Tannehill

[fyo]

I think you're deliberately ignoring the parts of this problem that make it hard.

Let's assume, for the sake of argument, that you are a team with a horrible offensive line. As a coach, what do you do? You do *everything* to help the o-line, cover up their weaknesses, limit the ways they expose both the running plays and the passing plays. This includes changing the play book completely, using more blocking TEs, keeping running backs in to block, etc. In other words, the coach pretty much does everything he can to wreck any correlation we would find using the metrics we are using. Maybe coaches actually do something?

You know me though, I'm going to rail at anything observational that doesn't allow for the systematic comparison of all the teams in the league, so I'll stick with your Adjusted Line Yards and contribute the fact that if Tannehill would've been afforded the league's best ALY by his offensive line in 2015, statistically speaking, his QBR would've most likely been 60.8 instead of 43.1.

Tannehill's QBR was just about that high last year...

The problem of course is that Tannehill is starting so low in the league in this area that it would take an uncommonly strong correlation between QB play and offensive line play, as well as a highly unlikely degree of improvement in the Dolphins' offensive line, to get him where he needs to be for the team to be highly competitive.

Why are you using solely this year as the baseline for Tannehill?

On the topic of coaching and play calling , which seems especially relevant with a new set of coaches taking over:

One of the things that has frustrated me as a fan of the Dolphins is that they just don't seem to call the plays that the offense has had success with. Every season it seems like a decision is made NOT to use any of the things that worked the year before and instead try something else or a bunch of stuff that didn't work. That always struck me as dumb, to put it mildly. If the detailed game charting stats were available, I'd love to run some math on them and see what the correlation (if any) is between overall team success (or offensive success) and keeping what you are good at doing.

[Dolfanalyst]

I think you're deliberately ignoring the parts of this problem that make it hard.

Let's assume, for the sake of argument, that you are a team with a horrible offensive line. As a coach, what do you do? You do *everything* to help the o-line, cover up their weaknesses, limit the ways they expose both the running plays and the passing plays. This includes changing the play book completely, using more blocking TEs, keeping running backs in to block, etc. In other words, the coach pretty much does everything he can to wreck any correlation we would find using the metrics we are using. Maybe coaches actually do something?

Does any of that compensation for a poor offensive line consist of things the quarterback might be called on to do, with whatever skills he possesses that can come to the aid of the offense?  Of course that would diminish the correlation, as well.

We do have the finding I pointed out earlier in the thread, where the year-to-year variation between better and worse quarterbacks is significantly greater than the variation within the two groups.  That alone suggests that quarterbacks' individual performances are able to withstand these sorts of deficiencies elsewhere on offenses, and that coaches with good quarterbacks would be wise to use them to compensate for such deficiencies.

Why are you using solely this year as the baseline for Tannehill?

Just because it's where the thread seemed to be when I entered it.  We can actually fart with any of these years, or with his whole career.

[Dolfanalyst]

This is sure interesting:

Tannehill needs a great tutor, counselor, mentor and guru because he has been playing his first four years under a coach who didn’t believe in him.

That’s right. Joe Philbin didn’t believe in Ryan Tannehill.

Everyone high enough in the Dolphins organization to know admits it now. I’m told Tannehill figured it out as well before this season began. The Dolphins don’t dismiss the idea that it played a factor in the quarterback’s performance.

So Gase, who has the blessing of Peyton Manning and Jay Cutler, the past two quarterbacks with whom he has worked, must hone and craft Tannehill into something better than he has been.

The Dolphins are confident it will happen.

“We’re convinced,” a source said, “you’ll see a different Ryan Tannehill next year. That’s how much Adam will affect things around here.”

That is the reason Adam Gase is here.

Read more here: http://www.miamiherald.com/sports/spt-columns-blogs/armando-salguero/article53964610.html#storylink=cpy

[CF DolFan]

This is sure interesting:

Read more here: http://www.miamiherald.com/sports/spt-columns-blogs/armando-salguero/article53964610.html#storylink=cpy

LOL ... Not much of an endorsement. Philbin's team that was predicted to challenge for the division quit on him and he was fired 4 games into the season and replaced by ... get this ... the tight ends coach. A guy who not only hadn't coached before but had never even been a coordinator or coached more than 4 players at a time!

