My Opinion on Ryan Tannehill

Wonderlic is BS. Poor test for intelligence. It is a good test for testing how well you do on the wonderlic. Waste of everyone's time. The NFL really needs to drop it.

 

yeah but do you think Dan Marino was a brainiac?

 

If you put any faith in total QBR by ESPN (which is the most accurate in my personal opinion), Tannehill has never cracked the top 15 in the league over his five year career. He was 24th last season and 27th overall the year before, which were what most fans refer to as his best years overall. And while I'm not debating that he drastically improved last year, he was still out-performed by all but 6 QB's in the league....including Kaepernick and Osweiler.

Total Rankings for QBR Over Tannehill's Career-

2016- 24th
2015- 27th
2014- 16th
2013- 21st
2012- 18th

Now, I don't have a lot of faith in any metrics if they don't consider the biggest stat of all- winning the game. So personally I wouldn't care if RT ranked 1st or 30th if we have a great season- I'm only sharing this for the folks who are hell-bent on proving that Tannehill is elite He may well be amazing and everything you guys say- there's just 24 QB's that did it better last season.

 

I place zero faith in QBR.

1) It is trying to fix something that isn't broken. I have plotted the data and the NFL passer rating has a .67 correlation to winning since 2002. That's a really strong correlation considering that it doesn't take into account rushing, team defense, fumbles, clock management or special team play.

2) it is a complex model. Generally speaking robust simple models outperform overly complex models. The ideal model captures the most important variables and doesn't worry about insignificant variables.
Complex models tell you more about the assumptions of the model maker than reality.

3) the formula is hidden. No one can check if the model's assumptions are valid.

4) They silently edit and retroactively adjust results. If you look at some of the examples mentioned in this article http://ftw.usatoday.com/2015/11/esp...tom-brady-behind-ryan-fitzpatrick-what-is-qbr you will see ESPN has gone back and adjusted some game QBRs.

5) It automatically penalizes QBs for tacking sacks, without talking into account any reason for the sack. it automaticallyi penalizes a pick 6 as being worse than another interception, but doesn't take into account whether the picks 6 is because of a bad throw, a great return, a tipped pass. These are examples of things that may or may not be outside the QB's control yet QBs are automatically rated the full penalty. QB rushing also seems disproportionately weighted according to some critics.

6) The model keeps changing. For example they apparently removed the 'clutch' part of the model at some point. So you can't get solid historical data on how it has performed over time.

 

Well, I just compared 2016 Total QBR, PFF and regular ratings- the same names are in the top 10 across the board. It looked like Luck danced around the most but for the most part, all three had any given player within 1 ranking of other charts. And in all three, Tannehill was in the mid to lower 20's.

So even if someone hates Total QBR, the argument is the same with any formula you want to tie in- experts see 25+ QB's that are more proficient than Tannehill. And that's fine...little Manning has rarely cracked the top 10 and he's wearing two rings, so raw statistics clearly don't tell the entire story when it comes to winning.

 

Clearly it is not the stats. It is how you get wins as ugly as they can be but a win is a win. As well as getting SB rings on your fingers.

 

"regular ratings" refers to standard passer rating? If so, Tannehill was 12th last year among starting QB's (I think that means 5+ games started on pro-football-reference):
http://www.pro-football-reference.com/years/2016/passing.htm

 

I really wanted to stay out of these type of threads....

First of all, I've yet to see anyone in this thread try to make a case for, much less say that Tannehill is elite.

Second of all, allow me to post this article that highlights how bad ESPN's QBR is as a metric.

http://ftw.usatoday.com/2015/11/esp...tom-brady-behind-ryan-fitzpatrick-what-is-qbr

 

I didn't list Pro Football Reference, I just looked at QBR on NFL.com. I think you're missing the bigger point I was trying to make though- the numbers only tell a portion of the story. Besides the first five games last year when the offense crumbled with no offensive line, I thought Tannehill had a solid year. But when you back out the first five (minus the 2nd half showing against NE) and then he misses the last four, we're really grading him on a body of work over a seven game streak. Statistics, on the other hand, grade out the entire year.

 

nfl.com isn't as convenient because it doesn't allow you to automatically list only "starting" QB's so I prefer pro-football-reference (which also allows you to download the stats). Anyway, among starting QB's Tannehill was 12th in passer rating (and I mean the "regular" passer rating.. ESPN's QBR is really not a stat at all because it has a subjective component).

Personally.. I think that 12th in passer rating over his entire 2016 season is just about where I'd place Tannehill. Average is ~16th and I'd definitely say Tannehill overall was above average, but no I don't think he was top 10 overall last year, so 12th seems really fair to me, certainly more than mid 20's (mid 20's would mean he's way below average and that just doesn't make ANY sense.. at least not in 2016). So in this case I think the "regular" passer rating passes the "smell test" a lot more.

btw.. there's nothing wrong with using "regular" passer rating to compare Tannehill's performance in games 1-5 vs. games 6-13.

