Just how important is "clutch", really?

When hall of famers play their best the games are typically in the 15+ category and when they play their worst the games are typically decided by 7 or less. There is a very clear pattern. On the other hand when Tannehill plays his best often the games are close and when he plays his worst often the games are blowouts (for them). So I don't agree with the conclusion that if he played at a hall of fame level he'd have an extra 0.5 wins a season, maybe the 7 points or less stats might look similar or slightly better, but the other splits would look completely different.

 

Pauly! Admire all the work you put into this!

Assuming the data is correct, I think your data shows 4th quarter passer rating and passer rating while trailing with <4 minutes left are important "clutch" stats because a correlation of around 0.4 is pretty good (it's similar to correlation of YPA with wins). Coupled with the 55% win percentage for close games and near zero correlation between passer rating and wins in those situations, it basically shows what matters (re: the QB) for winning close games is what the QB does near the end of the game, and not as much what he does during the rest of the game.

That all jibes with intuition so that's nice.

Regarding the stats themselves, I'd point out that analyzing just based on "best" and "worst" teams often skews correlations and many other stats, so I'd include all the teams you have. Also, the last list is missing Dalton and the win percentages are the ones from the list above that so you'll need to edit those. And just making sure.. when you "adjust" passer ratings from previous years to 2015 averages, that needs to be done by multiplication, not addition, or it will underestimate the variance.

 

I need a grant before I read this thread...

 

If I had two hills, both at the same height, incline and built of the same elements, then at the top of those hills I placed the exact same boulders and set them rolling down the hill towards the exact same 2 barriers, with the only difference being one barrier's structural integrity was severely weakened with holes....would it be reasonable to expect the compromised barrier to breakdown faster then the intact barrier?

I'm asking this, because I've gotten into numerous arguments with cbrad in the past about how the oline can adversely affect the QB's performance towards the end of the game when behind. Common football sense tells us, when behind in the 4th teams are going to pass, because time is of the essence and passing takes less time than running due to clock stoppage rules. If a defense knows the other team is always going to pass, then they'll have an easier time defending because they know they don't have to worry about the run. Having that knowledge, let's them take more risks and go after the QB harder. Teams with better olines or QBs with elite level elusiveness can counter that.

Its not just possible that our line and play calling are the reasons for Thill's numbers, its likely.

 

and how do you apply the factor of the team believing their QB can lift them to victory? Marino I would say the team believed 100% Marino would somehow bring them back an win the game. With Tannehill ...not so much

 

or a degree

 

Nah, I could follow. Anything at or below calculus, no problem. It's the post joint minutia that gets too much in the way of Panda's feelings...What's that you say, star crested unicorn in lotus position? I agree, I'm already rolling it...

 

Fun read....

Just watching games can tell you Ryan is subpar on 3rd downs and 4th quarters. IF he is to ever be a reliable NFL QB, this must change and fast.


Sent from my iPhone using Tapatalk

 

Yeah, I never got the feeling you understood my argument. I'll just state it here for others again, but I really don't want to get into a back-and-forth over what I feel is you not understanding my argument.

Anyway, long time ago, I pointed out that the drop-off from Tannehill's <4 minute or <2 minute tied to trailing stats are abnormal relative to the rest of the league. Importantly, I'm talking about the drop-off, not the absolute passer ratings in either condition. And no, he's not the only QB with such a huge drop-off but his drop-off is rare.

If we just take the list of QB's Pauly used and compare career <4 minute ratings (it's even worse for Tannehill with <2), you get an average <4 minute tied rating of 79.6 and an average <4 minute trailing rating of 74. Standard deviations are 21 for tied and 7.5 for trailing.

Tannehill's <4 minute tied rating is 97.5 and his trailing rating is 63.3 for a difference of 34.2. So with a crappy OL throughout the entire game, he performs well above average while tied and way worse below average when trailing. This is important because the OL is crappy ALL the time.