[Dolfanalyst]

From another thread:

Just out of curiosity, do you have any statistics to back that up?

Pro Football Focus has Tannehill as the third best quarterback in the league under pressure (for 2015, anyway):

https://www.profootballfocus.com/blog/2015/05/20/qbs-in-focus-under-pressure/

I found a different page on PFF where they break down the 2014 numbers. Here they have Tannehill as the fourth best on scrambles and third best on rollouts.

Basically, I'm struggling to find any statistic that substantiates your claim. Looks like Tannehill is almost elite on those plays, just going by PFF. (That's not necessarily my own impression, but whatever).

Actually there is something related:

http://espn.go.com/nfl/qbr/_/sort/cwepaSackedCondensed/order/false

Clutch-weighted expected points added (or in this case lost) on sacks.  League-wide it had a -0.19 correlation with the percentage of pressured pass dropbacks experienced by quarterbacks in 2015, and a -0.14 correlation with Adjusted Line Yards (Football Outsiders).

So the extremity of Tannehill's performance in the league in this area is nowhere near explained by either 1) the percentage of pass dropbacks in which he was pressured, or 2) a measure of offensive line play that correlates fairly strongly with quarterbacks' DVOA.

The most parsimonious explanation is that Tannehill, in comparison to the league's other QBs, takes more sacks on "clutch" plays.  Again, this is the opposite of what we saw the other day, when both Rodgers and Palmer experienced a great deal of pressure on clutch plays, but nonetheless evaded that pressure with movement and hit receivers downfield for very impactful plays in the game.

[Spider-Dan]

Bottom-ten clutch points on sacks, from your link:

1   Blake Bortles, JAX
2   Ryan Tannehill, MIA
3   Teddy Bridgewater, MIN
4   Tom Brady, NE
4   Philip Rivers, SD
6   Jay Cutler, CHI
7   Aaron Rodgers, GB
8   Russell Wilson, SEA
9   Matthew Stafford, DET
10   Matt Ryan, ATL

And bottom-ten in total sacks taken, from here:

1   Blake Bortles      51
2   Aaron Rodgers      46
3   Alex Smith      45
3   Ryan Tannehill      45
3   Russell Wilson      45
6   Matthew Stafford   44
6   Teddy Bridgewater   44
8   Philip Rivers      40
9   Tom Brady      38
9   Marcus Mariota      38

Seems to me that generally, the QBs that took the most sacks were the leaders in clutch-weighted EPA lost on sacks.  (This should be obvious by the inclusion of Rodgers, the QB you are specifically citing as clutch, as the 7th-worst QB in the league when it comes to "taking sacks in the clutch.")

There are some exceptions, of course: Alex Smith was t-3rd in sacks taken but only 14th in EPA lost, while Jay Cutler was 16th in sacks taken but 6th in EPA lost.  But for the most part, the lists pretty much match up.

If you're citing a statistic that has 2015 Tom Brady and Aaron Rodgers as the 4th and 7th worst QBs in the league (respectively), perhaps you should reconsider how useful this statistic is as a tool for evaluating QBs.

[Dolfanalyst]

Bottom-ten clutch points on sacks, from your link:

1   Blake Bortles, JAX
2   Ryan Tannehill, MIA
3   Teddy Bridgewater, MIN
4   Tom Brady, NE
4   Philip Rivers, SD
6   Jay Cutler, CHI
7   Aaron Rodgers, GB
8   Russell Wilson, SEA
9   Matthew Stafford, DET
10   Matt Ryan, ATL

And bottom-ten in total sacks taken, from here:

1   Blake Bortles      51
2   Aaron Rodgers      46
3   Alex Smith      45
3   Ryan Tannehill      45
3   Russell Wilson      45
6   Matthew Stafford   44
6   Teddy Bridgewater   44
8   Philip Rivers      40
9   Tom Brady      38
9   Marcus Mariota      38

Seems to me that generally, the QBs that took the most sacks were the leaders in clutch-weighted EPA lost on sacks.  (This should be obvious by the inclusion of Rodgers, the QB you are specifically citing as clutch, as the 7th-worst QB in the league when it comes to "taking sacks in the clutch.")

