So, you'd agree that Brady and Rodgers aren't "clutch" QBs then, right? As to my golf example, the novice really has no bearing on my "clutch" level. It doesn't matter how stable his play is, what matters is that I pulled out the win, on the last hole.
Holy crap! Are you like being deliberately obtuse? How many times in just the last few pages (much less over who knows how many times we've debated this) has it been said that on average the drop-off in performance for most HoF (or future HoF) QB's from normal to pressure situations is less than average drop-off? In many key stats they have less drop-off, in other key stats they have normal. And in a few stats (some weak examples like overall average to "tied" situations) it's worse than normal. On average, they are much better than average. You yourself used this "generally" it's true argument when I pointed out you were wrong about Peyton Manning. Now you have selective amnesia? If you don't want to give a contrary argument a fair shake so be it, but don't act like arguments that are printed multiple times in black and white were never made!
What if when we played I only won the holes where we had large stakes on it and those were the only holes you lost? However, you still shot a better score than me? That's my point. In those situations, human nature take over. There are just moments in life where people get more tense, things seem more challenging, etc. etc. Some people have a way of handling themselves better than others in that situation. Others have a way of staying the same in that situation while others falter. However, those moments are ABSOLUTELY affected by human nature. Again, hitting in the bottom of the 9th in the WS with the GW run on base is much different scenario than that same situation in June. The person may have the same outcome. They may not produce like they do in May, etc., but there is absolutely a difference of feeling to that batter when he goes up to the plate. How he responds in that situation is important quality to have.
I agree with you that there is a difference. Where we may part ways is that I believe that playing in those situations enough times makes the situation become normal, thereby removing the extra pressure that those situations can bring.
I agree with that. If you play in enough postseasons, Lebron James, Derek Jeter, etc. you become normalized to that type of situation. It doesn't mean you don't get nervous/amped/anxious, etc., but you're much more apt to handle that situation than someone else who hasn't been in that situation. Again, I think there is something to being able to handle defeat/failure period, which seeps into this "clutch" idea. I think it's important for an athlete to really be comfortable with that.
Now, while I agree there are people who choke, i.e. consistently play worse under pressure, I think that good to great players maintain their play, even someone like Brady who sees a greater drop in play, still plays much better than an average QB under pressure. He then gets called "clutch," but he's just keeping a high level of play, which is his norm.
Yup, you're just repeating falsehoods now. Brady on average has a smaller than average drop-off in performance under pressure.
WHAT?? You already agreed that he drops 13% vs 11% for the average QB. So his drop is greater. However, his rating after the drop is still better than the average QB.
Unbelievable. I'm not going to rewrite everything dude. Maybe just read the posts? Both Fineas and I agreed that drop was around average (technically slightly below) because that's what you'd conclude statistically. But his <4 trailing vs. <4 minute tied+ahead is much better than average. Same with Brady being ahead 50% of the time at the end of the 3rd quarter in games that end with 0-7 score differential, yet ending up winning 69% of them. He is way better in game winning drives and 4th quarter comebacks (some of these stats overlap of course), etc... Seriously.. please read the posts first (and not just Fineas's.. I know you like them but my responses nullify a lot of what seem to otherwise be good arguments). Brady on average has a smaller drop-off in pressure situations. That's a fact, so please start with that.
Well, part of this debate is whether playoffs really are a separate condition. Some seem incredulous at the suggestion it is not, but the general similarity between playoff stats and regular season stats across all sports (at least football, baseball and basketball -- I don't know a lot about hockey or soccer) suggests that it is not a meaningfully different condition. As to the NFC South numbers, your point about it possibly being due to specific matchups is similarly true of Flacco's playoff stats. For that 10 game stretch, it was in a 4 year stretch. Of those 10 games, he played NE 3 times and the Steelers twice. There may have been specific things about the matchups with several of the teams he played in those 10 games that made them good matchups for him. Or maybe some of those teams were missing key defensive players for those games. Etc. The fact that it was the playoffs is not the only variable or factor that could have affected his play and production.
You don't look at similarity of stats FIRST to determine whether you can define two sets of data to come from different conditions. You have to be able to independently identify different conditions. Clearly, you can do that with "playoffs" because those games are in tournament format vs. round robin. Whether or how that affects players is a question you ask after identifying different conditions. And regarding the NFC South vs. Flacco playoff argument, I'm not saying you couldn't theoretically find ulterior reasons for explaining away Flacco's statistical improvement in his playoff numbers. All I'm saying is that given the data we have now, there's only a 0.21% chance Flacco didn't actually improve his playoff performance.
