Here's What's Wrong With Cutler...Give Me A Minute...Wait I Know There Has To Be Something...

You don't know that. He's played the way he has due to the types of systems he's played under. He has all the athletic ability and smarts to thrive in the Pats system. That's why every QB that has played in that system has thrived. Some who were/are much less athletically gifted than Cutler.

 

I'm not sure that mentality changes to that extent.

And you are also discounting COUNTLESS great clutch throws Brady has made in his playoff career.

Not sure why he is so underestimated around here.


Sent from my iPhone using Tapatalk

 

Lol... Do you think Peyton is clutch in the playoffs?

(For the record, I don't believe in clutch)

 

What does Peyton Manning and his playoff performance have to do with the amazing throws Tom has made over the years?

You said Jay would win as many SB's as Brady, I asked why you are underestimating Brady's clutch throws. What u talkin?

And I'm ok with you not believing in clutch, I have no problem with people who think the earth is flat or that we live amongst aliens. I appreciate a free thinker.


Sent from my iPhone using Tapatalk

 

So would you say that based on the part of your post I boldfaced above, Dolphins fans on average would rate Tannehill as significantly worse than Rodgers, without more variance in their ratings than Packers fans ratings of Rodgers?

 

I think that's the "expectation" as a statistician would understand it, but as I said in the last paragraph of the post you quoted.. who knows about any particular case.

One thing to add though: as you reduce the number of rating categories, the percentage of items with low or non-existent variance increases. So if we just gave respondents a choice of 0 or 1 rather than 1-10 or so, then your original intuition is more likely to be accurate: lots of variance for Tannehill, very minimal for Rodgers.

 

You should play Poker.









Sent from my iPhone using Tapatalk

 

So if you were to use a rating scale of let's say 1 (very bad), 2 (average), or 3 (very good), and you asked the fans of the Dolphins to rate Tannehill, and the fans of the Packers to rate Rodgers, what would you expect to happen?

 

Personally for such a small rating scale I'd expect almost all 3's for Rodgers (out of 1000 people you'll probably still get the rare 2!) while for Tannehill I'd expect a normal distribution around 2.

 

But that's what I was getting at above, in that in that case, your kurtosis for Tannehill would be far greater than for Rodgers, and your skew would be far greater for Rodgers. You see what I'm getting at now?

 

Yeah I understood it the first time. But try the same thing with Tebow who for many people is an absolutely terrible QB and you'd probably get lots of 1's, 2's and maybe 3's too (based on that Tebow thread lol). So for Rodgers my intuition is that for small rating scales you're right, but in general?

More interesting question is which QB would almost everyone rate a 2? According to what I'm been saying there should be a few of those over NFL history.

 

But are you doing the same thing with Tebow among the fans of the teams he was on? If so I'd suspect he'd show low kurtosis and high positive skew. If you're doing it with everyone, irrespective of team following, then I think what you're saying above might be true, where we have lots more variation.

Here, however, we have only the fans of the Dolphins, and I suspect what we're dealing with is a relatively normal distribution with high kurtosis regarding Tannehill, using the "eye test" alone. Again that to me suggests we're talking about an average player.

 

Oh.. yeah that's something I didn't respond to before: if you actually wanted to do this study you'd want to make sure there aren't unknown between-group differences that could confound your results. That is, there could be biases in which portion of the rating scale different teams' fans use (e.g. due to how well their team is doing).

So you really want to use NFL fans in general and allow them to rate whichever QB's they want to rate as opposed to restricting the rating of a QB to just fans of that QB's team. They don't have to rate every QB though.. just enough other ones so that Rasch analysis will infer the "average thresholds" = average boundaries between rating categories so that all ratings are comparable.

 

Lol... Brady dinks and dunks. Of course out of his ~30 playoff games he's made a few great throws. He is an NFL QB after all. You asked earlier why he doesn't get much credit around here and it's because he cheats. Plain and simple.

Clutch is believing in magic. Flat Earthers and alien conspiracies are most likely the ones to believe in magic. Lol

 

But what I was getting at here originally was the value of the "eye test" in comparison to numerical data, in discussions about a player among hardcore fans of a team (like the people here). For that we'd need to restrict the analysis to those fans alone, and not expand it to fans of every team in the league.

