How Does Gase Have The Audacity To Be Irritated?

and im certainly not calling PFF infallible, im just saying, especially for the QB position, a subjective stat watching film is much more important than a blanket stat taking into account only blanket stats.

 

So just to prove my point, if I throw a pass that would have led to a TD if thrown in front of the reciever and it would have led to a TD if properly led the reciever but i throw 3 yards behind the reciever and it only leads to an 8 yard gain, and i perfectly throw a pass in between two defenders that if thrown perfectly at most would have gained 8 yards, does QB rating distinguish between the 2? No. and that is exactly the point.

 

Read post #67. We already went through this. In that particular case of course passer rating would be worse than pff. But on average passer rating can be proven to be better in terms of how much it captures what matters for winning.

 

dude, what dont you get about there being 21 other positions on the team that better determine winning than a position/stat that has no room for nuance? You argument in post #67 says "how often does that happen?" It literally happens every god damn play. A QB being given the advantage just because of his supporting cast CAN happen every single play. That is the exact point. One takes this into account every single play, the other does not.

 

Dude what don't you get about the inability of pff graders to demonstrably do better in figuring out what aspects of play the QB influences matter most for winning?

And you're not winning an argument with me by creating a strawman.. we went through that one in post #64 already.

 

Seriously? OK.. just so we know where you stand. You are now claiming the post I responded to, post #63, happens "every god damn play". What did you say in post #63? This:

Every pass behind the receivers eh? Happens on every god damn play eh? Yeah right.

 

and again, what dont you get that there are hundreds if not thousands of other variables that go into effect of just winning percentage than just QB play. If drew brees plays the same game as trevor simien last year on the opposite teams, you realize their winning percentages are completely different because there are so many other variables such as the entire other side of the ball?

Again, you are neglecting, because of the use of 1 stat, to take into account the starters at 21 other positions. I can make it home from school in 20 minutes on average and that might be completely and directly correlated to when i leave school, but that completely fails to take into account independent variables like weather, general unpredictability of traffic, how big of a hurry im in, etc. What dont you get about the fact that plain blanket formulas are not sufficient to take into account things that are so reliant on things that are completely out of the control of the QB that Qb rating doesnt take into account like O-line play, WR talent, Defensive talent, etc.

 

No, but you know what does happen on every god damn play? a subjective system that views that possibility as entirely possible on any given play and takes that into account in their grades? You know what doesnt take that into account? a blanket formula like Qb rating.

 

OK either you just don't have the ability to understand the arguments I've given or you don't want to move away from your strawman argument that I'm somehow claiming other positions don't matter.

Doesn't matter anymore. You're simply wrong. I've shown passer rating better captures what matters for winning than pff QB rating. I'll let you argue with yourself.. unless you stop the strawman arguments and demonstrate you can actually comprehend what I'm saying.

Peace out dude.

 

and again, you've yet to prove that QB rating and its correlation to winning % accurately takes into account the hundreds of other variables that effect winning % merely than just that stat. It's literally just like me saying that it rains more when i drive home than when i take the metrorail home without taking into consideration the myriad of other factors. Again, correlation does not equal causation, and you have yet to refute my points that the same qb rating can be achieved by a myriad of both great QB play and horrific QB play. Hence why the admittedly subjective stats at PFF are better than blanket formulas that have no room for nuance.

 

What exactly is my strawman? That PFF takes every play individually regardless of supporting cast and qb rating is a basic stat formula?

 

Read post #64. And then read your post #87. And then maybe read the explanations I gave in-between.

 

Fair enough, but I don't think its a coincidence that Ryan's best years came at the same time as their best offensive line ratings (minus first year, when we know he was a project). It might not amount to much in QBR rating as an average, but what are we looking for? We aren't hoping for a normal adjustment based on the average of expected result. We are hoping for that diamond in the rough. We are hoping he has something special that elevates him higher then what the normal expected result would be.

If our line moving up 3 spots (as they did from 2015 to 2016) on the protection list gives Tannehill +5 QBR rating(which he had), I would take that all day long! I know I know, that's not how statistics work, and the math is lazy at best. But there is always the hope that Tannehill is a statistical oddity. I mean his rating really does seem to dance in step with the line protection ranking...

I'm also not saying that its a lock that we will see this transformation, I'm just saying if we get the line humming right around top 10-15 and Ryan all of a sudden looks like Joe Montana back there, we will all be pretty happy we held on to him.

We know what Ryan is behind a bad O-line. Some of us think he did a pretty good job in a bad situation, some of us thinks he didn't do enough to earn franchise QB status. I honestly can't fault either view right now, cause as I said, I don't feel like I've gotten a good chance to evaluate him entirely. I personally want to see him play behind a solid offensive line. My fear is that Miami doesn't gets it together, and Ryan moves on and has his best years playing for another team.