There are some exceptions, of course: Alex Smith was t-3rd in sacks taken but only 14th in EPA lost, while Jay Cutler was 16th in sacks taken but 6th in EPA lost.  But for the most part, the lists pretty much match up.

If you're citing a statistic that has 2015 Tom Brady and Aaron Rodgers as the 4th and 7th worst QBs in the league (respectively), perhaps you should reconsider how useful this statistic is as a tool for evaluating QBs.

It's a piece of the puzzle.  Alone, it isn't a good measure in that regard.  However, the difference with the league's best QBs, the inevitable exceptions to the rule aside, is that they make plays in the passing game (and/or with the run, i.e., Russell Wilson and Alex Smith) that compensate for what's lost on sacks.

For example, when you look here:

http://espn.go.com/nfl/qbr/_/sort/cwepaPassesCondensed

...you see that clutch-weighted expected points added on plays with pass attempts ("PASS EPA") illustrates that.

So, it's not only what happens with regard to sacks on clutch plays, but also what happens on clutch plays that don't result in sacks.  Tannehill struggled in both areas in 2015.  The two years prior, he did better with regard to the latter, but he's struggled with regard to the former from 2013 to 2015.

If you're going to take lots of sacks in clutch situations, you'd better also make plays in those situations, when you aren't taking sacks.  If Tannehill is capable of the former to a great degree, then he needs to also be capable of the latter to a great degree, if the Dolphins are to be successful in the sorts of games like the Arizona-Green Bay one the other day.

[Spider-Dan]

Again, on your newly cited statistic, Rodgers is 14th best, while Tannehill is 20th best.  And both of them are behind such luminaries as Kirk Cousins and Jay Cutler.

This is hardly convincing evidence of clutchness (or lack thereof).

[fyo]

"Clutch" isn't the piece of ANY puzzle, for a variety of reasons. Overall, statistics based on it don't correlate well with future performance in the same metric, they don't correlate well with wins or points -- and throwing out a quarterbacks performance 90% of the time in a sport where there is almost always too few data points to begin with makes zero sense.

A quarterback has something like 500 passing attempts in a season. The number of data points necessary in order to be reasonably confident about much of anything depends on the width of the distribution of  the variable you're talking about, but "several hundred" is a good starting point. The 500 attempts already contain data from many, many different game situations and are quite noisy due to external factors (receivers dropping the ball, weather, whatever).

Looking for a quarterback's effectiveness in the clutch is pretty much futile if you are basing it on statistics. There just aren't enough data points.

[Dolfanalyst]

"Clutch" isn't the piece of ANY puzzle, for a variety of reasons. Overall, statistics based on it don't correlate well with future performance in the same metric, they don't correlate well with wins or points -- and throwing out a quarterbacks performance 90% of the time in a sport where there is almost always too few data points to begin with makes zero sense.

A quarterback has something like 500 passing attempts in a season. The number of data points necessary in order to be reasonably confident about much of anything depends on the width of the distribution of  the variable you're talking about, but "several hundred" is a good starting point. The 500 attempts already contain data from many, many different game situations and are quite noisy due to external factors (receivers dropping the ball, weather, whatever).

Looking for a quarterback's effectiveness in the clutch is pretty much futile if you are basing it on statistics. There just aren't enough data points.

"Clutch" in this context pertains to how much probability of winning is associated with the play.  There are far more plays in that regard than one might expect from only a superficial analysis.

For example, a play (a touchdown, a long gain down the field, etc.) that contributes largely to putting a team up by three scores late in the third quarter might be "clutch," in that it makes the other team very improbable to win at that point in the game (i.e., "the nail in the coffin").  Likewise, avoiding a sack and throwing the ball away, keeping one's team in field goal range to go up by three scores in the same situation, for example, might be just as "clutch."

In other words, we're not talking just about the hail-mary touchdown pass with no time left on the clock here, although that's obviously one well-known example of a "clutch" play.

[Dolfanalyst]

Again, on your newly cited statistic, Rodgers is 14th best, while Tannehill is 20th best.  And both of them are behind such luminaries as Kirk Cousins and Jay Cutler.