Again if the hardcore fans of a team are producing a relatively normal distribution with high kurtosis (indicative of lots of disagreement among those fans) when using the eye test with regard to a player, we're likely talking about an average player, if the eye test among hardcore fans of a team is indeed a valid measurement of a player's ability.

If on the other hand we produced a distribution with high negative skew and low kurtosis, we'd likely be talking about an exceptionally good player, again if the eye test among hardcore fans is indeed valid.

Notice there is little or no disagreement among people here about how good Cameron Wake is. I suspect in his case we're dealing with high negative skew and low kurtosis in the eye tests of people here, in comparison to the normal distribution and high kurtosis seen in discussions about Tannehill.

 

I remember when I had my first beer.


Sent from my iPhone using Tapatalk

 

OK, that's fine. There are methods of analysis to deal with that too, such as the Partial Credit Model which essentially does Rasch analysis separately for different items, or Item Response Theory (IRT) which has something called an item-discrimination parameter that allows you to estimate item-dependent measures (QB measures) allowing for different variances for each QB.

The one thing you give up if you use those methods is that the scale is item-dependent. That is, estimated measures of QB's are no longer directly comparable. But so be it. Let's play in the world you're describing.

It turns out that the "expectation" within any group (fans) is still the same as I stated before, just that the raw ratings are no longer comparable across QB's. In other words, across NFL history this means we should see multiple examples of QB's that fans of that QB's team almost universally rate a 1 (very bad), almost universally rate a 2 (average), and almost universally rate a 3 (very good).

As stated, I'd agree you'll probably get Rodgers as almost totally 3's. Same with Brady, Marino, etc.. by fans of their teams. I'd also think you'd have guys like Ryan Leaf with almost only 1's.

So then back to the question I posed before: which QB's do you think fans of that QB's team would almost universally rate a 2? If you can find as many examples of those as for "very good" and "very bad", then there's your evidence that variance shouldn't be expected to change as a function of distance away from median rating, even if you decide to forego estimating QB ability on an invariant scale.

btw.. the utility of an "invariant scale" is that you can claim you are actually measuring something, so it's not something to lightly get rid of, but there are applications where people don't care about measurement per se, but instead care more about fitting a model to data as best as possible, and that's when IRT shines.

 

Lol... Irish Whiskey tonight my man!

Hey cbrad... do you believe in "clutch"?

 

Well, it's well known in psychology that for sufficiently difficult tasks you tend to get better up to a point as pressure to perform increases, but after that point performance declines as pressure increases. That's true for everyone. So in a general sense, "choking" under pressure or its opposite "clutch" can and is studied.

Problem with football is you can't directly measure the level of pressure on an athlete so it's hard to know where on that performance curve the athlete is. Best you can do is to just assume that pressure to perform is actually higher in one condition (e.g. playoffs) than another (e.g. regular season) and see how the player performs. Problem is that to do statistical tests you kind of have to assume the increase in pressure between the two situations is identical for everyone because you can't independently measure pressure.

But let's say you make that assumption. Then you can measure "clutch" by looking at the distribution of the decrement in performance for the player you're interested and compare it to that same distribution for all players historically (e.g. all starting QB's from regular season to playoffs).

Calculate the probability your QB's decrement stats would come from the league-wide distribution, and if that probability is less than 5% (that's the standard for hypothesis testing in statistics), then you have statistical evidence for "clutch" or "choking" beyond what random variation could explain.

Some players will definitely fit that description, so in that sense yes I think there's statistical evidence for it. However, keep in mind the problem is the assumption that the condition of regular season to playoffs represents the same (or approximately the same) increase in pressure.

 

I think in sports "clutch" is defined as playing better than normal under pressure situations.

Every study I've seen show that clutch is a myth in this sense. Basically, an athlete will play up to their level in all situations, on average, if all situations have a big enough sample. Statistically Brady plays much worse in the post season than he does in the regular season. However, that's probably due to the level of competition.

Even Reggie Jackson, "Mr October", never played better in October than he did in the regular season. It's just that people see things that they want and want to believe in magical things like "clutch".

 

Also, "choking" is not a myth. A player definitely can play worse than their norm under pressure, but a person can only play up to their max ability.

