data, I agree that in principle one should be able to account for these biases or assumptions. However, in practice, I think it's more common that bias is introduced because the author of the model is very often blind to the bias in the first place; it may be a factor the author hasn't even considered or is unaware of or simply sees as common sense. It often takes someone with a completely different point of view to notice these things. Ideally, that's what peer review is for; alas, in the real world, group-think is also a thing.

tl;dr: Bayes: good; cognitive bias: bad and stubbornly pernicious.


July 10th, 2018 at 11:49 AM ^

I don't think the data is worthless for decision-making at all (or even close to worthless).  I just think, as someone else said, it requires responsible consumption.  It is one data point in the many-layered process of decision-making - its lack of completeness shouldn't be confused for worthlessness.  It's a data point that suggests NFL defenses are pretty damn good at defending goal-line fades vs. other kinds of passes, and rather less prepared for run plays.  It's not a full-blown model, nor do I think it pretends to be.  But I do think a coach that has this data is better-equipped to make a decision than a coach without it, which puts it somewhere well above worthlessness.


July 10th, 2018 at 7:42 PM ^

You make good points but these are 2 point conversions. It's nothing like a pitcher using a slider and fastball to complement one another. 2 point fades aren't setting up anything.

If you don't think coaches routinely call suboptimal plays then how do you explain the lack of 4th down attempts?

Of course we'd like randomized, controlled studies but that doesn't mean the information is entirely useless. We know QB stats are influenced greatly by the team around them, but they're still worth something. We can evaluate why fades and rollouts are less effective simply by looking at the play. Brian has been bagging on the rollout for years. He had a big row with Space Coyote over it. He gave a compelling argument about why he thinks it's less effective simply based on play design. As someone else said, fades require highly precise throws and timing. 

Leatherstocking Blue

July 9th, 2018 at 7:48 PM ^

Another way to improve soccer is to have the overtime PK's at the beginning of the game so at least the team who has fewer PKs is playing for the full 90 minutes knowing they lose if the game is tied after overtime.

Pepto Bismol

July 10th, 2018 at 9:51 AM ^

This idea is dumb. You're going to spend a half hour before every single game shooting meaningless PKs in front of a crowd that kind of cares, but not really, because more often than not the entire exercise is a complete waste of time.

And then at the end of 120 minutes of actual play that is dead even, when both teams and the crowd are clenching their sphincters for fear of defeat or anticipation of a shootout, you're going to take away any excitement. You're going to tell them who wins a tie, that way we can watch dipshit soccer players drain the clock with endless stalling and fake injuries instead of pushing for a winning goal. 

So you want to kill the excitement at the end of OT as well as neuter the excitement of a game-deciding PK shootout?  That sounds like the dumbest idea of all time. And why? I have no idea. What is the problem that pre-contest PKs solves? Are we having too much fun at the end of soccer games? 


July 9th, 2018 at 8:18 PM ^

Living in Australia for a year made me hate the way we do replay even more.  The NRL in Australia has reviews that take less than 30 seconds, and the the TV broadcasts the review official talking through what they're looking for, what they see, and their ultimate decision as they go. They may not always get it right, but it's fast, and you always know why they made the ruling they did.  I would KILL to have Rugby League-style replay in the US. 


July 9th, 2018 at 10:48 PM ^

Everyone knows the worst part of soccer, and until that's remedied anything that increases the reward for obtaining a yellow card for is going to incentivize more phantom dives. 

Train the refs to understand what a phantom dive is, and use replay. Use your idea to fix this problem - give the other team the ball wherever they want after a dive. 

Soccer would be really compelling if some percentage of hockey toughness was imported to it. 


July 9th, 2018 at 11:54 PM ^

Hell, even Spartan Bob could run a better clock than the FIFA folks.

Soccer VAR time seems wildly off to me. Or, maybe more to the point, the time spent with the review is not being counted as only the review. Think about a typical time VAR is used.

Play in question happens. Team gets hurt feelings because they feel ref got it wrong. A horde of angry players surrounds ref, pointing, cajoling, and getting in his face. Another horde of angry players goes and messes up the penalty spot or carries on some other, similar in spirit, shenanigans. In the meantime, the VAR folks are looking carefully at the play, deciding that the ref done screwed up. They queue up the play, with 2 or 3 clear angles and call the ref over. (Start 538's clock) So all the ref does is roll over and have the VAR say, "lookie here, yew donne screwed up." The ref then goes, "yep, I did." and fixes the call.

Queue the angry horde of players on the other team to repeat the process in reverse.

By the way, I have no patience for millionaire crybabies. And, in my eyes, soccer players take the cake on that one. This is why I could never be a soccer referee (or in any sport for that matter). I'd fling the red card out at the kind of shit these guys pull. Every game I reffed would end as a 7 on 6 contest as long as that crap went on.