That's not fact nor have we agreed to it. His drop from regular season to playoffs appears to be pretty average. His drop from 1st 56 minutes to last 4 minutes when tied appears to be below average, albeit on a small sample size. As I've stated, the drop you seem to be referring to -- from leading with less than 4 to trailing with less than 4 -- is largely irrelevant. There is much more pressure on a QB leading by 1, 3, 4, 7 or even 10, then there is on one down by 17, 21, 28 or 35. I just don't think that is a meaningful measure.
Sure it's fact. I'll ask once again. Where are the comparable "clutch" stats to the ones I and others have listed (just take those from the post you quoted) where Brady is well below average? He's either around average or above average in key clutch stats. He's below average in stats that aren't so important.
I'm not looking at the stats to determine if playoffs are a different condition. It has been my belief all along that it is not and the facts IMO support that. One can find distinctions between virtually any sets of data, but that doesn't make them meaningful. For example, there is a theoretical distinction between "Games played in the first week of a month" and "Games played in the last week of a month" but that doesn't make it a meaningful one. Same for "Games against teams with blue jerseys" vs. "Games against teams with red jerseys". Etc.
Yes, I was just rereading, as cbrad requested, and I don't see the two of you in agreement. Cbrad wants to include 4th quarter under 4 min with a lead in the stats, which gives less of a drop for Brady.
Fine, don't say playoffs isn't a different condition. Just try to argue it's not a meaningful one. But to suggest that taking any sequence of 10 games is no different than taking all playoff games is not acceptable. First case is cherry picking. Second case you can identify external conditions that are different.
Well if you read carefully you wouldn't have written the posts you wrote. Fineas thinks the <4 minute trailing vs. < 4 minute tied/ahead is irrelevant. Of course it's not because you've conditioned on whatever <4 minute left does, so if anything it's more relevant than comparing to overall average. Either way both are acceptable.
Now you are moving the ball. Your previous statement was "Brady on average has a smaller drop-off in pressure situations. That's a fact, so please start with that." Now you are insisting that I have to prove that he is "well below average?" I'm not claiming that he is -- I am just saying it is not a fact that he has a smaller dropoff than others on some such stats. Like the tied with less than 4 stat. I suppose one could probably argue that his dropoff is greater in Conference Championship games, which would seem to qualify as pressure situations, right? His career rating is 96.4, but in conference championship games he drops more than 20 points to 76.3. I don't know what the average drop (if there even is one) in conference championship games is, but I'd be willing to bet a substantial sum that it is far less than 20 rating points. It might also be worth looking at his playoff stats on the road (non-neutral site). His last 5 such playoff games he posted ratings of 74, 57.6, 79.5, 93.9 and 56.4 and the Pats went 1-4 in those games. To the extent his playoff dropoff looks smaller than other QBs (and I don't agree that it is), it might be at least partially due to the fact that he is usually playing at home in the playoffs. When he's been on the road, he generally hasn't been that good and usually doesn't win.
You are missing the point entirely. The 10 game sequence in 2010 is random, or at least appears to be. Who knows, maybe he was eating Wheaties before each of those games, but switched to Capt'n Crunch thereafter. But it happened and seems to be random. And it proves the point that anomalous runs or apparent patterns of 10 games or so can be totally random. As can patterns like the NFC South numbers. Or the retractable roof numbers. Apparent correlations aren't necessarily real and even when they are, they don't mean or imply causation.
I gave you credit for finding some stats where he's below average, but none of those are arguably as important as the <4 minute trailing, 0-7 games record (while being ahead 50% at the end of the 3rd quarter), GWD, 4th quarter comebacks, etc.. There are many key stats where he's average. But what balances out the key ones where he's above average? And of course there are many other less important stats where he's above average. Put it together and yes on average his drop-off is less (weighted by relevance of the stat obviously). I never argued anything based on Flacco's 10 game stretch. In fact, I'm the one who's been consistently saying don't choose such a 10 game sequence anywhere, playoffs or not because it's cherry picking. Only thing I argued with Flacco was the 0.21% probability he didn't improve in the playoffs, and him not having the usual drop-off in passer rating in the playoffs.
Well.. if you're going to dismiss what I think are some of the best clutch stats we have as not valid, then there's no reason to debate further. However, the sample size argument once again won't hold because I took that into account. That's what the data suggest. But as I said before no way to pin down what's really going on.
Yes, he's showing why the "best clutch stats" are really not valid, no matter how many times you say they are.
