But they're not "clutch" in a statistically significant way. Assuming a normal distribution, roughly half of all players will have stats that are better than the average. But as you know, that could be pure random chance. Indeed, for many of them it certainly is. And even when it is unlikely to be entirely due to random chance, there could be a lot of reasons for it that have absolutely nothing whatsoever to having an inherent ability to perform better under pressure. It could be due to matchups, coaching tendencies, playing at home, opposing player injuries, the play of teammates, etc. And if you are going to use a term like "clutch," there should be some causation between the apparently elevated play and some inherent ability to rise above the pressure, etc. If we use your definition of clutch, Timmy Smith may well be the single clutchest player in professional sports history. In fact, it probably isn't close. Yet, if you really believe Timmy Smith was truly clutch and simply elevated his level of play from journeyman scrub who averaged 3.17 ypc in his regular season career to 22 carries for 204 yards (9.27 ypc) and 2 TDs in the Super Bowl, you are either (1) on delusional, (2) plain stupid, (3) didn't watch that Super Bowl, or (4) some combo of the above. The holes he got in that Super Bowl were massive and any NFL RB could have run through them with equal success.
I agree that they are weak sauce. Very, very weak. Which is why, when combined with the small sample sizes and the myriad of other potential explanatory factors, they are meaningless. If you want, as cbrad seems to, to label as clutch any player who, no matter how small the sample size and how small the margin, produced some number in some situation that might arguably have slightly more pressure than the norm at above the league average rate, then I guess you can. But I think that makes for an absurd definition of "clutch." It's akin to defining as "awesome" "any player who once did something that didn't absolutely suck to such an extent that it warranted his immediate release from the roster."
First of all, in many cases we are not looking at normal distributions. Any QB clutch stat that correlates highly with win% will likely not be normally distributed if we look at all QB's (without truncating) because the distribution of win% among all QB's is highly skewed with mean win% being well below 50% across the league (I showed that on page 4, post #72). With that type of skew in win%, you're looking at well less than 50% that is above average (precisely 50% is above the median by definition, but much less than 50% is above the mean). Of course, if the clutch stat itself truncates QB's because only QB's with many attempts would even have those stats, the effect of win% skew will be less. Secondly, whether it's distributed normally or not says nothing about the causal factors that led to the distribution. When you say "random chance", you're referring to causal factors other than any mechanism internal to the athlete that responds differentially to "greater pressure" situations. The problem is this: we know there is a league-wide drop-off from less-pressure to more-pressure situations, so some differential response to greater pressure situations is almost certainly involved (untenable to suggest otherwise). In any biological system, there is random variation in every trait, meaning the default assumption here is that random variation in how people inherently respond to, or learn to respond to, greater pressure situations is part of the variance observed in the drop-off distributions. That's completely independent of the type of distribution observed, and is the causal mechanism implicitly being referred to here for "clutch" players. So yes, I agree it's hard to argue other factors aren't involved. But I think it's equally hard to argue differences in how people respond to pressure aren't involved.
I have compared 6,000 0-7 point games and comparing HoF to Non HoF QBs. I compared ALL quarterbacks in the NFL with 100+ starts. How is that a limited sample? Quite frankly it's easy to say "it might be something else" but doing nothing to prove it is true. I can only think of one time in this entire thread where you got off your arse and posted some facts to back up what you were saying. Cbrad and I have both come up with definitions of clutch that can be measured mathematically and we can demonstrate in large sample sizes that some QBs are consistently better in those situations than their peers. You on the other hand are being like Humpty Dumpty in 'Through the Looking Glass' saying 'clutch' means just what I choose it to mean—neither more nor less" and changing your definition of 'clutch' whenever you feel like. If you think Cbrad's definition of 'clutch' is inadequate please enlighten us with what you think the definition should be.
