Thoughts on the Dolphins/Ravens Bloodbath

When quarterbacks have had as many pass attempts as these guys we’re talking about, the most parsimonious theory is that they are simply either good or not so good players individually. Situational or environmental factors are very unlikely to be responsible for their performance.

 

Well.. here's a pretty obvious reason why Luck would have far more passing attempts:



Just look at the distribution in point differential for all games Luck played vs. all games Wilson played.

When Wilson loses, the point differential is often quite small while Luck is not only on the receiving end of huge blowout losses but a lot of his wins are close games with small point differential. So it makes a lot of sense that Luck has far more passing attempts per game.

Either way, passing attempts (like rushing attempts) shouldn't be used as an independent variable for measuring QB ability because you don't know causality: the QB helps create the conditions that leads to more rushing or passing attempts late in the game. That is.. Luck is partly responsible for creating the conditions where he needs to play catchup. So none of this is changing the overall narrative that Wilson > Luck.

 

Its relevant data, the problem is now youd need to find a way to find out how much of that is the fault of Luck vs thebteam surrounding him compared to Wilson.

I almost guarantee you Wilson has an advantage in that department due to defense keeping other teams from gaining the lead in the first place.

 

LOL!

This has been like dodge ball. Wait for a new game to start and get back in there. :-) I'll be watching from the bleachers.

 

If you look on the page linked below, you’ll see that Russell Wilson’s passer rating in 2018 was 111.6 when his team was playing from behind, and 119.3 when his team lost by between nine and 16 points.

Why is the simple explanation that the guy is just good discounted so heavily?

https://www.espn.com/nfl/player/splits/_/id/14881/type/nfl/year/2018

 

That's precisely what I did in post #152:
https://www.thephins.com/threads/thoughts-on-the-dolphins-ravens-bloodbath.94575/page-4#post-3198455

Wilson > Luck even after removing the effect of the defense and the difference is pretty big.

 

When removing the effect of defense does it account for field position as well or just points allowed?

Btw I'm not arguing who is better, I'm trying to see how deep the statistics run.

 

Have to look at ratings only if you have 150+ attempts. It's the same critique you can make of DolphinGreg's post #160 where he looked at stats by quarter or by number of passes. Too many of those stats are unreliable because they're less than 150 passing attempts which is a bare minimum. You really need a lot more.

 

Points allowed only. For defense that should be sufficient because the primary effect of field position you're interested in is points allowed.

That is.. you want to know whether you have to score 0 more points, 3 more points or 7 more points to win the game. That's the main effect and trumps issues relating to starting field position which probably won't affect efficiency stats like Y/A, Comp%, TD% and INT% that much given that we're talking about starting at the 20, 30 or 40 on your side of the field (I could see passing TD% being affected if you often started in the red zone but on your own 35 vs 20?).

From a pure statistical point of view, it's problematic to try and adjust passer rating for field position because you'd have to look at passer rating as a function of field position – you have 100 separate passer ratings for each QB, one for point field position. The sample size will be way too small to get anything reliable out of this. So it's best not to go there when talking about passer rating.

The best approach for incorporating field position is something called expected points, which tells you on average how many more (or less) points will be scored NEXT by the offense instead of the opponent. And you can calculate how each play changed that expected points giving you expected points added (EPA). They don't publish raw EPA for QB's as far as I know but that would answer your question, not just for defense (where I think points allowed is sufficient) but for down, distance and time left, etc...

 

Right, but those statistics certainly don’t permit the definitive belief that his performance is suppressed when he faces situational adversity. In fact those statistics, although tentative in nature, suggest just the opposite, that he thrives in those situations.

 

I don't think your interpretation of that information is leading you to the proper conclusion.

 

Russell Wilson has 3,281 career attempts.

 

Exactly, which is why the z-scores I calculated are reliable and show that Wilson > Luck.

However, a lot of the passer ratings you quoted in post #160 are with less than 150 attempts. You listed them yourself: 87, 133, 96, 137.. for the quarter by quarter ratings. And if you look at that 31+ attempts stat you quoted it's with 21 attempts. Those ratings aren't reliable.

 

The main point of that was that you can't divorce the QB from the situations in which he's involved. It's not like there's a separate "team" that's functioning independent of the QB, that creates situations the QB is subjected to without any influence on them.