My logic (explained before) is this, and tell me if you disagree: for any given score differential, one can argue there is less pressure when that differential is positive than negative because you NEED to score with a negative differential. That is, would you agree that in ANY situation, being ahead +3 means there is less pressure than being behind -3? Replace 3 with any number and I think that always remains true. So there's no justification for excluding ahead data. If anything, you should exclude tied data because that could fit into both categories.
He's not showing anything. He's just asserting they aren't valid. Look, if we could actually directly measure pressure, we could settle this issue. All I need as proof is to show there is more pressure on average in <4 minute trailing situations than in <4 minute tied/ahead situations. I'd think most people would agree that's true. He says no, but there's no proof or evidence he's correct. So it's a difference of opinion for now.
If you're trying to prove "clutch," why would you include stats for situations that you're admitting aren't pressure situations? That's why Fineas is arguing that you can't simply throw tied/ ahead vs trailing. I also think that tied has an enormous amount of pressure with it.
Sidenote, why mesh tied and ahead together? Would make more sense to me to look at 4 min ahead, 4 min tied, and 4 min trailing.
As a control. You want to compare in two situations: more pressure vs. less pressure. That's why trailing vs. tied/ahead makes so much sense if you agree there's on average more pressure in trailing situations. No problem with that. But to argue we shouldn't use ahead is IMO completely unjustifiable.
If you want to look at pressure in only two situations, I think you should either look only at ahead vs trailing, or look at ahead vs tied/trailing. Tied and trailing each have more pressure than being ahead.
Probably minor quibble, but the bold is not true. If you're behind 3 or more scores (for example) with a small amount of time remaining, chances are the pressure is off as it's generally pretty impossible to make that comeback. Also, you could have the lead and lose it with not enough time to recover.
It doesn't matter how you slice it as long as you can argue one condition has more pressure than the other. Also, Brady's stats are more impressive if you compare ahead vs. tied/trailing or ahead vs. trailing. League average for 2015 becomes 104.1 for ahead vs. 78 for combined tied/trailing. Compare that to Brady's career 88.3 for ahead vs. 82.75 tied/trailing. You have a difference of 26 for the league and only 5.5 for Brady. For ahead vs. trailing it's 104.1 vs. 78.5 compared to 88.3 vs. 83.7. That's essentially the same, showing the tied stats don't budge things much, at least with Brady.
Just replace "less pressure" in the sentence before the one you bolded with "less than or equal pressure" and it should take your objection into account. Being behind by 21 points with almost no time left or ahead by 21 points probably means there's no pressure either way.
I don't see how one "improves" at something and them immediately reverts back to being who he was before such "improvement." What Flacco does is play QB. If he improved at it at some point in the playoffs, that should carry over to the regular season. He should stay better. He should be better in those clutch moments. But that's not what we see with Flacco. He was the exact same guy the last 2 seasons as he was before he won the Super Bowl MVP. It appears that he was actually notably "unclutch" as he seems to have repeatedly failed to come through in the clutch last season. And as discussed earlier, there is no indication Flacco ever did anything to explain his "apparent" improvement in the playoffs. He went from awful to great instantaneously. I joked about the clutchness fairy sprinkling fairy dust, but nobody has offered any better explanation. When Flacco had that 9 game stretch in 2010, it wasn't a sign of improvement. If it was it would have been sustained. It was just a random stretch of good or hot play. The Heat's Josh Richardson was a .318 3 pt shooter in 4 years of college. Up to March 4 of his rookie NBA season, he was shooting .300 from 3. And then he went on a tear, shooting an incredible .584 over a 16 game stretch from March 4 to April 5. That 16 game stretch is 60% bigger than Flacco's playoff streak. And if one stopped on April 5 and looked at his play over that stretch, one using your logic would have concluded that Josh Richardson had improved as a 3 pt shooter to such a degree that he had become, by far, the greatest shooter in NBA history. His 3 pt shooting in that stretch makes Steph Curry and Klay Thompson look as if they can hardly shoot. If you ran the same analysis you ran for Flacco's playoff games for Richardson using his prior college 3 pt shooting numbers plus his first 4 months of his NBA career through to April 5, your probability analysis will say that the chances of his shooting like that being random chance are infinitesimal. From that, you'd presumably be saying it must be due to improvement as a 3 pt shooter. And he may have improved some, but certainly not that much. And that became apparent when he then shot .327 from 3 from April 5 to the end of the playoffs. In other words, he became pretty much the same guy he was in college and in the first 4 months of his NBA career. It's pretty clear to me that the vast, vast majority of that run of great 3 pt shooting had nothing to do with improvement -- he just had a hot streak. Like Flacco had in the playoffs (or if you don't like that because those aren't consecutive games, like Flacco had against the NFC South). If Josh Richardson returns to being a 58% 3 pt shooter for the remainder of his career (or even a season), I will certainly reconsider the Flacco issue. But I'm pretty confident I won't need to do that.