No, when I say random chance I mean random chance. Other causal factors are a whole 'nother layer on top of that. There isn't an overall "league-wide" dropoff from less pressure to more pressure situations for all such "clutch" stats. There are dropoffs for certain stats and I have explained several times why those drop-offs are due to a myriad of factors that have nothing to do with pressure. The main one I think you are referring to with the dropoff is the playoff stats, but there is a lot more going on than just increased pressure. Playoff defenses are much better, on average, than the overall average defenses (including non-playoff teams). As I explained, the weighted average of passer rating allowed among playoff teams during the regular season is 8-10 points lower than the league average. In 2015, 11 of the 21 playoff games by QBs were against Top 5 pass defenses. That makes a big difference. I also pointed out the weather factor and you came back and noted a 2-3 pt drop in passer ratings in December. That may well be a bigger drop in January. So any apparent drop in playoff ratings has a lot to do with factors that have nothing whatsoever to do with pressure. Indeed, a few of those factors appear to account for virtually all of the playoff drop. My point is that (with appropriate sample sizes) you just don't see players that consistently play much better or worse in pressure situations. You see a mix of great performances, average performances and terrible performances for virtually everyone. And those happen at about the same rates in which they happen in non-pressure situations. When you look at stats that are largely independent of the impacts of other players and the opposing team, you really don't see a drop off in close games. For example, in 2015 NFL kickers made 77.77% of their 40-49 yd FG attempts in the last 4 minutes of close games. But they made such 40-49 yd FGs at a 76.89% rate overall. Those numbers are virtually identical. Similarly, in the NBA players make "clutch free throws" (defined as under 2 minutes with neither team up by more than 3 points) at the same rate (.751 v. .753 for last year I saw such stats) as overall. QB play, of course, is not at all like that. It is inextricably intertwined with the play of 21 other players on every play.
Dominating physical play by the Redskins OL (which still hasn't gotten it's just due as one of the greatest O-lines of all time - along with the Cowboys in the early 90's, the Packers in the 60's, and the Dolphins, Raiders, and Steelers in the 70's). This is always the problem - whether arguing clutch or other factors - in analyzing players in the NFL. So many other factors come into play. That said there have been QBs who have been able to perform as well or better near the end of games and performed under intense pressure. Elway was always a better player in the 4th quarter than other times (though Marino was as good or better in the 4th in comebacks. He just never had a defense. However, Dan was also better than John in the 1st-3rd quarters). You can see others - like Montana against Cincinnati and other big games; Brady in several very close, big games - who performed exceptionally well and made needed plays to win their games. In the previous era, Roger Staubach was the best I'd seen at end of game/ 4th quarter moments. Oh, also, Big Ben has performed well under clutch (But also rescued by some great plays by receivers). You can also see players who did not perform so well in the clutch - who were nervous or threw bad passes. But, sometimes a failure by a QB is noticeable. I don't think you can measure it by statistics, directly. You can simply see players make great plays in the heat of a tight game. And, that's enough for me. You can't "rank" players in some direct order. You can simply say, player X was able to relax and play his best under the most pressure filled moments. On that count, Marino, Elway, Montana, Brady, and Staubach are all at the top for me as being able to perform well or even improve their performance under critical moments. I'd put Ben R just below them. Favre was pretty good at times as well - he always played relaxed though, sometimes too relaxed. Rodgers has had some great games and great plays near the end of games and is moving into that category. Let's see what it looks like when his career is over, though.