This is hardly convincing evidence of clutchness (or lack thereof).

In effect what you're saying here is that the gold standard for determining the validity of a statistic is "Spider-Dan's" personal assessment of how good quarterbacks are in that particular area.

In other words, we're starting with "Spider-Dan's" personal assessment, and the statistic is valid if and only if it jibes with it.  If it doesn't jibe with it, it's not that "Spider-Dan" is wrong, it's that the statistic must be wrong.

[Spider-Dan]

I'm not even using my own personal assessment; you're the one going out of your way to cite Rodgers and Palmer as Pillars of Clutchness compared to Tannehill.

Rodgers does poorly on the same metrics you're using to bash Tannehill.
Palmer almost single-handedly lost the game for the Cardinals multiple times in the 4th quarter of that game.

These are your preferred examples, not mine.

[Dolfanalyst]

I'm not even using my own personal assessment; you're the one going out of your way to cite Rodgers and Palmer as Pillars of Clutchness compared to Tannehill.

Rodgers does poorly on the same metrics you're using to bash Tannehill.
Palmer almost single-handedly lost the game for the Cardinals multiple times in the 4th quarter of that game.

These are your preferred examples, not mine.

No, I'm not.  I merely said those were examples of winning plays Tannehill has very, very rarely made.

[Dolfanalyst]

"Clutch" in this context pertains to how much probability of winning is associated with the play.  There are far more plays in that regard than one might expect from only a superficial analysis.

For example, a play (a touchdown, a long gain down the field, etc.) that contributes largely to putting a team up by three scores late in the third quarter might be "clutch," in that it makes the other team very improbable to win at that point in the game (i.e., "the nail in the coffin").  Likewise, avoiding a sack and throwing the ball away, keeping one's team in field goal range to go up by three scores in the same situation, for example, might be just as "clutch."

In other words, we're not talking just about the hail-mary touchdown pass with no time left on the clock here, although that's obviously one well-known example of a "clutch" play.

Just a further elaboration of this -- actually ALL plays are included in these statistics (the clutch-weighted ones), because all plays are associated with an increase, a decrease, or no change in a team's probability of winning.

This is how in fact the clutch "weight" is determined.  If a play is associated with no change in a team's probability of winning, then that play has zero weight.

So while the point made by "fyo" is a good one (i.e., how a small sample size would adversely affect the reliability of a statistic), that point in fact does not apply in this case.

In fact the beauty of the clutch-weighted statistics, on the other hand, is that they account for situational variables.  A garbage time touchdown pass -- one that does almost nothing to a team's probability of winning (or losing) -- is given far less weight than the plays during a two-minute drive when the game is tied, for example.

Last year the Packers played the Dolphins in Miami, and with 1:44 left in the game, the Dolphins were leading 24-20.  The Dolphins were 95% probable to win the game at that point.  From that point on, Aaron Rodgers engineered a drive culminating in a touchdown pass to Andrew Quarless with only 0:03 left to play.

If you look at the "win probability" graph on this page:

http://www.pro-football-reference.com/boxscores/201410120mia.htm

...you'll see that Rodgers was associated with the reversal of a 95% improbability to win (at the 1:44 mark) to virtually a 100% probability to win (at the 0:03 mark, after the touchdown pass to Quarless).

All of the positive plays involved in that drive, especially the touchdown pass, would be weighted heavily, and that's where a clutch-weighted statistic distinguishes itself from a traditional one.

[Spider-Dan]

So then, coming back from a 95% improbability to win is a good thing?

http://www.pro-football-reference.com/boxscores/201412210mia.htm - comeback from 94.70% improbability to win
http://www.pro-football-reference.com/boxscores/201309220mia.htm - comeback from 94.00% improbability to win
http://www.pro-football-reference.com/boxscores/201310310mia.htm - comeback from 97.90% improbability to win

The irony here is that you are citing Aaron Rodgers as your chosen example of clutch performance, when the knock on Rodgers for most of his career has been his lack of 4th quarter comebacks and poor record in close games.

Of the great QBs in the league, you picked the worst one to make your point with.