Now, for any OL, you should see some drop-off when defensive pressure is higher, for example from <4 minute tied to trailing situations. That average drop-off is 5.6 passer rating points with a standard deviation of 22.7. Tannehill's drop-off is 34.2 or almost 1.5 standard deviations below average. That's what I was alluding to.

You can't argue the OL was good while tied but crappy while trailing. It was crappy throughout. So all the other factors (Tannehill included) were more than compensating for the crappy OL while tied with <4 minutes left. Those other factors could not compensate while trailing. Point is, once you start off with the assumption the OL was crappy the entire game, you can't explain the drop-off with the OL. It HAS to come from one of the other factors, and since passer rating is so dominated by the QB, a large portion of this drop-off has to be due to Tannehill.

Anyway, like I said I'm not getting into another back-and-forth if it's clear my argument isn't being understood. Either way, only one other starting QB has such an abnormal drop-off: Russell Wilson, so you can use this argument for him too. But no one else comes anywhere close.

 

Well, cbrad, I think we do understand your argument. But, when tied or ahead, there's at least the chance of a run (although Lazor did a pretty good job at taking that thought away with his abandonment of the run). I think the disconnect is that you attribute the drop off solely to Tannehill, and we attribute much of it to what FinD said: that is, late in the game, when trailing, the defense tees off even more on our porous oline. That, and in clutch situations, like late in games, it was common to see holding our false starts by the oline (a sure sign that that know they're getting worked and they're trying to get an advantage), or various other penalties by players, consistently putting us in unmanageable situations on third and fourth down. I remember in several games last season where commentators were talking about how often the Dolphins offense was in 3rd and 7 or more to go. When the offense is consistently in 3rd and long, and you have receivers running routes short of the sticks (well documented on here last season by CK and others), I find it incredibly disingenuous to blame the QB for a lackluster 3rd down percentage.

 

I never put all the blame on Tannehill. I'm saying there's no way Tannehill isn't a major reason why you see that drop-off if you assume the OL is crappy throughout. And the extra defensive pressure is a condition all teams in that situation face, so that's not specific to the Dolphins. Also, lack of rushing is hard to use as an excuse when we're usually middle of the pack in rush offense.

And just mathematically speaking, if the OL is already crappy, then there's less room for it to get even crappier (that is, if you are operating at a high level, you can fall much further than if you are operating at a low level already). So even without a good independent measure of OL ability, odds are it's not the OL. Is it coaching? Maybe, but this is career ratings and over multiple coaches. Point is, it's hard to not assign a good portion of the blame to Tannehill.

 

I just don't think that from the standpoint of QB ability, that 3rd down in the 4th quarter, some QBs just forget how to play. I think it far more likely that other factors, oline, poor receivers, poor coaching, get exploited more late in games.

Again, we get into the oline badness, and, as we've discussed and agreed, the stats around pressures are far too limited. We all saw teams rush 3 or four linemen last year, and get pressure from multiple angles. Now, let's say it's the fourth, it's 3rd and 15 after an oline false start. Defense doesn't have to bring a blitz, cause they've been getting quick pressure with three or four all game. They bring three. Pressure comes from both sides, and all the short of the sticks routes (all the routes we ran, seemingly, lol) are covered. But that's on Tannehill?

Look, I'm not saying there weren't times that Tannehill didn't make a mistake, but to continue to act like the oline hasn't been a massive issue on third downs and in the fourth is just crazy. Everything gets exacerbated in the fourth quarter, so, I don't agree with the premise that the oline is equally crappy in the fourth as they are in the first. Common sense tells you that the oline should be at its best early in the game, when they're fresh and haven't been getting manhandled for 3 1/2 quarters.

 

Wait a second.. I'm talking about <4 minutes trailing stats, not 3rd down stats. I can help you out with the 3rd down inefficiency actually. Check these stats out:
https://imgur.com/a/jlWmo

Dolphins (last year) had on average more to go on 3rd down than all other teams.