That's illogical. Everything due to "random chance" has a cause. So random chance is part of "other causal factors". Random variation within "differential response to pressure" is already being taken into account when we do sample size calculations. We already went through this. I told you that with playoffs one factor that hasn't been adjusted for is strength of competition (after your comment on weather), but that these types of arguments don't work for why Brady wins 69% of close games, both in regular season and playoffs, when ahead at the end of the 3rd quarter at most 50% of the time. They also won't work for <4 minute trailing stats. You simply can't provide evidence that "strength of competition" or other factors excluding QB performance were responsible for those stats. Of course you tried to argue it was the rest of the team, to which I said there's no way you can't give the QB a big dose of credit. I pointed out I once estimated based on YPA that 15-20% of an entire game, or around 30-40% of offensive performance, was a decent estimate for how much the QB affects the final outcome, and you yourself agreed that was a good estimate. So yeah, we're actually in agreement here that a good portion of the observed variance is due to differential response to pressure by the QB. You keep bringing this "small sample size" argument into the thread repeatedly, yet you seem to have selective amnesia on this issue because I have repeatedly shown you how to take sample size into account. That 0.21% chance Flacco's playoff record was NOT due to improvement was calculated GIVEN a sample size of 15 (results are different if it's 16 or more). In the Tannehill vs. Luck thread, I calculated an absurdly low probability Luck's close game record would have occurred even once in the entire history of the league given a sample size of 27. Didn't we just get through talking about Lemieux where I showed the probabilities of different playoff records in different seasons for the given sample sizes? Point is, sample sizes are either being explicitly taken into account, or they aren't an issue because we know it's "large enough" based on previous research (most stuff with passer rating falls into this category), or they aren't an issue because we aren't just looking at a single stat. Remember the argument is how an athlete performs on average across many stats. Let's say there is a high enough probability of something occurring randomly in each of several stats. The probability ALL of them occurred randomly will be much smaller (the product of all those probabilities, assuming all those stats are measuring different things.. otherwise you have to adjust a bit). Seriously.. if you want large enough sample sizes, I can easily manufacture that for this argument by taking multiple stats into account, so sample size isn't a valid counter-argument here.
What you are talking about are isolated, anecdotal examples that are not supported by the full sample set. For example, you say "Elway was always a better player in the 4th quarter than other times," but he wasn't. His 4th quarter rating was worse than his 1st quarter and 3rd quarter stats for the time period for which data is available (they don't seem to have that for his full career). Joe Montana had some great playoff moments, but he also had several times when he completely and utterly "shat" himself -- he had 3 playoff games in which his passer rating was below 43. That's pretty terrible. He was a great QB so he had more good games than bad ones in the playoffs, just like in the regular season. Brady, Marino, Elway and Staubach all had some massive stinkers in the playoffs too. You have to look at the whole picture and not just the great moments that you remember.
No. There will be some natural and random deviation in results even when controlling for all variables. If 100 people each flip 100 coins the average/mean and median number of heads will be somewhere around 50. But not all 100 people will get 50 heads. Some may get 38 and others may get 63. I'm glad you brought up that Brady 69% stat because it is questionable at best. According to the search I did, Brady's Pats won 34 of the 75 close games in which they were tied or trailed after 3 quarters. So they lost more than they won. Of those 34 wins, they were tied after 3 in 12 of them, so those really weren't comebacks at all. And of the actual comebacks, only 1 was when the Pats were down by more than 10 points after 3. The average "comeback" was 3-4 points. And you completely mischaracterize our agreement and whether I agreed that Brady is entitled to a big dose of credit. He certainly played a role, but I don't know about the big dose at all. As I pointed out, the Patriots' late game pass defense through his career has been outstanding, giving up a passer rating of around 66 if I remember correctly. That plays a huge role. The Pats' unprecedented stretch of outstanding kicking also undoubtedly played a role. Certainly Belichick and the coaches are entitled to credit too. As for the sample sizes, as you undoubtedly remember, we had a misunderstanding as to what the Flacco % was supposed to convey. You agreed that it does not represent the likelihood that his regular season-playoff differential is due to clutch. You acknowledged it is just looking at the playoff sample itself and showing the likelihood that it is due to some kind of real "improvement." By response to that, to which you haven't responded, is why Flacco then seems to have lost such clutchness the last few years. Why his late game numbers are and continue to be below average. And why he can't seem to translate what you say he "learned" into any kind of lasting improvement. If it was actual improvement, it should have been sustained over the last few years. It hasn't been. My point is that he had a nice run, just like he had in 2010 and just like he had against the NFC South. There is no indication it was caused by pressure or clutchness. Sample size isn't an issue for everything, but it is when you look at a single player's numbers over a handful of games. Or when you exclude other relevant stats for the convenience of the argument. Like Flacco's first 5 or so playoff games.
