Yeah you really need to learn about continuous distributions. I don't think you understand what that is. You said "average" is defined as a rating of 91. Like I said.. the probability of randomly choosing precisely a 91 rating from an infinite number of possible ratings is 0%. So.. here are your probabilities GIVEN your definitions for "average" and "above average": 4 out of 5 "average" AND 1 in 5 "above average" = 0% 3 out of 5 "average" AND 2 in 5 "above average" = 0% 2 out of 5 "average" AND 3 in 5 "above average" = 0% 1 out of 5 "average" AND 4 in 5 "above average" = 0% 5 out of 5 above average = 3.12% And the reason the first 4 are 0% is because it's 0% probability of getting a rating of "average" according to your definition. Anyway, if you don't understand that you need to first read up on probabilities with continuous distributions. There's no point in continuing the discussion until you do that because this is apparently a completely alien concept for you and your intuition doesn't apply at all.
How many QBs have better passer ratings when they are throwing that much? One of our main complaints was that Tannehill, even as a rookie, was being asked to throw way too much. It's harder to have a good QB rating when you're throwing 30+ times vs throwing 20 times.
There's a confound there, in that teams that are behind on the scoreboard have to pass more, and they could be behind at least in part because of the inadequacy of their QB. That said, the correlation between pass attempts and passer rating on a team-by-team basis since 2004 is 0.45, meaning that as teams pass more, their passer ratings increase.
You're combining passer ratings across years. That gives you misleading results because of passer rating inflation. If I combine unadjusted ratings from 1970-2018 the correlation between passing attempts and passer rating is 0.4016, similar to what you're finding. But that makes no sense because teams increase the number of passing attempts when they're behind in the 4th quarter, and the fact better passers tend to be relied on more shouldn't be able to vastly overcome that. When you calculate the average correlation between passing attempts and passer rating across years (so first calculate it for each year separately then average), you get a more accurate result of 0.0732 correlation so basically nothing. You'll get a similar result if you first adjust ratings and then combine. Either way, the number of passing attempts on its own shouldn't be used to categorize QB's.
I'm leaving this thread here in the Dolphins forum because there's an obvious opportunity to evaluate the Dolphins during the Tannehill era seeing play with someone else. But keep it civil.
OK guys.. you got your statistical significance. With a 131.25 rating today Tannehill's set of 7 ratings with Tennessee is statistically significant for BOTH tests I said I'd use. The first test, the t-test for 2 samples with unequal sample sizes tests "Miami" vs. "Tennessee" conditions, so 88 games vs. 7 games while the 2nd test, ANOVA, tests whether at least one year out of his career was statistically significant (the only candidate is 2019). Right now both are. For the t-test the probability is 0.75% Tannehill's 7 ratings are due to random variation alone while for ANOVA it's 3.83%. It will take a really bad set of performances for that t-test to come back negative at the end of the year.. but the ANOVA is still fluid.. could change dramatically based on a game or two, so we'll see. Anyway, for those seeking statistical evidence that Tannehill's play can't be explained by random variation alone, at least for the moment you have it. I'll update this after each game until the end of the season.
I recorded the Fins game so I watched most of the Titans in the first half...but then Fitzpatrick got spunky and I had to catch up to live ball in that game. From what I saw, Tannehill looked sharp completing 8 of his first 10 (with two well placed drops). I missed both 2nd half TD throws live (since they happened 2 minutes apart) but that 2nd pass was beautiful! If we're looking for a difference between the two teams and what RT is doing, I think it's hard to look past Henry and the demand he places on the front 7 staying home. To me he looks like a generational player with that size/speed combo and I think he'd make any QB look a heck of a lot better off play-action. Not trying to take anything away from Tannehill but man, that dude Henry is an absolute beast!
I Agree. It sounds like someone doesn't understand the difference between cause and effect, but think they do understand it. It's like listening to my grandson argue passionately about things he knows nothing about. I find this lack of rationality common among millennial's and Lemmings. All you can do is ignore millennial rants and the Lemming's juice and their droppings.
Whether passer rating is a cause or an effect (or as the other fellow called it "a result") is immaterial with regard to whether it can be used as a predictor of winning, which was the context of the post you quoted. The temperature outside is certainly the effect (or the "result") of a whole host of atmospheric factors, but we can nonetheless use it to predict many things, including what you might pull out of your closet to wear this morning. So, ignore if you will, but you might miss out on learning something. Or do you know everything already, like your grandson?
This is a fun game. Debate something for a week. Two weeks later you start the same discourse, while pretending we haven't been down this road several times.
When I moved to North Carolina I saw an interesting bumper sticker on the car in front of me. It said "I may be slow, but I'm ahead of you". I feel it describes our situation in this thread, except I'm the guy with the bumper sticker and your "the guy" reading it! LOL
I hope cbrad will weigh in on this because he’s far more knowledgeable than I am about these matters, but the last time I looked at that, it was a little more than 5 passer rating points for every one win or loss added to or subtracted from the average 8-8 record for the season. So based on passer rating alone, ignoring hypothetically all of the other team variables that would be involved, a team would be expected to have a regular season record of 6-8 if its quarterback’s passer rating was around 80-82, which is about 10 points lower than the current league average. Now, what I could be mistakenly recalling is the figure for passer rating differential, rather than just offenive passer rating, and again hopefully cbrad will correct me on that if so.
So we should be able to agree that currently, there is a gigantic unlikely event that has been occurring all season long in NE.