Even the oft-cited relationship between the offensive line and the QB is not unidirectional. The QB's play influences the play of the offensive line as well. Take a QB for example who is good at executing and getting out to early leads (i.e., Tom Brady). That QB's offensive line is going to have an easier go of it than one associated with a lesser QB, because opposing defenses will have to stay honest versus the run and the pass and won't be able to tee off in the pass rush as easily. That might be one of the main reasons why it's been so difficult for teams to get a good and consistent pass rush on Tom Brady over the years, for example. The Patriots certainly haven't been churning out HoF offensive linemen right and left.

 

LOL, that's because I took 30 seconds and grabbed data from one year (to show it was available).

Why can we not just do it for his whole career to get above 150?

You're saying I'm wrong on the basis I threw out a snapshot of the available data. I take offense at that.

The data is there for anyone to go through. 3,280 / 4 = 820 > 150.

I mean seriously...I feel like you want to win an argument here when I'm going out of my way to try and be polite and explain my position. I think you're purposefully avoiding my points sometimes dude.

 

What you're saying is appreciated by everyone who watches football.

I don't want to demean your point by saying it's common sense per se. But I think what you're saying doesn't really need to be pointed out. We all understand it.

Some QB's make their teammates better. Some hold them back. There's no doubt about that. Peyton Manning made a lot of not-so-awesome TEs look really valuable, LOL.

But when someone can't see that the Colts teams of the Luck-era were pretty bad and actually comes at me with a position that says Andrew Luck was a part of that, I have to laugh.

That's what happens when you don't watch games and just follow the box score. There are people here who will look at passer ratings across divisions and tell you that Ryan Tannehill was the equal of Andrew Luck.

I think the art lies in knowing when to let the stats tell us what to think. However, we should not automatically allow what are often superficial statistical arguments to automatically dictate what we think because they are consistent. That isn't the right way to approach football IMHO. In many cases we have to look at the narrative that an statistical argument supports and ask ourselves whether that narrative seems plausible given other information we have.

 

You said this:
"the notion that Wilson is worse as he passes more (31+ attempts) seems to be true some years and maybe not so true others"

Can't combine across years if that's the claim you want to make. So for the claim you made the data are not reliable. Also.. saying the data aren't reliable isn't the same thing as saying you're wrong. It just means you can't make the claim you made based on the data.

Which points did I avoid? I directly addressed issues about adjusting for defense as well as your "attempts" argument.

 

I appreciate your response, but the "common" sense (i.e., "common" meaning what seems to be implied in many people's posts) seems to be that the QB's individual performance is a function of what's going on around him, rather than his having tremendous influence on what goes on around him. It's as if many people believe the relationship between the QB and the rest of the team is unidirectional -- the rest of the team influences him, and he doesn't influence them.

 

That's an incorrect assumption though.

A played who has that many attempts behind a poor offensive line will likely have worse numbers than someone with that many attempts.

You can NEVER completely factor out the influence of a team around a player. Especially as the players become more similar statistically.

Of course he influences them too. The point isnt that you cant compare QB to one another via statistics.

The point is that as the QB you compare become more statistically similar, the more those influences actually matter in determining the worth of a player.

Who performs better? Tom Brady behind the best offensive line in football or Tom Brady behind the worst? Talent wise? They are the same player but do they have the same stats at the end of the season?

 

As sample size increases you can factor out random variation to a greater degree regardless of the source of that variation. So the way to think about this is NOT that "environmental influence" is being factored out. The environment is always there so you can't factor it out! No, you're factoring out variation in environmental influence.

And the degree to which you can factor out random variation increases with larger sample size. For example, you can factor out variation in opponent strength over long periods of time (no team constantly plays only the strongest teams or only the weakest) to the point that in most cases you can show that opponent strength over a decade or so was close to average.

Same thing with other units. You tend towards average with the majority of environmental influences, with some exceptions of course like Brady always playing with Belichick. But in general, as sample size increases passer rating will reflect "true" QB ability more.

Oh.. and the more similar two players are statistically, the larger the sample size needed to remove the effect of random variation to see whether their "true" abilities differ (quantified by what are called confidence intervals).

 

Pick a quarterback you believe is significantly worse than Tom Brady, and then look at his and Brady’s last five seasons of statistics. Then determine whether you believe the difference in their average level of play over those five seasons, as measured by those statistics, was due to their own individual talent or their surroundings.