A technical note: you have to treat all stats (e.g. completions, attempts, etc..) as part of a single game for each set of games (that is, you can't just average the passer ratings because that's not how they compute a QB's passer rating for an entire year.. they treat all stats as part of a single game).

So, treating stats from games 1-5 as part of a single game you get a 83.6 rating, and for games 6-13 you get 100.13. How does that compare?

Tannehill in games 1-5 played like he was the 24th ranked QB while Tannehill in games 6-13 played like he was the 6th ranked QB. Of course.. this isn't really fair to all other QB's because we're not segmenting their seasons into a "best" and "worst" part so the rankings are probably exaggerated in both directions, but still.. nothing wrong with traditional passer rating.

 

People have taken issue with "traditional passer rating" in these Tannehill threads by the way

 

Another issue with averaging as opposed to computing the paser rating is thatbthe average passer rating tends to be higher than the computed average. It mainly seems to be because teams pass more when behind so the PR in losses tends to account for a higher proportion of your passing total than treating each game as having an equal amount of passing which is the effect of using n average.

 

And as well they should, since it has its flaws.

That said, ESPN's QBR is a lot worse. It's inconsistent and is just not a good metric.

 

but its the same for every qb

 

People take issue with traditional QB rating when people want to act like rating is only about the QB. For a QB to compete a pass, a receiver has to catch a pass. So, passer rating is really a stat for two players.

I dont think that's such a hard concept.

 

No, it isn't. It is HIGHLY subjective.

 

all qbs are evaluated in this particular system under the same rules no?

 

No, because everything about that system is purely judgment calls. Regular QB rating is all about quantifiable stats that there's no real subjectivity about...TDs, INTs, completions, attempts....etc. Those are easily countable. ESPN's QBR is almost entirely subjective...based on weird observations that no one but the coaches and players could accurately say who is at fault for. As such, players like Brady will ALWAYS get the benefit of the doubt for example, while a Cam Newton rarely would. No one uses their QBR. Not teams, not writers outside of ESPN, not football pundits...no one.

 

I get that its not the gospel, I'm just saying they all get judged under the same rules.

 

I wonder if the top 10 guys teams have issues with it

 

Not really though. The subjectivity does not allow it.

 

The rules are fluid Deej. CKap & Wilson could have done the exact same thing, and CKap will get judged harsher than Wilson because of bias or because Wilson's superior defense put him in a better position than the 49ers did for CKap.

You have a weird blind spot with stats and grades based on stats. It is as if you don't really understand them so you consider them all valid.

 

It's not that it's "not gospel". It's useless as a QB evaluation tool. It's an offense evaluation tool. When the offense is doing well then it incorrectly claims that all the QB's decisions were better and their mistakes are minimized. Conversely it does the opposite when an offense is doing poorly.

 

Compare Rodgers and Brady. One of those guys depends much more on receivers making plays happen, yet he gets a crazy high QBR, I'm guessing they're probably similar in QBR, but one is making more difficult throws, and we know Brady isn't as important to winning, by looking at win % when he plays well vs when he plays poorly.

 

I think that's probably the reason, though incidentally if you compare the average of Tannehill's games 1-5 you get 80.94 while combining you get 83.6 so occasionally it goes the other way.

In any case, both approaches have their merits. If you want to look at how efficient a QB is without worrying about game-to-game deviations, then you combine. But if you literally want to know the average QB rating that should be expected from a given QB, then you have to take the average.

So both approaches are correct. They just answer different questions. And paradoxically I think most of the time when people talk about average PR they really want the true average instead of the PR of the combined stats haha.

 

Yeah they do get judged under the same rules at any given time. So from that perspective you're correct.

The problem is that the "rules" are based on some subjective judgment that's hidden in the math. Basically, they take advantage of the fact that there are so many different ways you can try to "adjust" for some variable when you don't have clean enough data. They choose one way instead of many other possible ones.. and in general none of the ways are justifiable because there are so many equally acceptable ones and you don't have sufficient data so that everyone agrees on the method of adjustment. That's the subjective part in their formula.

As Pauly pointed out, they change that formula over time based on .. well who knows?!? The real problem is lack of transparency because I think it's very clear they'd be more than forthcoming with their formula if they thought it would stand up to scrutiny. Because it would probably be shot down to pieces if they were transparent about it (because people would point out they are just arbitrarily using one method instead of another to make adjustments), they don't publish it.