I'm being technical here. IF you truly control for ALL variables, there is NO random variation left. The proposed mechanism of differential response to pressure won't be the same for every time and place, obviously. But like I said that's already accounted for in those probability calculations that take sample size into account. So when you say "random chance" as something separate from the mechanism I'm referring to, yes it is part of "other causal factors". Brady had 91 close games (0-7 point differential). He won 63 of them, giving you 63/91*100 = 69.23% win probability in close games: http://www.pro-football-reference.com/players/B/BradTo00/splits/ So absolutely nothing questionable about the 69% stat. Furthermore, there's no way your stat is correct. You say he won 34 of 75 close games where he was tied or trailing at the end of the 3rd. That means 16 games (according to you) he was ahead at the end of the 3rd because there are a total of 91 such games. Even if he won all 16, that means his total wins in close games is 50 out of 91. Problem is he won 63 out of 91 so your stat is verifiably wrong. Once again, I went through 90 of those 91 games (the boxscores) and found 44 where he led at the end of the 3rd. I'd very much like to see someone double-check that, but your stat can't be correct. Whether you agree to the term "big dose" is irrelevant. Point is you agreed 15-20% is a good estimate (post# 645). And let's remember that if the QB gets 15-20% of credit for the entire game (that's including everything), then saying he gets 30-40% for the offense is very conservative because that puts credit for the offense at 50% total. If you think offense should be less than 50% total because of other factors like special teams and coach, then it's 15-20 percentage points out of less than 50, which will lead to more than 30-40% of offense being due to the QB. 30-40% of offense due to the QB as a conservative estimate I personally would say is a "big dose", but whether you call it that or not, you implicitly agreed that conservatively 1/3 of passer rating and other offensive stats (that we would say are QB stats) is due to the QB. That's more than enough for us to agree that a significant (i.e. 1/3 or more) portion of the variance in those offensive stat drop-offs is due to differential response to pressure of something internal to the QB. Agree to that and our argument is over. Who cares what the explanation is for Flacco's statistically highly likely improvement in the playoffs in this context? Just because we don't understand the causes behind everything that we observe doesn't mean you can't calculate the probability something was due to chance. The important point here is that the stats I'm using (or Pauly's using) for this clutch argument don't suffer from some "small sample size" problem, especially since combining multiple stats will for certain make the probabilities different sets of stats are due to chance more than small enough.
I think we are talking about different things, but I don't really care to have a technical argument so I'll leave it at that. You link isn't working, but my search shows different results. I'm getting 107 close games (incl. playoffs) with 73 wins, which comes out to 68%. Basically the same percentage as yours. Of those wins, the Pats were trailing at the end of 3 quarters 25 times (I previously incorrectly included a 2008 game that Brady didn't play in). They were tied at the end of 3 quarters in 9 of those games and were winning at the end of 3 quarters in 39 of those games. So most (almost 2/3) of those close game wins came with the Pats either tied or winning after 3. Of the ones when they were down after 3, it was generally by small margins. Only 1 was a 10+ pt deficit. Only 8 were deficits of 7 pts or more. These were not incredible or monumental comebacks. Sorry if you don't like those numbers -- bring up your complaint with the folks at Pro Football Reference. Not sure when you did it or if you included playoffs, but as of today, and if you include playoffs, your numbers are wrong according to the Pro Football Reference search engine. I agree that, in general terms, QB play is probably somewhere in the general vicinity of 15-20% on average. Really, what I said was that I wouldn't quibble with your estimate as a general proposition. But the general proposition doesn't mean that in this particular sample Brady is responsible for 15-20% overall. Looking at that 25 game sample when the Pats came back from a (usually small) 3rd quarter deficit, it becomes quite clear that the defense was a big factor. In those 25 games, the Pats defense never, ever gave up more than 7 4th quarter points. They pitched a 4th quarter shutout in 8 of those 25 games and gave up 3 4th quarter points or less in 16 of those 25 games. That is 4th quarter defensive dominance. So having looked at those stats, I am even less willing to assume Brady was a big factor. This is where we have a huge disconnect. If those Flacco numbers are caused by factors completely unrelated to handling pressure or rising to the occasion, then they are not evidence of clutchness. That's the whole point. The reason for those good performances matter, just as the reasons for bad performances matter. Again, without such reasons, one is forced to conclude that Timmy Smith is the clutchest player in pro sports history because, according to the numbers, he elevated his game in the Super Bowl in a manner never seen before or since. But a half-blind 6 yr old watching his first football game could probably discern that it was all about the OL and the giant holes they created.
