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

More insight in the clown show that was around Tannehill: http://www.thephinsider.com/2016/1/...ilbin-relationship-details-continue-to-emerge


Just disgusted that Philbin and Lazor basically set up an situation/environment for Tannehill to fail. Tannehill may not break out but I know for sure he's a viable starter in this league if he can put up decent to good numbers when the head coach and the offensive coordinator is trying to take you down.

 

Yeah, the optics aren't looking good for Philbin regarding Tannehill. Still.. there are many other parts to the story we're probably not hearing so who's right, who's wrong, what really happened, who knows. Either way, the proof that Tannehill is a viable starter has to be demonstrated on the field this year.

 

Not sure how we can come to that conclusion based on hearsay, and add to it that Tanne's best year, and biggest improvement came under Lazor.

Philbin was a lousy HC, but I find it hard to believe that he would be so stupid as to sabotage his own career because he thought Carr was a better option.

 

everything is on the record somewhere on this site, if you want to know what I said about a player let me know, wrong or right, the game is to be right more.

 

the qb plays under control when he needs to and unleashes playmaking ability when he needs to...

 

lol wth..you can mos def eval a player in isolation regardless of how good or bad the team is. and project that talent alongside better variables...sometimes its easier when the team is bad..

 

And when does Tannehill ever do that? The offense doesn't help the defense with the slow starts and going 3 and out.

 

Sure, technically you can do it, but it makes no sense and doesn't give you useable results.

I could isolate and evaluate Fighter X against one opponent and then isolate and evaluate Fighter Y against multiple opponents at once. If I were to then compare X & Y, my comparison would be meaningless, because I'd be comparing two different scenarios that require different skills. Now the individual evals have merit, as long as they are couched correctly, like "Fighter Y has trouble with multiple opponents", and not "Fighter Y is a poor fighter."

Which is why there's a problem here with Thill. You still refuse to admit there's different kinds/levels of pressure. It weakens your entire argument. The reason it weakens it, is because it shows your zealotry for the subject matter. When you can't even admit there's a massive difference between being rushed by 3 guys at once and being rushed by 1, then you're clearly not in a position to discuss this...nor can your eval be trusted.

 

You can do this, however:

There was a 0.62 correlation between Football Outsiders' Adjusted Line Yards (a measure of offensive line play) and quarterbacks' DVOA.

If the Dolphins were to have the very best Adjusted Line Yards in the league, then based on the correlation between Adjusted Line Yards and quarterbacks' DVOA, it would most likely be associated with a change in Ryan Tannehill's DVOA from -10.6 (his 2015 value) to 12.61.

Based on the 2015 correlation between quarterbacks' DVOA and win percentage, that would likely be associated with a 51.5 win percentage, or roughly an 8-8 record.

So an improvement in the Dolphins' offensive line, such that the team would have the very best offensive line in the league, via its impact on Ryan Tannehill's play, would likely be associated with two more wins in a season.

Now, let's look at what's more likely. Let's say the Dolphins' offensive line improves to one standard deviation above the league average in Adjusted Line Yards. Less than the best in the league, but significantly above average if you will.

That would likely be associated with an increase in Ryan Tannehill's DVOA from -10.6 to 7.73, which would in turn likely be associated with a 48.6 win percentage, or, again, roughly an 8-8 record, or two more wins in a season.

Now, let's add an improvement in the Dolphins' offensive line play to one standard deviation above the league mean, to the effect on Ryan Tannehill most likely to be produced by the presence of Adam Gase.

Based on the deviation from Peyton Manning's career average DVOA associated with Gase's presence, averaged with the deviation from Jay Cutler's career average DVOA associated with Gase's presence, Tannehill's DVOA should increase from -10.6 to -0.68 with Gase, which, based on DVOA's 2015 correlation with win percentage in the NFL (0.48), would be associated with one more win in a regular season.

The combination of improved offensive line play (to a standard deviation above the league mean) and the presence of Gase -- if all goes as expected based on the above historical information -- would boost Ryan Tannehill's DVOA to 28.25, which alone would likely be associated with a 61 win percentage, or just under a 10-6 record.

This is the sort of change the team is looking for, obviously.

 

While I recognize this supports my argument, I don't feel like its an accurate way to look at things.

All QBs don't produce the same way under the same circumstances. They are human. QB X might play better than QB Y if they were behind the same bad oline, but QB Y plays better than QB X with poor receivers...so on and so forth.

The quality of the OC and his play calling, design and use of the run game factor in a lot as well. As does being able to audible. All things, Thill had to deal with.

 

Right, but there will be no more reliable "crystal ball" available than what I provided above. Everything else will pale in comparison, with the observational judgments of deeply emotionally invested fans coming in right after last.

 

Phenomenal post. 5 stars.

Sent from my SAMSUNG-SM-G920A using Tapatalk

 

I just don't think you need a "crystal ball" in this scenario. We already have evidence he plays considerably better with an average performing line. We also know that being able to audible and run the ball makes any QB better.

These are all common (football) sense concepts.

 

I like the idea here, but Adjusted Line Yards seems to be a totally subjective stat, at least according to the brief description here:
http://www.footballoutsiders.com/info/methods

They're trying to separate out the effect of the OL from all other factors and they don't mention any principled way of doing it, so the stat for me is bogus. I think it's pretty clear that depending on the (subjective) weights they put on the OL influencing the outcome you get totally different results in your analysis.

The second problem with that stat is they claim it's ONLY a measure for the running game. I think people are more interested in what the effect would be if there was less pressure on the QB (due to the OL), so if someone can show expected gains in wins, points or passing efficiency based on OL "pressure" stats (even using sacks would be a start), that might be better than using such a subjective stat as Adjusted Line Yards.