No, because again, it’s not only a team’s passer rating, but the passer rating it surrenders to other teams (i.e., passer rating differential), and that’s where New England, prior to yesterday, had been stellar all season. Obviously if you keep opposing teams from doing well with regard to something that’s strongly predictive of winning (offensive passer rating), you stand a good chance of winning yourself.
So you're arguing that is common for a QB with a below average rating to have a better rating than the opponent? Since we already know what the projected odds are for a below average rating to win, then it can't also be common that below average passer ratings result in a winning differential most of the time.
Apparently I misunderstood your question. You’re right that what New England has done wouldn’t be common or expected. Their pass defense in terms of passer rating surrendered (again prior to yesterday) has been off the charts good. I believe before yesterday it was in the low 50s. But again, that nonetheless shows us the power of passer rating differential in predicting winning. If you can obliterate other teams’ passer ratings, you can win almost every game you play, even though your own passer rating is no better than average.
Aww crap, the Patriots lost...just saw that! That hurts twice- once to our draft pick with Texas, and again with Tennessee taking the division. The one time in YEARS I've actually wanted the Pats to win and they choked!
But you can’t predict that. LOL. Rating and rating differential are results...and if your QB is averaging below league average, the most likely occurrence will be a loss.
Using ONLY passer rating, this is the graph you want to go by after adjusting ratings to 2018: The equation suggests the best way to remember this is: 1 extra passer rating point = 1 percent increase in expected win%. So over a 16 game season 1 extra win = 100/16 = 6.25 extra passer rating points. NE's (and Brady's) current passer rating in 2019 is 88.2, which becomes 88.2*1.021 = 90.05 in 2018 numbers. The expected win% for 90.05 in 2018 using that equation is 46.9% which equals almost exactly 7.5 expected wins in a 16 game season. NE's win% as of right now is 83.33% and you can see in that graph that that's approximately the best any team has done with that kind of offensive passer rating (technically, the best is 13 wins = 81.25%). So what NE is doing is at the absolute tail end of the historical distribution.. so far. More important for me: this is the FIRST year Brady may end up below average in rating! That has to be celebrated because it's never happened before no matter his surrounding cast and it might suggest he's declining (not statistically significant though).
btw.. since people have been talking about predicting games, it doesn't matter what stat you use you can't predict them very well. The best you can do is predict the outcome to 59-60% accuracy using either win% or point differential, and the best predictive power is seen using either win% or point differential based on the last 10-11 games, not fewer and not more: I haven't done this using passer rating (I actually can't since the game-by-game database I have doesn't have that) but given that only about 40% of the variance in win% is explained by offensive passer rating and about 63.2% of variance in win% is explained by passer rating differential, those won't do as well as win% or point differential. Point is.. no matter what stat you use, the best you can do is predict about 6 out of 10 games (maybe some combination of stats might get it slightly above but I bet not by that much) so it's mostly unpredictable. Note from that graph something interesting though: to predict win% based on performance over multiple seasons, point differential is better than win%.
Do you believe that Brady is doing things different this year? What I see is he is missing Gronk as a seam threat. Historically, Brady throws much lower ratings without Gronk. Last night they doubled Edelman, and the other receivers were useless.
That’s true, but in saying that (quoted above), you’re using only the predictor of offensive passer rating and not passer rating differential. Passer rating differential is comprised of not only offensive passer rating but also the passer rating surrendered by the defense.
Again. What is the likelihood that if your QB is throwing for below average, that the opposing QB is throwing even further below average? ESPECIALLY since the Patriots are winning...passer rating almost always favors the winning QB, because successful plays are what win games. So if your offense is not having successful plays, but they're still winning, then it's showing why using passer rating is not an accurate predictor. It's incredibly unlikely that a QB is going to be below average and yet still be significantly better than three opposing QB.
I think if you were to combine passer rating differential with turnover differential, removing of course the overlap involved in interception differential, you might come up with a much higher number, because I suspect you would be combining the strongest systematic predictor (other than points) with the strongest random one.
Haven't watched NE games this year except that Ravens game so I don't know. But Brady's been without good receivers multiple times in his career, and I don't think anyone thought this was going to be his first year with below average ratings.
That also begs the question...I've been told for years that one example of Brady's greatness was being successful with garbage. But I believe that Brady, while definitely having no name receivers, did not have garbage. He had receivers that excelled in that system. This year he is missing Gronk...just like he always has when Gronk was injured.
What you’re essentially saying there is that you believe offensive and defensive passer rating are strongly correlated on a season-by-season basis, but that isn’t the case. In the 160 seasons played by NFL quarterbacks between 2014 and 2018, that correlation is a mere -0.09, meaning that those variables are almost completely independent of each other. So, having an average offensive passer rating while completely dominating other teams defensively in terms of passer rating surrendered isn’t an unexpected result, at least in terms of the relationship between those two variables.
NE prefers precise route runners that can juke you out of your shoes...but probably lose in a foot race. That's why folks like 7-11, Edelman, Welker, etc. have thrived in that offense even though they didn't last on other teams. That definitely doesn't make them "garbage" though...it simply means Brady needs receivers to get open quickly. Chunk yards have never been a priority in NE and it's obviously served them well. But you are right- without Gronk (their only real chunk-yard threat), it is easier to scheme against them and slow that offense down. This season they're playing Brady like teams played Tannehill a few years back....they give him the short stuff and dare him to beat you deep.