 

Ive read your moronic post #64. It continues to base itself off the opinion of this beyond stupid opinion of "Win% is precisely what matters most, because you are interested in how a player affects the final outcome you care about. So yes there is a correct answer here. The correct answer is whatever pff's intentions or approach, they can't estimate the type of QB ability that matters for increasing win% better than passer rating."


And again, using team win % as a way to determine the effectiveness of a QB stat is beyond stupid by anyone who understands math or the game of football. What you continually seem to not understand is that, despite what might be a "correlation" is not a causation. Just like in the extreme example of it might rain every time i go grocery shopping, me going grocery shopping is not the cause of the rain. A QB can have a great QB rating due to factors wholly out of his control that have an effect on winning percentage.

For yet another extreme example to show how wrong you are, imagine a team gets the ball first from the 1 and the QB throws a screen that goes right through the hands of the CB for a would be pick 6 and the WR runs 99 yards for the TD and the opposing QB throws a pick 6 every time he throws the ball. The QB would have a perfect 158.3 QB rating without taking into consideration that he only threw one pass or the context of the fact that that 1 pass should have gone for an easy pick 6. Watching the context of that one play (every play) would show that nuance that a formula stat wouldnt show, which is that the QB actually performed terribly over that 1 pass (1 game) but the team win percentage remained the same despite the fact that he performed terribly and got perfect results wholly out of his control.

 

Actually everyone who understands how to apply math to sports uses one of the two metrics I described to determine how good a stat is: prediction of future outcomes, or correlation to win%.

But you lack the capacity to understand why so we'll just end it at that.

 

no, you seem to not understand that individual stats and their correlation to win % and future outcomes does not = causation or give credence to those stats.

When you can give me an explanation for how the same exact statistical play can be attributed to the fact that a QB can have completely different impacts on a play will end this discussion. Yet you cant. You cant differentiate between a 99 yard TD pass which is the sole result of the QB being incredible or the 99 yard TD pass that is the sole result of a mistake of a QB which still results from a CB error + WR great play. Both result in the same QB rating. Ill end my argument when you can differentiate between those two and realize that they come out with the same exact QB rating despite complete and utter different QB play. That is why the individualized PFF rating, despite subjectivity, is so much better than a basic formula.

 

In 2015 PFF had Cutler ranked 15th in tier two and Tannehill ranked 19th in tier three.

 

Crippled? Now that's funny.

Apparently you're completely redefining the definition of that word.

 

Ahhhhh the ol "he ruined our Luck chances" nonsense.

Now I get your sourpuss act regarding Moore.

 

It
It's graded on a curve Cbrad. I know us smart kids hate that, but 126 is close enough for Dolfans

 

The whole idea of perfect passer rating is absurd anyway. There is no theoretical upper limit to passer rating so why don't they just list the actual passer rating when it goes above 158.3??

Here's a list of all perfect passer ratings:
https://en.wikipedia.org/wiki/List_of_NFL_quarterbacks_who_have_posted_a_perfect_passer_rating

If you calculate the actual passer ratings for each one of those QB's, the absolute highest ever (get ready for this) is Frank Ryan in 1964 with a 271.96 rating LOL!!

The highest in the SB era was Fran Tarkenton in 1970 with a 228.94 rating. Overall he comes in #3 on that list.

Tannehill comes in 12th all time with a 213.05 rating haha!

 

Oh stop! You had me at absurd

And Fran was a bad mofo. In todays game...sheeit!

 

And Trump won the nomination for the Republicans. Go figure.

 

OK... I just went back and read all the posts in this thread and I just want To say.

Cbrad and Phins18................................


BEST NERD FIGHT EVVERRRR!!!!!

 

PFF's ratings do that (the boldfaced portion) in theory only.

What tells you that -- in practice -- their ratings actually do that validly?

 

As with what I stated just above, PFF purports to control for those sorts of extraneous factors in theory.

What tells anyone that PFF is accomplishing that task validly? That they're simply trying?

 

Over the short haul, yes. Over the long haul, no.

If the question should arise of whether Peyton Manning for example was any good, nobody says, "wait -- let me look at the film first." A simple glance at his career statistics tells the tale. Looking at film alone on the other hand, one might choose by accident a particularly poor sample of Manning's play and mistakenly conclude that he was terrible.

Conversely, in just one game for example (or any such relatively smaller sample of play), it's possible that a QB's passer rating was artificially inflated or deflated by the play around him. Then yes, a look at the film tells all.

Each approach has its strengths and weaknesses. My point above, however, is that there is no way of knowing whether PFF's film-watching validly accomplishes its goal. They could be trying to parse QB play from the play of those around the QB, and failing. How do we know they aren't?

 

And you're implicitly echoing there what I've stated above about large versus small sample sizes, by focusing on a very small sample (one play), for which the film would provide the more accurate portrayal than the statistic.