But there's no arguing the <4 minute trailing argument I gave. In fact, that argument is so strong statistically speaking that almost every other stat you'd use to argue Tannehill is a good or improving QB is far more likely due to chance than the one I put up. We're talking a huge effect size here (effect size = difference in means divided by standard deviation). For example, Tannehill's passer rating improvement over his first 3 years I'd argue is good statistical evidence he improved, but from a purely statistical point of view that's far more likely due to chance than this <4 minute trailing argument.

Point is, it simply strains credibility to dismiss this <4 minute (or even worse the <2 minute) trailing argument by saying the QB was not a major factor. resnor.. just accept it. It's not an indication Tannehill is bad in general or won't make it. It just says he very likely performed worse than an average QB would with the same tools (including OL) in those pressure situations. Accept it and move on

 

That's for a non-statistics thread honestly. That's unquantifiable and I'd say unlikely to even be a factor. NFL players are motivated by personal success as well. The idea that someone is going to potentially lower their future compensation because they don't think a guy is going to throw a TD is sort of ridiculous. I had a boss who I didn't think was the best person for the job. I actually worked harder to make myself more attractive to other employers/departments so I could get out of there faster. Add to that, we KNOW why our O-line has sucked, no need to speculate. Dallas Thomas is trash, Billy Turner isn't that good, Albert was hurt, Pouncey was generally Pouncey, and James was pretty much James.

Now, I think O-line issues are overstated relative to QB play considering a guy like Wilson has had a pretty weak line over his career and the Patriots had a terrible line last season and still made the AFCCG. You want to see Tannehill overcome some of these issues, much like other top flight QB's are able to. There are a lot of variables at play here and each side just focuses on the variables that allow them to blame who they want to blame.

 

I understand your argument. I always have. I don't think you understand my argument actually.

I'm not saying the line gets worse. I'm saying the line is always bad...but the opposing defense knows what's coming in those situations which is a pass. When the defense knows what's coming its easier to defend.

Say you were playing a game...that started with a coin flip. If it landed on heads you have 20 different options and if it lands on tails you have another 20 options of what to do. Now, imagine your opponent has a double headed coin....doesn't your job become easier? Of course it does. The defense knows we're gonna pass when behind in the fourth. So they don't have to hesitate or guess, they can come at our porous line at full speed.

So back to the hill analogy...if the defense is rolling down the hill at full speed, which barrier are you going to stand behind, the regular one or the compromised one?

 

I agree, Tannehill needs to be able to overcome, but he needs some tools. The main one, I believe, is coaching. I think the coaching during his first four seasons was bad and detrimental.

Wilson can be successful with a terrible oline, right now, because he has Barry Sanders athleticism. Tannehill doesn't. Brady had receivers who would routinely turn 4 or 5 yards passes into 10-15 yard gains. He he also has stellar coaching. Tannehill had neither of those.

So, yes, I want to see Tannehill do better, but I think that a big party of that is getting better scheme and coaching. I think that's happening.

 

Yeah, I went with 3rd down, as we often get that lumped in. Lol. However, 3rd down is often encountered late in games, so dismissing it when talking about 4 min left, or 2 min left, is silly.

 

While I applaud the extensive stat analysis, IMO there is no combination of stats that will correlate how clutch a QBs play is. Especially the stats in "close" games. That's a meaningless stat. You have 43 other players on the field, plus special teams, plus coaching decisions that all play factors into winning close games, or better yet, how close those games end up being when the clock expires.

 

Joe Montana had a higher playoff rating in the playoffs. 95, which is remarkable considering the era he played in.

Peyton Manning's two Super Bowl wins were pathetic performances by him, and FANTASTIC performances by his defense (Indy led by Bob Sanders, Denver last year). He was pretty good when they lost to Seattle.

If there is no such thing as clutch, there is no such thing as choking.

 

You have to show the increased level of predictability in those trailing situations was: 1) abnormally large for the Dolphins, and 2) that it explains most of the observed drop-off (i.e some mathematical relationship between the "predictability" stat and wins/points/rating, etc...).