Both our numbers are correct (now that I see you were conditioning on winning and including playoffs). The 91 was regular season, I did the 16 playoff games separately. You're showing the distribution in the cases where he won, while I showed that he was leading less than 50% of the time (just about 50% in the regular season, and less than that in the playoffs) at the end of the 3rd quarter in ALL close games (games ending with 0-7 point differential). Since he was leading at most 50% of the time in all games, but won 69% of all games, that provides a "clutch" stat you can't explain away with weather or strength of opposition, which is all I was pointing out here. That he won more often when he was ahead at the end of the 3rd is no surprise. If the offense HAS to score, how is it suddenly viable to credit less to the offense than in most non-close games where NE was ahead at the end of the 3rd and the offense does not have to score? If anything, the offense should get more credit, which would raise Brady's overall contribution since his contribution to the offense isn't changing. I think it's perfectly reasonable to credit the QB with something like a minimum of 1/3 of overall variance in offensive stats that are commonly called "QB stats", like passer rating, even in these clutch situations. This particular Flacco stat was solely aimed at countering your suggestion that the numbers were possibly due to chance, nothing else. 0.21% probability they're due to chance. To suggest none of the statistical improvement was due to Flacco requires proof on your side, especially given how much of variance in passer rating is at a minimum likely due to the QB (1/3 from the discussion above).
Fineas, Until you put your opinion on the line about what clutch is or isn't you're behaving like a troll. You're repeatedly saying other people are wrong, but give no indication of what could satisfy your conditions. When specifically challenged to provide a definition you just rehashed old arguments and tried to muddy the waters. Again.
OK.. re-checked the stat, and this time I got all 91 regular season games. Brady was ahead at the end of the 3rd in 45 out of 91 games that ended with a score differential of a TD or less. So he's ahead 50% of the time at the end of the 3rd in games that end close, but wins 69% of those games. Only 7 times was it tied at the end of the 3rd. So everyone can check themselves, here's the data (the last column shows score differential from NE's point of view): 2001-09-23, NYJ, L 3-10, 3rd: -7 2001-11-18, STL, L 17-24, 3rd: -7 2001-12-02, NYJ, W 17-16, 3rd: -2 2001-12-16, BUF, W 12-9, 3rd: +3 2001-12-22, MIA, W 20-13, 3rd: +17 2002-09-22, KAN, W 41-38, 3rd: 0 2002-09-29, SDG, L 14-21, 3rd: -7 2002-11-10, CHI, W 33-30, 3rd: -11 2002-11-17, OAK, L 20-27, 3rd: -11 2002-11-24, MIN, W 24-17, 3rd: +7 2002-12-29, MIA, W 27-24, 3rd: -8 2003-09-21, NYJ, W 23-16, 3rd: +7 2003-09-28, WAS, L 17-20, 3rd: -10 2003-10-19, MIA, W 19-13, 3rd: 0 2003-10-26, CLE, W 9-3, 3rd: +3 2003-11-03, DEN, W 30-26, 3rd: -4 2003-11-23, HOU, W 23-20, 3rd: 0 2003-11-30, IND, W 38-34, 3rd: +7 2003-12-20, NYJ, W 21-16, 3rd: +11 2004-09-09, IND, W 27-24, 3rd: +10 2004-10-24, NYJ, W 13-7, 3rd: +6 2004-12-12, CIN, W 35-28, 3rd: +14 