Brady CHOKED in their 18-1 season during the Super Bowl. Lol

 

One more thing, if a person defines "clutch" as someone who can play up to their ability under great pressure then ALL NFL players are "clutch".

 

Well Brady's definitely not going to fit the bill. That ~8 point decrement in passer rating from regular season to playoffs is right smack in the middle of the average decrement across the league.

Guys like Eli or Flacco would be more interesting to look at as they've actually performed better overall in the playoffs by many measures. I think I once did an analysis that showed that IF you assume Flacco was a different QB from 2010 onwards then that consecutive 10 game playoff stretch he had since then was statistically speaking significant relative to some decrement distribution I was looking at back then (can't remember exactly what it was).

Either way if you know of studies of "clutch" or lack thereof in football I'd be interested in looking at what they've done.

 

The studies I've read were mostly MLB related. I think I read one where they studied kickers, but I may be mistaken.

 

http://archive.advancedfootballanalytics.com/2013/01/clutch-persistence.html

 

Hey great find!

If this Brian Burke guy thought of comparing actual to expected win probability added kudos to him! However.. his use of correlation from year-to-year to won't determine anything other than consistency. You have to directly calculate the probability of obtaining that sequence of "clutch-WPA" from league-wide distribution of clutch-WPA. That is, he needs an actual hypothesis test in there, not some correlation, to determine how "clutch" something really was.

Oh yeah.. and one problem with the clutch-WPA metric is that WPA won't distinguish between "pressure" vs. "no pressure" situations like "playoff" vs. "regular season" or under which situations WPA was above expected (maybe the increase mostly occurred in non-pressure situations?). Nevertheless, that's a very good idea.

EDIT: Here's a suggestion for a better metric. Calculate (1-WP)*WPA at every point you have WP in a game. The first factor (1-WP) is a larger number (closer to 1) the less likely it is you will win the game = crude measure for greater pressure, while the second factor WPA is how much win probability you added.

Now if actual (1-WP)*WPA is greater than expected, that tells you the QB did better on average under greater pressure situations (more or less). You can look at that distribution league-wide and do hypothesis testing to see the probability that QB was more "clutch" than others.

 

What if Welker catches the ball, or Tyree drops that helmet catch?


Sent from my iPhone using Tapatalk

 

What the ..........

 

What if Brady threw a better pass to Welker?

What if the Pats defense didn't get 3 turnovers vs. the Rams scoring on 1?

What if Adam V. missed the 50 yard game winning FG?

What if Adam V. missed any of his 3 Super Bowl winning FG's?

 

It doesnt surprise me that you believe everything you're told. Lol

 

Football Outsiders paints a similar picture with respect to Cutlers best seasons relative to his peers. Team passing rank for Cutler led teams:

2007 9th
2008 7th
2009 23rd
2010 28th
2011 30th
2012 25th
2013 7th
2014 16th
2015 13th

Whats interesting is he seems to be a bit of an outlier in that most QBs that start off as hot as Cutler did go on to become the leagues best.

Roethlisberger
2004 6th
2005 6th


Brady
2001 10th
2002 10th


Rodgers
2008 10th
2009 9th


Wilson
2012 4th
2013 8th


Rivers
2006 3rd
2007 13th


P. Manning
1998 13th
1999 4th


Matt Ryan
2008 4th
2009 13th


Romo
2006 7th
2007 4th


Palmer
2004 12th
2005 5th


Warner
1999 1st
2000 3rd

Kaepernick would be the closest example of a hot start going off the rails as Cutler did. Looking at this measure there are 3 different versions of Cutler:
2007-2008 a hot young gun poised to be a future star
2009-2012 a total hack that had no business as a starting QB
2013-2016 an above average 2nd tier type

If he can stay healthy and Gase can somehow bring out what he showed he was capable of early in his career we may be pleasantly surprised.

 

To be very simple-minded, I assume one reason the Dolphins picked up Cutler was that they did not think Moore could get through the entire season in one piece. If they thought that, it wouldn't really matter how good a quarterback he was in a single game, right?

 

Oh look, danmarino in ANOTHER Brady debate. Shocker.

We are on to Cutler now. No need to tear down better QBs for the sake of RT for at least another year. Gase is a QB whisperer. Real games are almost here.

