Where does Tannehill rank among quarterbacks today?

He probably ranks 33 1/3rd

 

It means that in 2012 he did sending that he had never done before, and hasn't done since. Hence I called it abb aberration...an anomaly...something different from his norm. It's exactly what I've been saying this entire debate.

 

I think you will find most QBs passer rating drops in the playoffs, which makes sense seeing as the average playoff opponent is going to be much tougher than the average regular season opponent. So the fact that Flacco's rating actually goes up in the playoffs puts him in the minority. Of all the active QBs with a significant sample size from what I can tell the only ones that improve on their rating in the playoffs are Brees, Eli Manning and Joe Flacco.

Coincidentally, all three of them are statistical outliers. Both Brees and E. Manning had abnormally slow starts to their careers, and for elite QBs (if you even consider Manning elite) neither of them follow the usual elite QB pattern of increasing win percentage as point differential increases. And Flacco is one of the few (only?) "elite" QBs that never got his offenses passing DVOA into the top 10 earlier in his career, accomplishing that for the first time in 2013 with the 9th ranked passing game.

 

Joe Montana owns an 85 reg season rating, but 95 in the playoffs, which is amazing for the era he played in. Lights out amazing. And that is over 23 games, not the smallest of sample sizes.

 

I'd love for him to be Tony Romo. Romo has worked his way into the upper echelon of QBs, below Rodgers, but just barely. Quietly amassing the 3rd highest rating of all time (behind Rodgers and Wilson, yes there is rating inflation).

I'd have two lists if ranking Ryan. Where I think they rank now, and the list I'd start a franchise with, where age is taken into account.

Where I'd rank him:

Rodgers.
Brady.
Peyton.
Roethlisberger.
Romo.
Brees.
Luck.
Wilson.
Matt Ryan.
Rivers.
Flacco.
Eli.
Newton
Tannehill
Carson Palmer.
Cutler.
Alex Smith
Kaep.
Stafford.
Bridgewater.

A few guys if they can unlearn their bad habits and play more fundamentally sound, would skyrocket like Newton, Stafford and Kaepernick.

That's where I'd rank them. I wouldn't be surprised to see Bridgewater leap frog a lot of people this year. He played quite well towards the end of his rookie campaign, and now with AP back, and Wallace, I wouldn't be surprised.

Tannehill has had a trajectory upwards, so I'd hope for a huge jump up, just a few more TD's and a nice jump in YPA and we'd have ourselves a nice QB.

 

I have derek carr shooting up a lot of lists.

 

Well this is your first ridiculous post. You don't care about standard deviation means you don't care about the range of values that are likely to occur. So, for you a QB that has a passer rating of 84 on average, but never goes below 80 or above 88 is absolutely no different statistically than a QB whose average rating is 84, but has ratings all the way from 10 to 125.6 (Flacco). Whether 118 over 4 games is significantly different absolutely depends on standard deviation.

At least the cat's out of the bag now. Stats actually don't matter for you.

 

His post isn't ridiculous at all. All you've effectively done is find the range a given player is capable of playing....which is something we already knew.

Of course Flacco is capable of a high level of play and a low level of play....who the hell isn't? Of course the high end of his rating is 118 and the lower end is 84, we've already seen him do that. And of course there's no abnormalities between his range and someone else's. No one suggested otherwise.

84 passer rating is not all world. If he does that more than he doesn't, then he's not a great QB end of story.

 

No, I don't care about standard deviation of your using it to argue that an 84 rating is not significantly different than a 118.

Do you believe that a QB with a career 84 rating is significantly current from a QB with a career 118 rating??

 

His post is absolutely ridiculous. There are valid and invalid ways of attacking a statistical analysis. You could ask whether the assumptions that were made hold, or whether the "level of significance" is something that's appropriate in this case.

What you do NOT do is say something that is necessary for solving a problem is actually unnecessary. Saying you don't care about standard deviation is saying you don't care about information that is necessary to solve a problem.

 

Career rating is totally different than a sample of size 4. Most people don't have an intuition for what to expect when taking samples with small sample sizes. Oh, and I NEVER argued 84 is not significantly different from 118.

 

No, he doesn't care about standard deviation as it pertains to this discussion. I don't either. Standard deviation does nothing helpful in this argument. Again, all you really showed was that the difference between his highs and lows are on par with other decent QBs. That was never up for debate, though.

 

Well then you're making the same error. Standard deviation is necessary to solve the problem of how likely a sample of size 4 with average rating 118 is, when the mean of the distribution the ratings come from is 84.8. You simply cannot solve this problem without a measure of the spread of the distribution. And that btw WAS the question.

 

No, that's not the question nor was that the solution.