2004-12-20, MIA, L 28-29, 3rd: +4 2005-9-25, PIT, W 23-20, 3rd: -3 2005-10-09, ATL, W 31-28, 3rd: +15 2005-10-30, BUF, W 21-16, 3rd: -3 2005-11-13, MIA, W 23-16, 3rd: +5 2005-11-20, NOR, W 24-17, 3rd: +14 2006-01-01, MIA, L 26-28, 3rd: -5 2006-09-10, BUF, W 19-17, 3rd: -3 2006-09-17, NYJ, W 24-17, 3rd: +10 2006-11-05, IND, L 20-27, 3rd: -7 2006-11-12, NYJ, L 14-17, 3rd: -4 2006-11-26, CHI, W 17-13, 3rd: +7 2006-12-03, DET, W 28-21, 3rd: -5 2006-12-24, JAX, W 24-21, 3rd: +3 2007-11-04, IND, W 24-20, 3rd: -3 2007-11-25, PHI, W 31-28, 3rd: -4 2007-12-03, BAL, W 27-24, 3rd: 0 2007-12-29, NYG, W 38-35, 3rd: -3 2008-09-07, KAN, W 17-10, 3rd: +11 2009-09-14, BUF, W 25-24, 3rd: -7 2009-09-20, NYJ, L 9-16, 3rd: -4 2009-10-04, BAL, W 27-21, 3rd: +10 2009-10-11, DEN, L 17-20, 3rd: +7 2009-11-15, IND, L 34-35, 3rd: +10 2009-12-06, MIA, L 21-22, 3rd: +2 2009-12-20, BUF, W 17-10, 3rd: +14 2010-01-03, HOU, L 27-34, 3rd: +7 2010-10-17, BAL, W 23-20, 3rd: -7 2010-10-24, SDG, W 23-20, 3rd: +17 2010-11-21, IND, W 31-28, 3rd: +14 2010-12-19, GNB, W 34-3, 3rd: -3 2011-09-25, BUF, L 31-34, 3rd: +7 2011-10-16, DAL, W 20-16, 3rd: 0 2011-11-06, NYG, L 20-24, 3rd: -7 2011-12-04, IND, W 31-24, 3rd: +28 2011-12-11, WAS, W 34-27, 3rd: +7 2011-12-24, MIA, W 27-24, 3rd: 0 2012-09-16, ARI, L 18-20, 3rd: -4 2012-09-23, BAL, L 30-31, 3rd: +6 2012-10-14, SEA, L 23-24, 3rd: +10 2012-10-21, NYJ, W 29-26, 3rd: +10 2012-11-11, BUF, W 37-31, 3rd: +7 2012-12-02, MIA, W 23-16, 3rd: +7 2012-12-16, SFO, L 34-41, 3rd: -21 2012-12-23, JAX, W 23-16, 3rd: +3 2013-09-08, BUF, W 23-16, 3rd: -4 2013-09-12, NYJ, W 13-10, 3rd: +3 2013-09-29, ATL, W 30-23, 3rd: +3 2013-10-06, CIN, L 6-13, 3rd: -3 2013-10-13, NOR, W 30-27, 3rd: +3 2013-10-20, NYJ, L 27-30, 3rd: -6 2013-10-27, MIA, W 27-17, 3rd: +3 2013-11-18, CAR, L 20-24, 3rd: -7 2013-11-24, DEN, W 34-31, 3rd: -3 2013-12-01, HOU, W 34-31, 3rd: -3 2013-12-08, CLE, W 27-26, 3rd: -8 2013-12-15, MIA, L 20-24, 3rd: 0 2014-09-21, OAK, W 16-9, 3rd: +1 2014-10-16, NYJ, W 27-25, 3rd: +1 2014-11-30, GNB, L 21-26, 3rd: -12 2014-12-21, NYJ, W 17-16, 3rd: -3 2015-09-10, PIT, W 28-21, 3rd: +10 2015-10-18, IND, W 34-27, 3rd: +6 2015-10-25, NYJ, W 30-23, 3rd: -1 2015-11-15, NYG, W 27-26, 3rd: -6 2015-11-23, BUF, W 20-13, 3rd: +10 2015-11-29, DEN, L 24-30, 3rd: +7 2015-12-06, PHI, L 28-35, 3rd: -14 2015-12-27, NYJ, L 20-26, 3rd: -4
Cbrad, I love what you've done here. But mt prediction is that fineas is going to latch onto some of the games where the 3rd quarter difference was quite larger than the final margins and say "see these examples prove you're wrong"
Yeah well.. I think we're nearing the end of any honest attack on our arguments, aside from playing the Devil's advocate and saying "it could be due to something else". We have a good statistical definition of clutch, many sources of evidence for a league-wide drop-off in various clutch stats, no serious counter-arguments due to sample size, and a plausible psychological/physiological mechanism in the stress response. So I'm also getting to the point where I don't think the debate is being conducted with the goal of coming up with the best possible hypothesis that fits the data. Once that happens I'm out . But I definitely did want to make sure the reported stats are correct.
