Oh yeah, I forgot. You're one of those guys who thinks playing the Jets, Bills and other bottom dwellers (like us) is the same as playing the best team in the NFL. With no difference in pressure or competition.
It was a flukey play. He made a nice play to get away from the defenders and basically threw the ball up for grabs and benefitted from a miraculous catch. With Eli doing everything exactly the same, there was a pretty good chance that ball would have been intercepted.
Yes, I agree that Super Bowl stats have little scientifically significant relation to clutchness. As for Flacco in 2012, that was not too unlikely to be explained by chance. Frankly, I'm surprised you would even make such an absurdly loose statement. Even when there is a .001% chance of something happening by pure chance, there is still a 1 in 100,000 chance of it happening by pure chance. That sounds like a remote chance, but it really isn't. People win lotteries by pure chance with far worse odds than that all the time. And its clearly not true that the odds are so stacked against Flacco having a rating in the 120 range for a single game. Of his 122 regular season games, he has had a rating higher than 117 16 times. In other words, he posts a 117+ rating in roughly 13% of his games, or almost 1 in 7 games. Those are hardly remote odds.
You should learn statistical analysis before acting like what I said is an "absurdly loose" statement. Listen.. for any data there are many possible hypotheses that fit the data. That is, your initial hypothesis isn't the only possibility. The real question is which of various hypotheses is most likely to be correct. That is, if we assume a candidate hypothesis is true what is the probability of the observed data occurring? The hypothesis with the highest such probability is the most likely to be correct. Problem is there are in principle an infinite number of different hypotheses, so what is commonly done is to set some threshold beyond which we say we should reject the hypothesis given the data. The most common threshold used is the 5% threshold: if the data would occur less than 5% of the time given that hypothesis, it's better to reject that hypothesis and look elsewhere. That threshold btw differs from field to field, but 5% is the most common. You're confusing "possible" with "probable". 1 in 100,000 is possible, but arguably too unlikely to keep holding onto the hypothesis for which that data would only occur 1 in 100,000 times. You simply don't want to ever change the hypothesis you started off with. I can't respect that.
As I'm sure you know, the teams playing in the Super Bowl aren't necessarily the best teams and certainly aren't necessarily the best defensive teams. The two Giants teams that the Pats lost to in the Super Bowl ranked 17th and 25th in the NFL in points allowed and 19th and 22nd in passer rating allowed.
Looks like from 2000-2015, the average rank (in regular season) in points allowed for the SB winner is 7.9, and there were 5 out of 16 years where the defense was ranked #1, and 8 out of 16 where it was ranked top 3. So clearly on average, the competition is much better (how would they get to the SB if they weren't?).
I'm not confusing anything and I know enough about statistical analysis to know that your assertion that Flacco's 124 rating in a single game is too unlikely to explain by chance is nonsense when he posts a rating in that same vicinity in 13% of his games. In other words, it is a regular occurrence. It happens 2-3 times each year on average. One time it happened to have happened in the Super Bowl. I'm not sure what you are defining as the likelihood of happening "by chance." Maybe you are getting cute there with your wording. But when something happens 13% of the time based on the full data set, it is hardly unexpected or miraculous when it happens again. The 5% statistical significance "threshold" isn't particularly useful or applicable when you are looking at a single event, as opposed to a pattern of events. If what you are saying is that a 124 rating only has about a 5% chance of happening in any particular game, that's fine. But that certainly doesn't mean that when it happens it is due to clutch or anything else. It simply means that given a set of 100 events, that result is likely to happen about 5 times. If they are randomly distributed, some will fall in the playoffs, some will likely fall in the regular season and some might fall in the Super Bowl. And that is by pure random distribution. And that Flacco Super Bowl performance looks very different if not for some flukiness on that Jacoby Jones 56 yd TD. [video=youtube;L2AwRAjweF8]https://www.youtube.com/watch?v=L2AwRAjweF8[/video] First, Jones got open by a country mile. That was important because Flacco's throw was just awful. A really, really bad throw. With any kind of respectable coverage that ball is picked off. But even giving him the benefit of the catch, if Jones had been tagged down at the 8-9 yard line, Flacco's rating would have been far lower. And if Jones didn't come up with the catch due to the pathetic throw, it would have been lower still. The point being that there was plenty of luck that went into that 124 rating.
Over the last 10 years the average QB loses 7% of their passer rating in the playoffs, so he's performed better than most QBs do in the playoffs. For reference that's bigger than the drop that Tannehill had between 2014 (where he was generally praised for breaking out with a 92.7 passer rating) and 2015 (where he 'regressed' to 88.7). If Tannehill had lost 7% he would have droppedfrom 20th in the NFL to equal 26th in the NFL with an 86.2 rating, Below Sam Bradford and equal to Blaine Gabbert.