People are saying that his playoff run was an aberration compared to how he normally plays in the regular season. They aren't saying that he was never capable of playing at his peak level. They weren't saying his peak level could never be 118. They were saying he plays the majority of his games closer to the 84 level then he does the 118 level. You flat out misunderstood the question.

 

No, you misunderstood the question completely. The question was whether his 2012 playoff run was an aberration compared to how he normally plays in the playoffs. I suggest you read through the relevant parts of this thread first before continuing because you've got the question all wrong.

 

Maybe I'm dumb, again, but you say right here that it isn't statistically significant.

 

No, I am actually arguing, that for his career, both regular season and playoffs, Flacco is an upper 80's rated QB, and thus, four straight games of 100+ is not normal for him.

 

Even if that's the case, you still didn't approach it correctly.

Your approach assumes the argument was that 118 rating is outside of his abilities. That was not the argument. No one was saying that 118 is beyond his reach. All you did was prove that it wasn't (even though he did it already so that just wasn't necessary).

The argument is and was, is the 118 rating outside of what he normally does (regular or post, it doesn't really matter).

Think of it like this you've got two QBs. QB A has a range of 85 -120. QB B has a range of 84-110.

After 100 games: QB A has played 4 games at 120 and the rest under 100. QB B has played 57 games at a level 100 and the rest below a 100.

Who is the better QB?

That has nothing to do with the numbers you brought up, but that is what the argument was.

 

OK.. the wording is very important here. What is NOT significantly different is the mean of the 4-game sample (118) from the distribution of randomly chosen means of 4-game samples, given that you randomly chose from a distribution with mean 84.8 and std = 37.2.

This is not saying career rating of 84.8 is not significantly different than 118. Why? Calculate the standard deviations over a career and you'll see they are very small in comparison. What is it? I don't know but it's NOT 37.2 (probably single digits), in which case 84.8 and 118 would be unbelievably different, as your intuition suggests.

It's also not saying the sample mean of 118 is not significantly different than a sample mean of 84.8, because that's not the question. It's saying 118 is still "likely enough" to come from a distribution of 4-game sample means where those samples where taken from Flacco's playoff ratings.

 

I never assumed 118 was outside of his abilities. NEVER. Why would one ever start a statistical analysis with the assumption you cannot do what was actually done?

The only assumption was that Flacco's playoff abilities were best estimated by his actual playoff ratings, and since we're only talking ratings here that assumption shoudn't be a problem.

The question I posed is exactly whether a 4-game sample with mean 118 is likely to come from a distribution with mean 84.8 and std = 37.2. That's the statistical equivalent of asking whether what Flacco did in the 2012 playoffs is "abnormal" or not. No, I'm not confusing the question nor its representation in math. And precisely because of two things: 1) the std is huge = 37.2, and 2) the sample size is small = 4, 118 is actually not that unusual an occurrence.

btw. to you and resnor: 84.8 + 37.2 = 122 and 122 is ONE standard deviation away, so getting a rating in a SINGLE playoff game of up to 122 is not at all unexpected for Flacco. You just need the Central Limit Theorem to answer questions like what happens when you look at more than one game.

 

Look, I'm not a math guy. If you throw enough of those terms at me in a post, my brain shuts off. LOL. However, no matter what may nee possible, given an infinite number of tries, you can look at Flacco's career, both regular season and postseason, and understand that what he did in the 2012 playoffs was far from his normal play. That is all I'm saying. I'm not saying it's impossible for him (obviously, because he did it). I'm saying that I wouldn't expect him to do it all the time, given his career averages. I'm not sure why that's controversial. Flacco had shown that overall, with his great games and his bad games, that he averages it to a mid-upper 80's rated QB. It just is what it is. Sure, he may flash something tomorrow, but then the following week he may have a rating in the 40's.

 

That's not what he's saying.

He's showing the 4 game stretch is not an aberration, compared to what he's done before. Perhaps you skipped over it, or there was some sort of miscommunication. cbrad never said a 4 game stretch of 118 or whatever, is not statistically significant than a career playoff record of 84. That's not what he said.

He set out two datasets:

1 - His entire playoff history, including his very bad first two appearances
2 - His playoff log from 2010 and on.

In either case, his 4 game stretch of excellent QB play in 2012 is not an aberration, and well within what could be expected given his entire playoff record, based on the concepts he applied.

 

Yeah, I'm trying not to deliberately throw math terms at you, but to be precise enough I kinda have to a bit

I fully understand your intuition here. All I'm saying is that sometimes intuition is off a bit.

In this case, what's counter-intuitive comes from the very small sample size of 4. Let me just point out that if Flacco had 5 consecutive games with mean 118 that is EXACTLY at the threshold (still not "abnormal" but right at the edge). If he had 6 in a row, then that would be considered statistically speaking too unlikely to occur. I bet your intuition wouldn't have told you that.