Actually this brings back somethings I wanted to discuss further. After going through the database the only other QB I could find at Luck's level of winning % in 0-7 point games was Daryle Lamonica with 29-7-10. (73.9% I believe). My problem with treating Luck's record due to something Luck controls is because I couldn't find any stats that show Luck is more than middle of the road when it comes to high pressure stats, yet has a fantastic winning % Since Lamnica was also a noted deep ball thrower too perhaps there may be something about being able to make 20+ yard plays that shifts the needle for winning close games. The other point is Warren Moon. According to the study that puts Flacco very high, they also rated Warren Moon as being one of the worst playoff QBs in history (3-7 record). However Warren Moon's passer rating in playoffs was 84.9 compared to his career average of 80.9. So in Moon's case he personally performed very well but his teams didn't.
The thing is, you've essentially redefined clutch into something that most people don't mean when they say clutch. Saying that someone is clutch, even when playing below their usual level of play, doesn't seem to be what most people mean when they say clutch. Unless you're arguing that good/great QBs are clutch all throughout games, and that why their stats are even better in the first three quarters...but then that makes their "clutchness" even less impressive late in games.
I think the interesting part, is that in trying to prove if "clutch" exists, they've answered the thread's initial question, with basically a "no". I mean if clutch (exists and) is important than Flacco should be the QB everyone wants. He isn't, which seriously undermines the clutch is valuable theory. In the end it still seems like to me "clutch" is the same as "cold".....in that they don't actually exist. Cold is merely the absence of heat. Clutch is merely the absence of choking. And also, I've yet to see anything in this thread account for the fact that unlike any other sport, a QB cannot make a successful play on his own by throwing the ball. It is not possible. Since that isn't possible, its intellectually incomplete to not account for those other players in determining "clutch", wins/losses, success rates, etc.
That's precisely the point I made in the initial post. Using Win% in 0-7 point games as the measure 'clutch' gets you an extra win every two seasons on average. Being super elite, like top 5 or 6 in NFL history gets you to an extra win a season. An extra 5 points on your career passer rating will get you 2 more wins per season on average. I am happy to say 'clutch' is real and measurable. It's far more important to be good than clutch however.
This is nonsense. I've expressed my opinions about what clutch is numerous times and I have backed up pretty much everything I have said with facts and stats. "Clutch" would mean the player legitimately improved his play in pressure conditions, taking into account all other factors that could lead to improvement in stats completely unrelated to stepping it up under pressure. I haven't seen you, cbrad or anyone else provide a single example of that.
I've started going through those 25 close games in which Brady was down at the end of the 3rd to see what actually happened in the 4th and whether, and to what extent, Brady deserves credit for the win. Working backward in time, through the most recent 9-10, its pretty remarkable how much other factors had so much more to do with it than anything Brady did. NYG 2015 -- Won by 54 yd FG with 1 second left. That's a very tough FG that many kickers miss much of the time. Without it, this win doesn't happen. SEA 2015 SB -- Malcolm Butler INT in end zone at end of game. Without it, Brady doesn't win. Brady certainly contributed to the result, but the biggest factor was that play at the end. NYJ 2014 -- Brady threw INT with 5 mins left. NE blocked a FG with 4 mins left. Very lucky to get this win. CLE 2013 -- Pats recover onside kick with 25 second left. Pass interference called against CLE in endzone giving Pats ball at 1 instead of 29 as clock expiring. Unusual circumstances that Brady had nothing to do with. HOU 2013 -- Pats won on 2 53 yd FGs in last 7:16. No Brady or offensive TDs in 4th quarter. DEN 2013 -- Denver muffs punt at own 13 in OT, recovered by Pats. Gostkowski kicks GW 31 yd FG. BUF 2013 -- Defense shuts Bills out in 4th. Pats get 2 FGs in 4th, but no Brady TDs. BAL 2013 -- Balt shutout in 4th. No Brady TDs in 4th. Brady throws INT with 4:17 left but defense gets stop. Brady gets ball back and goes 3 and out. Ravens miss 32 yd FG. I'll get to the earlier games as I can.
