Merge: Report: Joe Philbin wanted to replace Ryan Tannehill with Derek Carr

"Fin D wrote:

So I'm clear, all QBs are not the same but the oline play effects them all the same?

If offensive line play affected them all the same, the correlation between offensive line and QB play would be 1.0.

Again, the weakness of the real correlation suggests exactly the point you're making, that there are QBs who are affected more, and QBs who are affected less, by offensive line play. Presumably the driving force in the degree to which they're affected by offensive line play is their own individual ability.

A QB with poorer individual ability needs a better offensive line, and a QB with better individual ability doesn't.

"Silverphin wrote:

Not a knock against Carr, because I legit like the cut of his gib, but that doesn;t mean much.

i know..

"Laurence wrote:

If offensive line play affected them all the same, the correlation between offensive line and QB play would be 1.0.

Again, the weakness of the real correlation suggests exactly the point you're making, that there are QBs who are affected more, and QBs who are affected less, by offensive line play. Presumably the driving force in the degree to which they're affected by offensive line play is their own individual ability.

A QB with poorer individual ability needs a better offensive line, and a QB with better individual ability doesn't.

No that's just one possibility. If it wasn't just one possibility then it means there's two degrees to an oline, bad and good.

Another possibility is that the QB isn't poor, but handicapped by other issues that limit his ability to deal with a poor oline, like making audibles and/or having a running game.

"Fin D wrote:

No that's just one possibility. If it wasn't just one possibility then it means there's two degrees to an oline, bad and good.

Another possibility is that the QB isn't poor, but handicapped by other issues that limit his ability to deal with a poor oline, like making audibles and/or having a running game.

That's definitely another possibility, but it certainly won't be reflected in the webpage that got this interchange started. What's reflected in that webpage is 1) a weak correlation between a pass blocking statistic and QB play, and 2) that a change in the Dolphins' offensive line alone, to the level of the Raiders' offensive line, wouldn't have done much of anything appreciable to Ryan Tannehill's DVOA.

So we can hypothesize about the effects of many different variables, but as we do that, we get further afield from any ability to measure them objectively, and further afield from any ability to confirm or deny them.

In other words, people proposing such complex interactions among many variables are really up against it in confirming those possibilities objectively. They're essentially in a world in which they're asking others to trust their expertise about a situation, when there is no known expertise to trust.

"Laurence wrote:

That's definitely another possibility, but it certainly won't be reflected in the webpage that got this interchange started. What's reflected in that webpage is 1) a weak correlation between a pass blocking statistic and QB play, and 2) that a change in the Dolphins' offensive line alone, to the level of the Raiders' offensive line, wouldn't have done much of anything appreciable to Ryan Tannehill's DVOA.

So we can hypothesize about the effects of many different variables, but as we do that, we get further afield from any ability to measure them objectively, and further afield from any ability to confirm or deny them.

In other words, people proposing such complex interactions among many variables are really up against it in confirming those possibilities objectively. They're essentially in a world in which they're asking others to trust their expertise about a situation, when there is no known expertise to trust.

This is the point where stats cause problems. They aren't the end all be all. They are just a tool. A foot long ruler is pretty accurate, but you don't use it to measure miles. There other tools in the kit that help understand how much a oline affects QB play. That tool is common sense.

"Fin D wrote:

This is the point where stats cause problems. They aren't the end all be all. They are just a tool. A foot long ruler is pretty accurate, but you don't use it to measure miles. There other tools in the kit that help understand how much a oline affects QB play. That tool is common sense.

Tell that to the person who posted the webpage and cited it as support for the idea that Ryan Tannehill would've benefitted from Derek Carr's offensive line play. Why did he bother posting that if common sense was all he needed?

Geez, dude. I can only assume you've returned.

Anyway, the webpage, if you're referring to the one I linked to, was specifically about RUNNING the ball. They also included a small bit about pass blocking. The Raiders oline was much, MUCH better at run blocking than Miami's. If you read the page, they explained that. So, is not as simple as looking at the pass blocking of each team, and taking the square root of the QB, and determining that if Tannehill had the Raiders pass blocking in Miami he'd only go up a couple points. In that scenario you are completely discounting the run game.

"Laurence wrote:

Tell that to the person who posted the webpage and cited it as support for the idea that Ryan Tannehill would've benefitted from Derek Carr's offensive line play. Why did he bother posting that if common sense was all he needed?

The only common sense needed is by you. I posted a link to mostly run blocking stats, and you went all ape on pass blocking.

POINT
|
|
|
|
|
|
|
\/
YOU

"resnor wrote:

Geez, dude. I can only assume you've returned.