It absolutely matters. Let's say QB A plays to a 100 rating when ahead, an 80 rating when tied, and a 75 rating when trailing. If you average 100 and 80 together you get 90. So, you would have a 15 point drop when comparing to trailing. Now, if average tied and trailing, you get 77. Now you have a 23 point difference when comparing the two. Which is more accurate? I think it's more accurate to say that the QB plays at a significantly lower level when trailing or tied. When you do it your way, it masks, or can mask, the level of play when tied, since it's tired to level of play when ahead, and most QBs play their best with a lead. Why do you have two different ahead ratings for Brady? Ahead vs tied/trailing Ahead vs trailing The ahead figure should be the same.
No, I don't agree with that. People respond to pressure differently. When down, some feel like the pressure if off, they aren't supposed to win, but if they do they are a hero and they let it loose as if they have nothing to lose. Others tighten up when ahead and may be saying "just don't F up," which sometimes leads to them doing just the opposite. I don't think either of those factors are much at issue with professional athletes. By the time they get there, they have demonstrated that they are the elite of the elite and have played in all kinds of pressure situations, whether that be state championship games in high school, conference championship or rivalry games in college, or the NFL playoffs. There simply don't seem to be consistent, repeat chokers at that level, nor does there appear to be people who regularly and consistently play above their normal ability in those situations. Almost everyone has a mix. As in non-pressure situations, great players have a lot of great games, a few good ones, a few average ones and a couple of stinkers. Role players, or non-great players, will typically have much fewer great games, more good to average and more stinkers. Playoffs and what you call "clutch" situations are inevitably smallish sample sizes, especially in football, so our perception of a player's clutchness often depends on how many of his great games show up in those situations and how many of his stinkers do. When a game is tied with less than 4 minutes less, we know the pressure is on. When a guy is up or down by 14+ points there just isn't much pressure at all.
Like I said, I don't know. I just follow the data. However, I think IF (just try this for once) you assume playoffs truly represent a different condition, then it explains why you don't see the carry-over into the regular season. You look at his regular season passer ratings (game-by-game) over his entire career and it's been real stable. Oh, and the abrupt increase in playoff ratings I'd argue is untenable only if those were games that occurred like right after each other. Such increases aren't uncommon in developmental biology/psychology, for example when you see how children learn things. In many cases you see that suddenly they get a concept they couldn't get for a long time. The typical learning curves (called acquisition or extinction curves) where the differential is greatest early on and then asymptotes are curves that describe learning across consecutive periods (e.g. trials, or blocks of trials, etc..)
You don't average ratings like that. You take total number of completions, attempts, yards, TD's, INT's etc.. and plug them into the passer rating formula. Otherwise sample size bias will play havoc with you (e.g. if you had a small number of attempts in "tied" but large in "ahead", it makes no sense to take the average). Oh, and the figure for "ahead" is the same in both conditions: 88.3. You even quoted that.
The general pattern of how performance changes as a function of pressure is generally pretty stable across people. It goes up with increased pressure up to a point after which it falls. Where each person is on that curve changes for different situations, but the general pattern is very predictable. Also, while it's true pressure changes as a function of point differential, the comparison of up by X vs. down by X shows that overall pressure should be greater in "trailing" vs. "tied/ahead".
No. Not only is it not true that there is more pressure when one is trailing, it is often the condition most conducive to piling up impressive, but meaningless, passing stats. When down by 14+ with 4 minutes remaining, there is no real pressure. If you comeback and win you are a hero, but if you don't win it won't be because of what you did in those last 4 minutes. It will be because of what happened in the first 56. That QB who is down 14+ is probably going no-huddle against a prevent or shell defense. They'll let him complete 8 yard pass after 8 yard pass as long as the clock keeps running. If that QB throws 10 4 yard passes with an extra 4 yards of RAC and maybe gets a gimme TD in the final seconds, that 10-10 for 80 yards and a TD stat line is a helluva passer rating booster. He'd have a 133.33 rating for those last 4 minutes. And it would have accomplished nothing. But people looking at those stats from your perspective would presumably conclude that he is clutch.