As the sample size increases, however, the film loses its ability to be more valid than the statistic. And again, there is no way of knowing whether PFF's purported use of film for large samples has any validity.

 

Everyone here is forgetting one intangible that can't be measured by statistics and that is leadership. So many times I've heard the argument that great quarterbacks elevate the level of play of the players around him. Now, whether or not Tannehill is a "great quarterback" is another debate all together but one thing is definitely certain...he has the respect of the team and of the offense. Matt Moore ALSO has that respect. Whether or not Cutler has that same level of respect and has earned the trust of the team, I don't know but in all reality, Cutler is a stranger in Miami.

Someone said earlier that last season Miami started 1-4 before Gase was able to get the team on track to make the run it did. Perhaps with a new quarterback, the team is going through the same growing pains as it did last year. Who knows...

 

If there's one part of football where stats don't/can't tell much of the story it's in the O-Line. I think the "eye test" (hopefully by someone who knows what they're looking at/for) is the best way to determine o-line effectiveness. Of course, rushing yards, sacks, passer ratings etc MUST have some correlation with o-line play, but it's damn near impossible to know how much.

 

You're leaving out the important part of PFF's method: the person watching the game. Person A may grade differently, even when given the same instructions, than Person B. Passer rating, as cbrad has proven, is better than PFF's methods.

 

Personally, I wouldn't accept a correlation of less than (the absolute value of) 0.5 for any candidate OL stat, whether it's to passer rating for pass blocking or to a key rushing stat like YPG for run blocking. The reason for the 0.5 is that for some reason almost all useful stats in football have linear relations to win% over much of their range (TD:INT ratio is the one exception because that's a ratio and you naturally get a skewed distribution).

And 0.5 correlation for a linear relation implies the square of that = 0.25 = 25% of the variation observed is due to whatever the stat captures. And I'd say the absolute minimum I'm willing to accept for OL influence on the passing game is 25%, and it clearly could be more.

 

I don't think it's better. I think its different. Clearly there are huge problems with simply going by passer rating, as phins18 has pointed out.

 

But if you had to choose just one passer rating is provably better. And I'll go so far as to say that PFF's method is really not worth reading. Now, if an NFL coach watched a play, or a game, and said, "The o-line played outstanding" I'd more often than not believe him. However, when some random guys on PFF say that I'll take it with a grain of salt. Why? Because I've actually seen some of their rankings and then watched the player(s) and saw something completely different. Not saying that I'm right or wrong, but saying that the guys working there often don't come to the same conclusions either.

 

I actually think it’s smaller than that, and the reason is because there is relatively little variation among the league’s offensive lines, in terms of quantifiable measures of their play, far less than the variation in the quantifiable measures of quarterback play.

As I’m sure you know, under those conditions a correlation is necessarily limited.

 

Well, I think to really dismiss PFF ratings, you'd need to look at all their ratings of say, QBs, then watch tape, and see if you still come away thinking that they're inaccurate as a whole, or of they're any more inaccurate than rating can be (for the reasons given already in this thread).

 

Umm.. there are two (serious) problems with this argument:

1) The variation in OL play measured by common stats.. say sacks or sack%.. is in different units than the variation in other measures, like passer rating or passing yards per attempt. So there's no way to directly compare and say which variation is larger or smaller because you're talking about different units.

For example, in 2016 the standard deviation in total sacks for starting QB's is 8.14 sacks, while the standard deviation in passer rating is 11.06 passer rating points and the standard deviation in yards per attempt is 0.699 yards. Which is larger or smaller? It's a meaningless question because the units aren't the same.

2) The claim that correlation becomes smaller as the variance in one variable gets smaller is simply false (except under the exceptional case where there is zero variance, in which case correlation is undefined). Easiest way to prove this: suppose you have random variables X and Y and you plot X vs. Y. Suppose the correlation is perfect at 1, meaning that the points lie perfectly along a line. What does that tell you about the slope of the line? Nothing!

That line could be ANY slope except 0 or infinity and still have the exact same correlation of 1. The slope matters because that tells you the relative difference in magnitude between the variances of the two random variables. Unity slope means the variances are equal while a higher slope means Y has a higher variance.

In general, multiplying one random variable by a non-negative constant will change its variance but NOT affect its correlation to another random variable.

 

Are you saying that if you have two normal distributions and one of them is sharply peaked and the other is extremely mound-shaped, the two variables involved could be strongly correlated?

 

Absolutely. That last sentence in my previous post is all you need to test. Just take any two sets of numbers you have available and calculate the correlation. Now, multiply every number in the first set by say 10 (which will demonstrably increase the variance of that set) and recalculate the correlation. They'll be identical.

 

OK thanks — appreciate the feedback. And so then one would need to correlate the Z-scores, correct?