Anyway, the webpage, if you're referring to the one I linked to, was specifically about RUNNING the ball. They also included a small bit about pass blocking. The Raiders oline was much, MUCH better at run blocking than Miami's. If you read the page, they explained that. So, is not as simple as looking at the pass blocking of each team, and taking the square root of the QB, and determining that if Tannehill had the Raiders pass blocking in Miami he'd only go up a couple points. In that scenario you are completely discounting the run game.

Of course now we could do the same thing we did above with regard to quarterbacks and offensive lines, where we talk about the measurable effects of the running game on quarterbacks, but in the end the explanation for Ryan Tannehill's performance will come down to an interplay of many variables that can't be measured (coaching, audibles, etc.), and whose relationships with each other can't be confirmed or denied. If that's where we're going to end up -- and we will -- then I'd rather just skip it.

We can all have our opinions. Again, what suggests the reality of the situation in my opinion is the consensus "balance" that's derived from all these interchanges, and that consensus balance suggests that we're talking about an average QB.

"Laurence wrote:

Of course now we could do the same thing we did above with regard to quarterbacks and offensive lines, where we talk about the measurable effects of the running game on quarterbacks, but in the end the explanation for Ryan Tannehill's performance will come down to an interplay of many variables that can't be measured (coaching, audibles, etc.), and whose relationships with each other can't be confirmed or denied. If that's where we're going to end up -- and we will -- then I'd rather just skip it.

We can all have our opinions. Again, what suggests the reality of the situation in my opinion is the consensus "balance" that's derived from all these interchanges, and that consensus balance suggests that we're talking about an average QB.

What consensus are you taking exactly?

Yeah, I suppose it's just better to pretend that the run game, or lack there of, has no impact on the QB, or offense as a whole.

It just boggles my mind that so many teams bother doing something that seems to have no impact on the offense and it's production.

"cbrad wrote:

Why are you lumping me into the group? What I said is true. What you said is also generally correct but my argument from the outset was DVOA isn't objective (like you're saying).

And my argument rests on nearly perfect evidence for that even without someone showing the formula: there is NO way they get the weights they do from any optimization method (which implicitly uses partial derivatives) unless someone adds subjective weights to them. No expected win probabilities can tease apart the contribution of the OL to a running back's performance, etc..

As for what will work, the point is nothing will if you just use common stats. That was my entire argument starting in post #216.

I was just saying that discussing the subjectivity/objectivity of FO doesn't change that the initial proposition was an inappropriate use of stats

"resnor wrote:

Yeah, I suppose it's just better to pretend that the run game, or lack there of, has no impact on the QB...

There is an impact, but statistically speaking, as with offensive lines, it's much less than some people seem to believe. Again though, when the end point of the discussion is an untestable and unfalsifiable position, there's no use in exploring any of that. You might as well just lay your untestable and unfalsifiable position out there, call it your opinion, say it once, and leave it at that.

"Laurence wrote:

There is an impact, but statistically speaking, as with offensive lines, it's much less than some people seem to believe. Again though, when the end point of the discussion is an untestable and unfalsifiable position, there's no use in exploring any of that. You might as well just lay your untestable and unfalsifiable position out there, call it your opinion, say it once, and leave it at that.

Is that your plan?

"Pauly wrote:

I was just saying that discussing the subjectivity/objectivity of FO doesn't change that the initial proposition was an inappropriate use of stats

There is nothing inappropriate about determining a predicted value based on regression equation.

"Laurence wrote:

There is nothing inappropriate about determining a predicted value based on regression equation.

Are you sure you're right about that? :shifty:

"Laurence wrote:

There is nothing inappropriate about determining a predicted value based on regression equation.

There's nothing wrong with using regression to make predictions. The problem was the input to the analysis, DVOA, which includes a stat that claims to measure OL ability. So that's not a knock on you per se.

I'd say the only part "wrong" with the use of regressions (what could be blamed on you) was adding the effect of OL to Gase. Adding them means you're assuming the two factors are independent which I'm sure you'd agree they are very far from being.

"cbrad wrote:

There's nothing wrong with using regression to make predictions. The problem was the input to the analysis, DVOA, which includes a stat that claims to measure OL ability. So that not a knock on you per se. I'd say the only part "wrong" with the use of regressions (what could be blamed on you) was adding the effect of OL to Gase. Adding them means you're assuming the two factors are independent which I'm sure you'd agree they are very far from being.

Right. It was more like a "best-case scenario" approach, based on the two external variables (coaching and offensive line) thought to be most influential. The more interesting finding in my opinion is that Gase's "effect" on both Manning and Cutler appears to have been much less than what's been touted in the media. They didn't deviate nearly as much from their career averages (beyond their rookie years, which I didn't include) than one would think, based on the media hype. In fact, both QBs had better years previously in their careers.