Anyway, analysis-wise the only thing I'll point out is you probably can't just add the effect of any OL stat with the effect of Gase's presence. That would suggest there's no interaction effect, which probably in this case means you have to subtract something from that expected win total.

In summary, I think the attempt is nice but I can't give it any credence because of the stats used as input to the analysis (even DVOA, while the principle behind it is fine, is shaky depending on implementation which they conveniently don't show).

 

You are correct that ALY is a measure of run-blocking. However, it is not a subjective metric.

Either way, you could use a subjective pass-blocking metric and get similar results.

 

How does an improved defense that doesn't give up tons of points affect the expected win/loss total?

 

It's certainly subjective based on their description. They are saying they can separate the effect a RB has on a play vs. the OL. That's complete bull**** given that the same RB and OL almost always play together and you're only using common stats (so no controlled experiments).

Anyway if you're going to argue it's not subjective, show me the formula and the justification behind it (the first part you probably don't have and the second part is where it becomes subjective).

 

How about the QB just make more plays. That would increase the win total too.

 

lol.

 

ALY is the offensive line statistic most strongly correlated with quarterbacks' DVOA. From there it drops very sharpy, with percentage of pressured pass dropbacks, for example, having but a mere -0.19 correlation with quarterbacks' DVOA.

So if you choose something else to measure offensive lines, obviously the sorts of adjustments I did above with regard to Ryan Tannehill would be far more conservative. For example, if the Dolphins had even the best offensive line in the league in terms of percentage of pressured pass dropbacks, it would do something virtually negligible to Ryan Tannehill's DVOA.

In fact, you could make the argument that ALY, based on its 0.62 correlation with quarterbacks' DVOA, represents the effect of the quarterback on the offensive line, rather than vice-versa, given that all of the other measures of offensive line play have far weaker correlations with quarterbacks' DVOA. ALY has little if any concurrent validity as a measure of offensive line play, in other words.

 

None of this makes it subjective. Its calculating probabilities.

 

Poster Laurence on fire. Where did this guy come from???

 

C'mon....really String? You really don't know?

 

Of course it's subjective. First of all, assigning numbers to something doesn't make something objective. If I say 72% of the yards gained by a RB are due to the RB and not the OL that's quantifying something but still subjective. Second, there's an entire area of probability theory called Bayesian inference where you update probabilities assigned to different hypotheses, but the initial ones you assign are completely subjective, and thus all further ones are too (only the updating method is objective). Third, in this specific case they are saying you could use the variance in yards gained to somehow (they obviously don't specify how) determine how much the OL contributed. Like I said, they don't show their work and there's no principled way of doing that anyway.

Yes it's subjective and the stat is bogus.

 

What numbers are being assigned? You clearly don't understand what FO does. They are not assigning any numbers to anything.

 

No idea. But he has had the best posts in this thread. Can't wait to see what else he has.

 

Of course they are. Read the link I gave you. And you just got through saying they're calculating probabilities (which are numbers if you didn't realize already) and you now you say they aren't assigning numbers to anything??

And yeah the problem is no one except them understands what they do. Why don't you clarify what they are doing? Right.. because you have no clue either. The stat's bogus.

 

Everything is derived from play by play data. Again, what number did they assign? Everything that goes into the DVOA model is objective play by play data. It is no more subjective than passer rating.

I understand perfectly what they do. It is exactly the same statistical techniques they use for DVOA. I'm not going to write out pages of python code to illustrate how DVOA works. Thousands of other people understand what they do, which is why they're so highly regarded in the analytics community and tons of work has been built off of what you claim is bogus.

 

I'm asking you to explain the math. That shouldn't take more than a paragraph. I can tell it's bogus myself so I know I'm right here. And computer code means nothing. You can assign tons of subjective weights to whatever the hell you want in them.

Explain the math or this conversation is worthless (it already is actually).

 

The fact that you believe weights are subjectively assigned shows that you do not understand. We'll leave it at that - you don't understand exactly how DVOA works, so you won't accept it as evidence of anything. Nothing wrong with that.

 

Right you can't explain the math. This conversation has shown me more about you than your entire posting history while I've been here.

 

I certainly can. I'm not paid enough to make a post on it.

Tell you what - lets put up $100 that says I can't explain the math. Money is paid to thephins.com.

 

Bull****. The time it takes to explain the math is measured in minutes. The fact you can't do it tells me everything. You can't even tell me which statistical techniques give you the ability to take the variance in yards gained by the RB and somehow derive how much is due to the OL. Answering that takes seconds.

It's simple dude.. the explanation they gave leads to subjective weights. You claim otherwise, show the math!

 

I think DVOA is interesting, and that they do a better job than PFF, but at the same time it isn't a set in stone definitive, and I think that's what Brad is saying.

 

Put your money where your mouth is.

 

Show us the math.

 

He's saying its subjective. I'm saying its objective. I'm not saying anything about DVOA's validity or utility. The result of the coin flip is objective, but it has very little utility when assessing performance.

 

Let's do it for charity. Make a Wish Foundation.

 

Stringer Bell doesn't understand this, but the only way you're getting "objective" analysis from DVOA or so is if the weights they get are derived from some kind of parameter optimization (you get a computer to run through a whole series of possible weights and choose the ones that help you predict something best) so that you can predict win % for example.

That would be objective. If they ever did that they'd say it front and center. Anyway, in science or math (and I've made discoveries in applied math and created models in neuroscience) you'll never see someone act like he does. They show you their work up front and it becomes a question of where the errors are, if any.

Anyway, I'm done with this. Until he shows the math it's a waste of time.

 

Thanks for clarifying how they are objective. I'd urge you (and everyone else) to donate $100 to the Make a Wish Foundation.