Sure, on average it is. But when we talk about individual games or even 2-3 SB games, the average doesn't matter much. It's the individual opponents and matchups that matter. Take the 49ers team that Flacco had his big day against. General perception is that the Niners had a great defense. And in some ways did. But Flacco wasn't alone in putting up big ratings against them late that season. In 2 of the previous 4 games, they gave up ratings of 114.8 and 115.3.
No, you clearly do NOT know enough about statistical analysis. You don't even know what the word "likelihood" means (it's a technical term btw). And the 5% threshold IS useful for single events. That's actually precisely how it's used almost all the time. Finally, I was basing the analysis on the distribution of league-wide drop-off from regular season to SB rating. I wasn't looking at the SB rating per se based on regular season distribution because once again that totally masks the effect of all other factors that are different from regular season to the SB. Naturally you from the outset assume there can't be any difference between regular season and SB so you go the wrong route. Anyway, I've heard enough. You don't know stats yet you want to act like I'm wrong with basic stuff, and you simply won't let the data dictate the hypothesis.
Yeah.. when half the time you're facing a top 3 defense yet you cherry pick 2 cases where the defense was worse than average, it's not very honest IMO to say "average doesn't matter much".
I absolutely do know what the term likelihood means. It has a simple, straightforward non-technical definition that is quite easy to understand. Merriam-Webster defines it as "the chance that something will happen." http://www.merriam-webster.com/dictionary/likelihood I wasn't using it in any technical sense because that is not what I intended, nor is it what anyone here (other than you) cares about. Basing it on the distribution of league-wide dropoff from regular season to SB rating is silly. We have a specific player. We know how often in 122 regular season events he posts a rating in the 124 range. To evaluate the likelihood he'll do it in any particular game, look at that well-defined data set for that specific player. Your claim that it "masks the effect of all other factors that are different from regular season to SB" is nonsense. It hasn't been shown or proven that anything is different. The Super Bowl is a game played by the exact same rules as any other game. Your approach is the one that tries to be outcome-determinative to support your thesis that there is a difference. But again, I see no compelling evidence that there is a difference, especially when it comes to pressure and/or clutch. As you note, on average, the defenses in the Super Bowl are better than in the regular season so in most cases you will tend to see somewhat lower passer ratings. But it isn't always the case -- several teams have made it to the Super Bowl with below average defenses. And when a guy has a big game in the SB against one of those defenses it is not miraculous or proof of clutch, it is just a below average pass defense giving up a big game like they often do.
Lol.. you tell me I'm "absurdly loose" with my wording, so I use words like "likely" in a technical way to make sure it's precise enough. Now you come back and say who cares what the technical definition is. Wow.. Right.. "average" doesn't matter. It doesn't matter that statistically speaking the playoffs and the Super Bowl on average include better competition. No.. just ignore all that and calculate the probability of the SB rating given the regular season distribution. Sure, works until you realize the conditions ARE different (I know you assume they aren't but there's data suggesting otherwise). Once there's evidence the conditions are different you got to remove the average drop-off. Anyway, if you get it fine, if not no purpose in debating. You can have the last word on this Fineas.
Using a term according to its common definition and usage is not imprecise. At all. I realize that you are someone who likes to argue for days over whether cold exists or whether water is wet, but not everyone does. And it doesn't make you smarter or wiser. Clearly, you are aren't paying attention. Averages matter when talking about overall trends, but not when talking about individual events. As I explained many times, the overall higher competition level in the playoffs is why you tend to see lower passer ratings. Not pressure or clutch. So it does matter when talking about overall trends. When talking about a single game, those averages don't matter much. It is one team against another. In such a situation, it doesn't matter what other teams in other years did.
Cbrad has been very precise with his use of language, and everyone who has read this thread knows it. Averages are, by definition, made up from a series of individual events. So either single events are useful and you incorporate them into averages to assess individual performances in specific events, or you assess every football transaction as an isolated one off event. The average helps when you compare what the person did in situation X to what they could reasonably expect them to do. When there's a variance you then look at explanations for the variance. You can't assess someone's performance unless you can say what you could reasonable expect them to do, and to do that you need averages. For example. If QB X has a regular season average of 100 points. Then we know that you can expect him to have a 93 rating against an average playoff defense. If we know that the opposing team Y's Defense is 10% better than average playoff Ds then a passer rating of 93 x 0.9) 83.7 is a reasonable expectation. So if in a playoff QB X gets a passer rating of 85, some people might say OMG he choked he was 15% worse than regular season, but statistically he's doing a little better than expected.
OK, but didn't you just argue with me that Brady playing to his avg in the SB was just that? Avg? And weren't you just arguing that the SB doesn't mean it the best teams? And you have the absolute audacity to accuse someone else of "liking" to arguing? Question, do you have any mirrors in your place?