"Laurence wrote:

There is nothing inappropriate about determining a predicted value based on regression equation.

There is when what you're using is a dependent partial derivative.

This is one of the best worked examples on teh web: http://ocw.mit.edu/courses/mathematics/18-02sc-multivariable-calculus-fall-2010/2.-partial-derivatives/part-c-lagrange-multipliers-and-constrained-differentials/session-43-clearer-notation/MIT18_02SC_MNotes_n1_2.pdf

"Pauly wrote:

There is when what you're using is a dependent partial derivative.

This is one of the best worked examples on teh web: http://ocw.mit.edu/courses/mathematics/18-02sc-multivariable-calculus-fall-2010/2.-partial-derivatives/part-c-lagrange-multipliers-and-constrained-differentials/session-43-clearer-notation/MIT18_02SC_MNotes_n1_2.pdf

Well.. that's just a fancy way of saying when the data isn't best described by a line (plane, etc..) linear regression isn't a great idea.

I don't think we need to be that picky here, especially since we don't know what the underlying data look like :wink2:

Skimming through.. Had to stop reading at "weak correlation between pass blocking and QB play". LOL. Dear lord that's rich.
Is it just me or is this thread is starting to feel very shouright-y?

"Fin D wrote:

This is the point where stats cause problems. They aren't the end all be all. They are just a tool. A foot long ruler is pretty accurate, but you don't use it to measure miles. There other tools in the kit that help understand how much a oline affects QB play. That tool is common sense.

One further thought on the bolded portion above: there are times when common sense is wrong and "uncommon sense" is right. If we don't explore that possibility by using something objective to measure the variables involved in the situation and determining how they interact, we're prone to making the mistake of thinking our common sense is correct, when in fact it might not be.

At the very least, that exploration could leave us with a "we don't know for sure," rather than too certain a belief that our common sense is correct about the matter.

In this case, for example, it's entirely possible that there is such little variation in offensive lines across the league that it can't possibly explain the much larger variation in QB play. Another possibility is that the individual ability of QBs is so influential that it outweighs the variation in offensive lines by a significant margin.

Either of those possibilities, if true, would indicate that the common-sense belief that QBs are affected significantly by offensive lines is incorrect.

Then there is another possibility, as well, that the variation in offensive line play is actually caused by the variation in QB play. That finding, if true, would also nullify the common-sense belief in the relationship between offensive line and QB play, by effectively reversing it. QB play would cause offensive line play, rather than vice-versa.

All of these are possibilities. We certainly shouldn't rest on believing our common sense is correct in the absence of even any mere exploration of those possibilities. Human beings aren't that great at knowing the truth about things based on our intuition alone.

"dcafiero wrote:

Skimming through.. Had to stop reading at "weak correlation between pass blocking and QB play". LOL. Dear lord that's rich.

What's even more rich is when people don't have the basic humility to admit that the issue should be subjected to some objective exploration, rather than thinking they're correct about the matter simply because they believe they are.

Subject the matter to exploration. You can still be right in the end, but at least it'll be based on something more than just a personal belief. If you "stop reading" before you do that, you may fail to learn something.

Sometimes these debates revolve around the participants' openness to learning. When someone shows a reluctance to learn for some reason, there is really no room for purposeful debate, because the possibility for the consideration and acceptance of new information is nil.

"dcafiero wrote:

Skimming through.. Had to stop reading at "weak correlation between pass blocking and QB play". LOL. Dear lord that's rich.
Is it just me or is this thread is starting to feel very shouright-y?

I don't know, but it does feel pathetic and sad.

"Laurence wrote:

What's even more rich is when people don't have the basic humility to admit that the issue should be subjected to some objective exploration, rather than thinking they're correct about the matter simply because they believe they are.

Subject the matter to exploration. You can still be right in the end, but at least it'll be based on something more than just a personal belief. If you "stop reading" before you do that, you may fail to learn something.

Sometimes these debates revolve around the participants' openness to learning. When someone shows a reluctance to learn for some reason, there is really no room for purposeful debate, because the possibility for the consideration and acceptance of new information is nil.

No.

Your "objective data" is incomplete. It doesn't tell the whole story, which is par for the course on stats and all tools.

You over rely on stats and have for years. The data is there, its been put into the debate, it didn't however end it.

You've literally ignored common sense and tape since you've been posting here, and every time, you get your ass handed to you.

Hell, you never even acknowledged that it takes two people doing their jobs correctly to create a completion. Its that kind of common sense denial that gets you banned every time you come back.