 

Last year he had a 4 game stretch, 120, 104, 106, 99. He's streaky as they come, but it's not out of this world impossible to happen again. He is as streaky as they come though.

 

Your response to my post #300 where I laid out the math shows at least one of the arguments you were making is exactly what I said it was. You say this: "I was simply using it to mean, 2012 was much different than his other playoff years".

And posts prior to all that show that was one of the arguments being made here (2012 playoffs being an aberration given what Flacco has done in the playoffs).

 

Yes, cbrad, but there is a difference between saying statistically, it's not an aberration, and saying that based on his usual play it is an aberration. I think both are correct statements, but we're both arguing different things, essentially. You're saying it's not an aberration, as it's statistically possible. I'm not denying it's possibility, I'm saying that when looking at his career numbers, it's not normal for him.

 

I'm saying statistically likely, not statistically possible. What's statistically possible is anything that's physically/mathematically possible, so a passer rating of 0 to 158.3 is statistically possible. The value of stats is that it gives you a way of quantifying how likely something is. So yes, I'm saying that 2012 playoff stretch was not unlikely enough to claim this isn't what you'd expect from Flacco in the playoffs.

 

Since five in a row is borderline against the odds, if you take the four from 2012, and the first from 2015, you have five playoff appearances in a row over 100 rating. Does that matter that it skipped a year?

 

So you believe it's likely, despite Flacco's years of performance, that he will again have four straight playoff games over 100? I really feel like this is a situation where stats on probability less us astray.

 

Doesn't matter that it skipped a year the way I set the problem up but it has to be consecutive. If you find 6 in a row, you've got statistics on your side here. With 5, a statistician will still report it's likely enough.

btw.. since both using stats and eyeballs it looks like Flacco was different from 2010, it might help intuition if you use mean = 104.7 and standard deviation is 19.3. Then it "feels" less unlikely. But of course that's throwing away data, so one has to have a good reason to throw it away (I think one can make the case, but that's another story).

 

Since we're talking about a 4-game stretch, I'd estimate from intuition (the math is more difficult here) that if you saw something like that on average 2 or more times every 10 years, some statistical test might say there's something fishy going on. I can't be sure, and my intuition here won't be as good because it's a bit complex, but I think that's the right ballpark. One out of 10 years will almost certainly pass, but 2? I have a feeling not.

Oh, and I'm assuming here that the overall playoff average still is 84.8 and we're talking 118 average for the sample, not simply 100+.

 

The past 3 appearances, he's played damn well in the playoffs. Let's leave it at that.

48 TDs is an aberration too. Hell, Dan Marino throwing for more than 30 TDs is an aberration, only happening twice his entire career. Brett Favre actually has more 30+ TD seasons than Dan (and this was pre-2004) and led the league in TDs more than Dan.

Dan was all aberration buddy!

 

When he came into the league, he really was an aberration (speaking intuitively only here, but I bet the stats bear that out) compared to what came in the decades before him. Only problem with any analysis involving Dan is going to be that HUGE rule change with defensive holding in 1978. I think that pushed up average ratings by double digits.

 

Yeah, that was an aberration. Did you expect Marino to throw 48 every year?

 

Sigh.

I didn't say you thought 118 was out of his realm, I said you thought the argument was that 118 was outside of his realm, hence your approach. You were trying to illustrate that it was not outside of his realm.

You are still missing the point. If Flacco had played every playoff game at around an 84 rating, no one would be talking about him. However, he had one magical post season, that he played to the top of his abilities, and now people act as if he is a qb that regularly plays at that level, when he doesn't. That is the argument. People are comparing Flacco to Tannehill based on one single 4 game stretch. Nothing more. That four game stretch, within the standard deviation or not, is not an accurate reflection of the rest of his career, yet it is the single thing used to say he's better than Tannehill....4 games.

Again, all you've shown was that 118 is not beyond his capabilities.

 

This is what you said:

"Your approach assumes the argument was that 118 rating is outside of his abilities. That was not the argument. No one was saying that 118 is beyond his reach."

And that is false. I did not assume that.

I'm not missing the point. You're just not understanding what I did. Let's just leave it at that. I've explained this well enough multiple times I think.

 

If that's not what you thought the argument was, why did you take the approach you did? Because what you showed is that Flacco is capable of going 84 to 122 and that is perfectly normal....when no one argued that.

 

Let me try this...It is possible for me to flip heads 20 times in a row. You could even find the probability for that. However, if I do it one time, you should not treat me as special, and judge my entire lifetime of coin flipping based on one occurrence.

 

I showed that if you take Flacco's entire playoff record as the best estimate of his playoff ability, then getting 4 consecutive playoff games where the mean rating is 118 is not something a statistician would say is so unlikely that it's inconsistent with what Flacco's done. Importantly, these 4 games don't even have to be the exact 4 in 2012; they just need to have a mean of 118.