In <4 minute trailing situations, the defensive pressure will be greater in general and the predictability of the offense will be greater in general. So it's not enough to just say these are possibilities, you have to provide the evidence something unique was going on with the Dolphins. I'm just basing my conclusions off the data we currently have.

Here the same thing. You have to show the increase (if there was one) in 3rd down distance in trailing situations for the Dolphins was abnormally large compared to other teams. Otherwise, yes the 3rd down stat can be dismissed for a <4 tied vs. trailing comparison.

 

I appreciate the work Pauly!

Personally, I believe that a QB's passer rating is based on the play of several players rather than just the QB. To wit, two QBs with identical skills would produce different passer ratings if one had a superior pass blocking OL, Calvin Johnson at WR or such a strong running game that the defense couldn't assume pass. Very few QBs can overcome team deficiencies regularly.

 

Really, man? We already know that their 3rd down distance was longer than almost every other team. But I suppose that changed magically in the fourth quarter.

 

This is what I mean by not understanding my argument. You'd never say what you just said if you realized once again I'm talking about drop-offs in performance, not absolute values. What else can I say..

 

I didn't say the the predictability was abnormally large for the Dolphins. I said the predictability allows for the defense to tee off on a horrible oline.

Also, if we're talking about predictability, then understand we ran the ball the fewest of any team last year. Certainly, i don't need to explain that passing so much in the first 3/4 of the game makes it easier for the defense to recognize a given pass play when they're lined up in the 4th.

 

That's precisely the point. You are NOT providing evidence the predictability was abnormally large for the Dolphins. All teams will have an increase in predictability in those situations. Yeah.. same response as with resnor. I just don't think you guys understand the argument.

Anyway, debate was pleasant! Till the next one.

 

Which kinda of leads me to my question.

Because football success is so predicated on multiple variables.... I almost think you need to look at whether or not a QB is clutch.. And how many other players you would define as clutch around him.

For example... Dan Marino is certainly clutch.. But also how many other clutch players were also on the field at the same time.

Look at Ryan... Not someone I would call clutch.. But right now.. Really only Jarvis Landry is someone I'd call clutch on the offense.

 

Something else... On the radio this AM they were talking about clutch in the NBA is 5 mins left and score within two possessions.

I wonder if there is something like that in football.

 

I'm not providing it because I didn't say it or imply it.

I do understand your argument. Its just wrong.

You aren't accounting for differences in oline ability. You want to people to show X is worse for the Dolphins, but you make no allowances for our oline. You make no allowances for our coaching.

 

Say everybody in the world played the flute, and everybody in the world lost one finger. Everyone's ability to play the flute would diminish, by some non-random factor. If, on average, everyone is expected to see a 15% drop in that ability, then people at or around that percentage drop can pretty much be excused. The Miami Flute Orchestra has a first chair whose ability diminishes by 45% without the finger. How good he was didn't matter. If he was really good or really bad, he still should have only seen a ~15% decrease in playing ability. cbrad has already given credence to the O-line and offensive predictability. What he's saying is that compared to his peers, Tannehill's decrease in play is WAY MORE than what can statistically be attributed to that O-Line decrease.

A qualitative analogy: Sun in your eye should account for slowed driving and maybe a swerve or two. You're trying to attribute 32 counts of vehicular manslaughter to having sun in your eye.

 

I understand his argument. However, every QBs oline isn't the same level. Did the QB with the best oline see the same dropoff? Tannehill's dropoff is extreme, but his oline is terrible. When your looking at dropoff, and comparing QBs to each other, you're now taking the oline out of the equation, and attributing the drop to the QB.

 

First, I want to say that I appreciate the effort you put into all this. That said, while you can define "clutch" however you want for this purpose, I don't think "getting better than 50-50 results in 50-50 situations" is a good definition of "clutch." Of course there will be some that get better than 50-50 results because you one can't expect everyone to get exactly 50-50 results with any significant number of participants. With any normal distribution, some will do a little better than 50-50 and some a little worse. If the sample size is big enough, I'd expect that range to be somewhere between 45-55%, which seems to be pretty much what you found. Just as if you had 100 coin flippers flipping 1000 coins you would see some get better than 50% heads and some worse, the same would be true in the situation you are talking about. Not to mention that the QB doesn't control all, or even most, of the variables that go into a win or loss. So I don't think the fact that some may win 55% of their 50-50 situations proves that anyone is clutch.