And to be clear, from a purely mathematical standpoint it's easy to get higher correlation to win% (add more variables, have a computer tune those variables to the data you want to predict, and voila!). But those extra variables or how you're relating them in the math probably are meaningless from a football standpoint. So just having a formula with greater predictive power isn't something you'd aim to keep proprietary. It's easy to produce something like that if you don't care about how the math represents the real world.

 

Ok, fair enough..

 

Isn't passer rating the same kind of BS?, I mean to them an int has no context under their rules, points dropped regardless..how bout yardage per game..no context for a hitch that the receiver takes it 50 yards or a bomb perfectly placed..all the same..

 

Well.. the first difference is that passer rating is transparent. We know what they're doing so it's clear what the problems with their assumptions are.

The primary problem with passer rating is that the relative weights on the different components (e.g. Y/A, comp%, TD%, INT%) are in fact arbitrary and were set by experts way back in the 70's. If someone were to build a passer rating formula today, they'd probably get those weights in a more principled manner by having a computer find out which set of weights best predicts the data. They'd probably also add more turnover data in the formula because turnover differential has a very high correlation to win%.

What you're talking about is that passer rating is only a crude approximation even if the weights were the best possible. That is, as you point out, there's no context for essentially every single event.

Here's the problem. While your critique is in principle valid, in practice no one knows how to put those contexts into a formula so that the formula is justifiable. This is precisely the problem DVOA and ESPN's QBR ran into. Once you try to put those contexts in.. say you precisely define a method of adjustment that you think is correct in a given situation (e.g. QB throws a catchable ball but WR drops it so you say for "catchable" balls you'll give the QB 100% credit).. you find out that there are TONS of unintended consequences you weren't thinking of.

That's precisely what's happening with the counterintuitive results of ESPN's QBR. People try to put the correct context in one situation, but that creates unintended results elsewhere.

This problem is often seen in scientific modeling. Put too many parameters in and you actually start to do worse! So simplicity is a virtue. Not only is everything transparent, it actually often works better. Having said that, I think one can do a lot better than passer rating but there's just this cultural inertia that's hard to overcome (first-mover effect).

 

Are you talking about qbr?

Deej is talking about traditional passer rating. That doesn't change, is not fluid, and there is no subjectivity.

 

You sure he wasn't talking about ESPN's QBR.....?

 

I just went back and it's ambiguous. Lol. The post he quoted talked about and criticized both.

Deej you talking about passer rating or qbr??

Raf's post sounds like criticism of passer rating hence more confusion

 

Raf is talking about QBR and not qb rating, I'm fairly certain.

 

Could be. If so, fark qbr. There should be a new rule in the forum, no citation of QBR it sucks hard. I automatically dismiss a post that uses it and move on

 

Honestly I would not. There's not much else I hold higher than my integrity. Of course I've lied in my life, but I'm not proud of it and I do my best to not make those same mistakes. However, in the game of football if one is willing to do what the Pats have done they are pretty low on the scumbag scale.

 

Hey wait a minute.. I'm on here a lot, I don't hate my job. I just... maybe have too much time on my hands. Although, in my defense if the IT guys see this.. I'm on my lunch break I have no idea why I just said any of this. Carry on.

 

I didn't watch much of Marino but the fake spike tells me all I need to know about his football smarts. Tannehill is not football smart, which means his ceiling is something less than elite.

 

Fake spike wasn't Marino's idea.

It was a combo of Kosar, Marino, Shula and Stevens. Shula even said they had practiced it all week.

http://www.espn.com/nfl/story/_/id/...-dolphins-unforgettable-victory-new-york-jets

 

I did a look into it last year the 4 components of the passer rating (yards/attempt, dompletion %, TD % and int%) were all rated equally when you use 1970s passing averages. What has happened over time is that with modern figures completion% and int% end up count as weighted twice as much as TD% and yards/attempt due to how the passer rating formula works.

What I haven't looked into is whether or not this is wrong. The fact that coaches and plays still seem to give serious regarc to passer rating, over and above alternate methods, such as yards/attempt, QBR, DVOA, or PFF rating, indicate to me that it is in the right ballpark at least.

The biggest specific complaint is that compleion% counts for too much. Many people have made convincing arguments to that effect, and that big plays should count for more. However it might be that you need to be on the field to make big plays, so completion% might actually be more important. Without doing the testing Cbrad suggested we really don't know.

 

The diifernce is thatbthe NFL passer rating is a measure os outcomes. The ESPN QBR is supposed to be a measure of process.

For example QBR removes yac from their rating because they don't consider yac to be a QB skill. However growing up seeing Joe Montana hitting Jerry Rice perfectly in stride to turn a 5 yard slant into an 80 yard TD I for one would be loathe to remove yac from my assessment of a QB. Yes sometimes you'll have Landry pinball his way through the entire opposing team, but if a QB is consistently hitting receivers in stride that counts for something and should be meaured if you are evaluating process not outcomes.