Yes, let me spell it out again. Overall, the competition in the playoffs is better and you tend to see that impact QB ratings. But when you look at small sample sizes like 6 or fewer games, those overall averages aren't what matter. What matters are those specific matchups. So in a few of Brady's SB appearances (Giants), he faced pass defenses and overall defenses that were in the bottom half of the league. Brady has been Brady in the SB, just like Brady is Brady in the regular season. He's really good. But he didn't play better in the Super Bowls. And his Super Bowl matchups weren't necessarily against great defenses.
That's nonsense. Let's take an extreme hypothetical to illustrate the point. Let's assume that the average SB defense is ranked No. 1 overall and allows an average passer rating of 55. QB A plays in one Super Bowl and in that game plays against the league's 30th best pass defense that allows an average rating of 96. In that game, QB A posts a rating of 90, which is well above the average in other Super Bowls. Is the clutch? No. H is rating is lower than what that defense usually gives up. Do the average ratings of players in other Super Bowls matter with respect to QB A? No.
No what you said is literally nonsense, as in it's mathematically impossible if you assume random variation (i.e. 30 teams aren't all ranked the same). How can you have the average SB defense be ranked #1 overall yet have a #30th ranked pass defense in one SB? That's impossible in practice. Either way, to do any kind of DPUP calculation you need to calculate some expected value and compare to that. We don't know in your example how much that 96 rating should be expected to drop in the Super Bowl. That's what you first calculate. Secondly, because there is random variation you need to know the distribution of such drop-off's (expected variance) to calculate the probability something occurred by chance, meaning you have to look at the entire league drop-off, etc.. Anyway, we're already at the point where I think either you just don't understand this or you won't let go of your initial hypothesis no matter what. I'll respond if there's some new point (technically the post I quoted qualifies), but I'm winding down the "debate" with you because it doesn't look like you want the data to drive the hypothesis.
Eli's internal clock and realization of the "moment" gave him the extra effort on that play. If this is week 4 at Chicago he just goes down. The recognition of the moment and what you do in those moments are what makes you "clutch" Sent from my iPhone using Tapatalk
If the NFL could be solved with "numbers" half you nerds would have better jobs. It's not that simple nor has it ever been. Too many variables. It's fun reading this stuff though, although it's taken with a 32 oz cup grain of salt. Sent from my iPhone using Tapatalk
Totally agree Fin-O. But there's no question IMO that stats-based analysis can provide useful info you'd otherwise miss, either because people have biases, incomplete memories, or the inability to make certain derivations. All of that depends of course on agreeing on what you're looking for, which is why choosing a definition of "clutch" that helps tease apart something you think is useful is important.
I meant to say that in all previous SB's the defense was ranked No. 1 and in this one most recent example its 30th. But with a large enough sample size (again this is a hypothetical), you could approach that. I get that you want to be hyper-technical so we can play that game. 50 prior Super Bowls with each previous one being No. 1 defense v. No. 2 defense and their average rating allowed being 55. In the 51st, you have No. 1 v. No. 30 and QB A is facing the No. 30 defense that gives up an average rating of 96. So the average SB defense rank is 1.77. Does that make you feel better? I don't expect that 96 rating to drop in the Super Bowl. If it is a very good QB, I expect it to rise because that 96 rating was given up to mostly average QBs, not a great one. As for the distribution of drop off, if there is any, that will be largely a function of playing great defenses, not pressure or clutch. Sounds good to me. I do want the data as I am confident it will show my position to be correct. But frankly, I'm tiring of this.
Exactly. In the NFL, every game matters unlike basketball or baseball. So if a qb can make a play and doesn't because "it's the 4th game" then that qb has no business being on the team.
Yeah, if the 96 is for the defense it should rise (I temporarily forgot this was defense). My question though is what kind of data would you want? If mean.. what happens if there are multiple teams that satisfy similar conditions? Then you'd have to take the average or? In other words, what other teams do will matter if you want to talk about "expected" change.
You call a speculation I call it commonsense. We can watch thousands of Eli drop backs and never see a play like that from him, yet he does is on the biggest of stages with minutes left on 4th down in the 4th qtr... Probably just a coincidence eh... I think you unlike some are smarter than that, problem is it goes against your theory, so instead of putting that into perspective you reply with "speculation"?? You are better than that Sent from my iPhone using Tapatalk
Absolutely. But, stats are more detailed and more widely available than ever before allowing a higher degree of analysis than previously possible. Also what we've been trying to do is find general trends which is possible. Stas also help you find out whether someone is performing differently from what is expected (average), but it won't, by itself, identify the reason for the differential.