Personally, I think "clutch" refers to someone who elevates his performance in "clutch" situations in a material and statistically significant way. Most studies of this nature that I have seen have concluded that this form of "clutch" doesn't really exist. When you look at sports actions with few variables and that are easily measured over large samples sizes, you don't really see players deviate from their norms in a major way. Take, for example, free throw shooting in basketball. If there was such thing as clutch, you'd expect to see players who over extended periods of time are much better free throw shooters in the playoffs or in late game situations. But we don't see that. Michael Jordan was the same FT shooter (actually very slightly worse) in the playoffs. Larry Bird was basically the same FT shooter in the playoffs (very slightly better). Etc. Some small variation is expected. Playing QB has many, many more variables than FT shooting. Indeed, late game QB play is in some ways a very different game than the rest of the game. The QB is often either in a hurry-up offense with 4-5 WRs and a defense in prevent, or trying to wind the clock down. So even if you just look at the QB's own performance in late and close situations it may just show that the QB is very good with 4-5 WRs against prevent defenses, as opposed to him really being "clutch."

 

5 minutes left? That's an eternity in the NBA. One reason I don't like basketball so much is because for so many games you can just tune in with 2 minutes left and you haven't missed much.

I really wish they'd get more creative and have another basket way higher than the normal one where you could get 4 or 5 points, or at least have not just 3 point lines, but 4 and 5 etc.. or do something to prevent that incessant fouling at the end. I don't know.. basketball just isn't as interesting for me as football.

 

As cbrad explained, there are two outliers. Tannehill and Wilson. Outside of that, everyone was within the band of values that can be deemed 'understandable'. In this case, understandable means owing to the factors that vary for all QB's. Obviously situations are not 1:1 and better or worse situational coaching and O-line technique, as well as random variation (e.g. slipping on turf, rainy weather causing drops), can cause varying degrees of drop-off. That's where you see people within a standard deviation above or below the average value. Generally, though, QB's are clustered around 5.6 rating point decreases. There are some guys who are lower than that, some who are higher, and that's the nature of averages with high standard deviations. The proble is that Tannehill drops off by 34.2 rating points. He's 28.6 points worse than average in that situation and far surpasses what you can reasonably attribute to the pratfalls of predictability and a more aggressive defense.

 

Yeah, what you're saying is true for a single randomly chosen QB, but the real question is how likely it is for many many HoF QB's to have an average of 55% win percentage. The type of math you'd use here is something called signal detection theory. It basically quantifies the overlap between two distributions: distribution (in this case) for HoF QB's vs. distribution for non-HoF QB's.

The overlap is quantified by calculating something called d' (d prime), which is just the difference in the means of the two distributions divided by a weighted standard deviation (of the two standard deviations for the two distributions). d' can be translated into the probability that a randomly picked HoF QB will outperform a randomly picked non-HoF QB. That probability (something called area under the ROC curve) is what we want.

Just for info

 

IMO outliers are most likely caused by factors other than the QBs themselves. In every long-term analysis of pro athletes I've seen, the vast majority tend to perform roughly the same in "clutch" situations as non-clutch given a large enough sample for each. Muscle memory is just too ingrained at that level so they perform according to their training/skill level. "Clutch" just isn't a factor at that level. Therefore horrifically poor OL play and/or play calling are more likely factors for outliers. For example, the Dolphin defense with ZT and JT were notorious for late game melt downs. I don't believe that these guys weren't "clutch" or played worse at those times. They simply played in very predictable defenses and that gave the offenses a big advantage. The same would happen for a QB playing in a very predictable offense. The D could just over-play the expected play and the offenses' odds of success would go way down.