As has been pointed out many times now, clutch, like anything else has a range, the absence of choking is clutch, and goes up to playing above their own avg, and just like anything else it's not constant. For every action there is an equal and opposite reaction, if choking exists, by definition, clutch exists, just like cold, or darkness, or white.( I hate to have to tell you this Fin D, but the absence of light, heat or color do exist.) For you guys that don't think clutch exists, go compete in something where something is on the line, and you will find clutch.
There are other "clutch" stats with Luck: 4th quarter rating tends to on average be lower than other quarters, but Luck's 4th quarter rating is higher than his average. Also, ratings for 3rd and 4th down are in general much lower for your average QB, but Luck performs best on those downs. Either way, it's untenable to say Luck had nothing to do with the win/loss record, so that counts as one clutch stat. And of course you have to look at how a QB does across many stats instead of just one.
I'm seriously not trying to get into our typical pissing contest stuff....but you don't understand... ...and this proves it. I didn't say there's no such thing as the absence of those things. In fact, I said the opposite. I said there's no such thing as cold there is only the absence of heat. Understand, I didn't make that up and its not my observation.
Yeah, the definition I'm using is the only one I can justify. You're right many people tend to think of clutch as performing above normal level under pressure situations, and you only find that in rare cases. But it's interesting though to note that the scientific literature also tends to define "clutch" as the opposite of choking, so people who have really thought about this stuff seem to come to similar conclusions.
But doesn't choking mean playing worse than your best in pressure situations? At this point, you've worked this into a way where choke and clutch are the same thing.
There's no way through stats alone to pin down how much of the outcome is due to the QB, but you can make crude estimates like I did using variance explained and YPA (which btw was talked about in this thread). More importantly, it's untenable to suggest the QB isn't an important contributor to the outcome. So while people can disagree on what percent of the outcome is due to the QB, it's just impossible IMO to argue these clutch stats are due to everything else except the QB, and that's really all I need.
You have said it twice now, that there is no such thing as cold, it is the absence of heat, that is the same thing as saying there is no such thing as darkness, it's just the absence of light, or there is no such thing as white, it's just the absence of color, but cold can be measured, can be generated and is tangible, so therefor exists.
Just like "cold" and "hot" are different ends of a spectrum, "choking" and "clutch" are different ends of a spectrum.
Not based on the definition I gave. If average drop-off for an individual is less than the league-wide average drop-off, it's on the "clutch" end of the spectrum. Very precise definition.
Please, just stop for a second. You said: Do you see how I said that the absence of heat does exist? Yes or no? PS, you should stop including white and the absence of color. In terms of light, white is all colors of the visible spectrum, put together.
As resnor pointed out, the numbers show that virtually everyone plays worse than their average level in pressure situations....which is the definition of choke. All you've shown is that some people choke to a lesser degree. That's not clutch. That's your numbers not giving you the result you want (which is definitive proof of clutch) so you've gone back and changed the definition of clutch to be choking at or above league average of choking. All you've done is bend whatever you can to fit the results you wanted.
Definitions get changed all the time, and one reason is when they lead to something that doesn't fit the data. That's btw also true with "hot" and "cold". Once it was found out you can't go below a certain temperature (absolute zero), those definitions changed. So.. we can either invent new words to describe what I'm defining or re-define clutch. I think the best approach is to re-define it because we're talking about differential responses under pressure situations, which intuitively correspond to concepts like "choking" or "clutch".
Which brings us around to my other point, incomplete data. We aren't factoring in anything other than the QB...and not because we can't, but because getting the other data is too hard. Basically, you're saying its perfectly reasonable to change definitions based on results from incomplete data, to fit your already held notions. Well, I'm sorry to tell you, that's neither science nor math....but I guess you can change those too.....
We are factoring in things other than the QB. Already responded to that in post #753. Also.. I didn't have "already held notions". I looked at the data first. And like I said, it's telling the scientific community looks at "clutch" as the absence of "choking". Nothing unscientific about any of this.