 

I suspect the real reason was that Cutler's skill set provides a much more complete use of Gase's playbook.

 

People make the mistake of confusing stats with QB evaluation. Stats are a result of team production. A QB like Cutler has some very good/even elite skills. But he also has some glaring faults. If you put him in a situation where he can maximize his skills and mask or minimize his faults then it's undeniable that he can be part of a very productive offense. That's what occurred the years the team passing rank was top 10. Add a good D and it could be a successful team.

That's also what people miss when looking at Tannehill. Tannehill also has some elite skills and fewer or at least less glaring faults. You put him in the right situation and likewise he will also be part of a productive offense. Last year once the offense settled in and they jettisoned those horrible Gs Tannehill was producing an elite rating (100+). The situation was far from perfect or even adequate, but it was also a vast improvement from the dumpster fire Tannehill has had around him his first four years.

 

After settling into Bill Lazor's offense and during a subsequent 12-game stretch in 2014 he produced at about the same level.

 

Oh look, jdang defending Brady again. Maybe you should see I mentioned more than Brady? Probably can't due to you having your head so far up his ***? Lol

 

We're not making that mistake. More importantly, stats allow you to quantify what portion of the variation in Cutler's passing stats are due to other stats, such as the running game. You can't do that using an "eyeball" approach.

For example, let's list Cutler's adjusted passer rating to Cutler's teams' adjusted rushing yards per game, from 2007-2015:

Adjusted passer rating (to the year 2016):
2007: 95.40
2008: 92.44
2009: 82.85
2010: 91.97
2011: 91.00
2012: 85.00
2013: 92.91
2014: 89.11
2015: 91.46

Adjusted (to year 2016) rushing YPG for Denver 2007-2008 and Chicago 2009-2015
2007: 120.09
2008: 109.33
2009: 87.09
2010: 96.10
2011: 117.05
2012: 115.66
2013: 110.28
2014: 88.14
2015: 115.98

The correlation between those is 0.4993. That means that the "variance explained" is the square of that = 0.2493. In other words approximately 25% of the variation in Cutler's adjusted passer rating can be explained by the running game, measured by YPG. That of course leaves 75% that's due to something else.

What about the defense? Ahh.. there you find something totally counter-intuitive. I'm just going to use defensive ranks by points allowed per game instead of going through the tedious calculation of adjusted points allowed for now:

Defensive rank for Cutler's teams from 2007-2015:
28, 30, 21, 4, 14, 3, 30, 31, 20

So just eyeballing it you can see Cutler did his worst with a good defense, and he did his best with a bad defense!! Correlation between the inverse of those rankings (i.e. 32 minus those rankings) and Cutler's adjusted passer ratings is -0.3629. Now this is where it gets interesting. Variance explained is 0.1317 or 13.17%, meaning you could predict 13.17% of the variation in Cutler's adjusted passer rating from looking his defense's ranking by points allowed.

But that makes no sense from a causal perspective (great example of correlation doesn't equal causation). What's going on there I have no clue, but one thing is clear: defense is NOT the reason Cutler did well in certain years and bad in others. In fact you have to give Cutler more credit than you'd otherwise give him. So however you look at it, most of whatever led to Cutler's passing stats has little to do with non-passing components of the team.

 

I would say the mistake people make is thinking stats have no place as a part of QB evaluation. Especially the people that will use stats when they support their argument then when they don't they be like: stats are for losers.

Stats and eye tests don't always match up. Thats OK.

"Predictive stats" aren't 100% infallible, as this particular measure was wrong on Cutler. Thats OK too, we're dealing with human beings after all. In Cutler's case he felt unwanted, that obviously had an effect on him. Another guy that started hot, Palmer, suffered a knee injury that derailed him a bit. Again, its the human element.

 

I think the human error also shows up when we see the media fawning over a player. People tend to jump on the bandwagon. I remember the press saying that Aikmen was the GOAT back in the 1990's. So many fans followed along with that and started claiming him as the best ever. I used to laugh at that assertion. And you'd be hard pressed to find anyone calling Aikmen the greatest nowadays. And the inverse is also true. The media claimed that RT was a bust more than once. And what happened? Most fans of other teams, and even some people here, followed along with that line of thinking.