 

Well you could also go back and adjust for the quality of defenses faced. Andrew Luck had a pretty big benefit in his first 3 years in the league because the AFC South (roughly 1/3 of his schedule) was epically bad in that time.
In 2012-2014 the NFL average passer rating was 86.9
In 2012-2014 the AFC South, excluding Indy, average passer rating allowed was 91.0.
In 2012-2014 the NFC West, excluding Seattle, average passer rating allowed was 83.0

To my mind that would widen the gap in Wilson’s favor.

 

No we know that all QB's perform worse under pressure once sample size is large enough. Have to look at ratings over multiple seasons because you often have too few passing attempts under pressure in a single season to get reliable data. This is from 2015-2017:



Notice that ALL QB's performed worse under pressure over that 3 year period. Brady's rating was 17.5 points lower under pressure than in a clean pocket so we do have statistical evidence that he would have worse stats behind Miami's OL, which as you notice has a higher percentage of "Under pressure attempts" than NE: 31.5% to 24.85%

 

I 100%

I dont disagree with what you are saying. However there is no rule that says a players supporting cast will greatly fluctuate in talent over time. No matter the sample size like you said, environmental factors cant be eliminated.

FWIW I'm talking about guys on the same tier who may not be separated by much number wise.

Obviously like I said before, you can tell a Peyton Manning from an average QB with numbers alone and no context.

 

When do you get to the point that you're watching a quarterback be inducted into the Hall of Fame and they say something during the broadcast about the exceptionally good statistics he accumulated during his career, and you attribute those statistics predominantly to his own individual ability? When does the threshold get met for "this guy was really good" versus "this guy benefited from what was going on around him"?

 

I dont claim to have that exact answer, but it doesnt mean that threshold doesnt exist. It just means I dont have all the answers to everything.

 

That’s a very good question. I think of someone like Tom Brady, who many would consider the greatest QB of all time. I’m not gonna deny that he’s great but when I look at him vs Manning and I had to start a franchise with one of them in their prime, I’d take Manning every time.

While Brady is great, he benefited from the best head coach/GM of all time who was always able to build a ton of talent around him. He’s had consistently good to great OL and defenses. So much so that when Brady got injured, his team went 10-5 without him. Manning on the other hand often times was running for his life, rarely had good defenses to win more Super Bowls. When Manning was out, his team went 2-14.

Really makes it hard to evaluate an individual in a team sport with so many variables that go into the supporting cast of making someone great.

 

I'd also start my team with Manning because Manning was a OC playing QB making it less important who the coach on the offensive side is. In fact, Peyton in his prime is either my top choice or one of my top choices precisely because of that.

However, the overall evidence for how Belichick does without Brady is a bit more mixed. We have 3 periods where this occurred:
1) the 2000 season where Belichick went 5-11 without Brady continuing into the 0-2 start in 2001 after which Brady helped win the SB.
2) The 2008 season except for game 1 where Belichick went 10-5 without Brady
3) The first 4 games of the 2016 season where Belichick went 3-1 without Brady

Overall that's 18-19 without Brady with better results the longer Belichick has been the HC.

 

Along these lines, can we come up with any history of quarterbacking in the NFL in which anywhere near a majority of QBs played at the average level or worse for a particular team for a sufficiently significant period of time, and then changed teams and played at a significantly above-average level for a sufficiently significant period of time? You can also apply the same scenario to a significant change in quality of play for the same team, perhaps after the team improved the QB's surroundings.

The point here is that if that pattern of career performance is the exception to the rule -- i.e., it happens very rarely -- then the probability with which any one QB will follow that pattern is very low, and the attribution of any one QB's performance should lean heavily toward his individual ability (or the relative lack thereof).

If this historical pattern of QB performance had any significant precedent, personnel men around the league would be paying a premium to dislodge QBs from teams whose surroundings were understood to be responsible for their relatively poorer individual performance, with the purpose of relocating them into surroundings in which they could thrive and realize their potential.

Those kinds of personnel moves are few and far between, however, and I have to think it's because personnel men around the league attribute QBs' play to their individual ability, and not to their surroundings, even though they would presumably know far more about the effects of those surroundings than we.

 

There are obviously some QB's that thrived under a new coach + system after doing real poorly elsewhere (best example is Steve Young.. real bad in Tampa then with Walsh became the most efficient QB in history by z-score rating), but most of the time you don't see any statistically significant differences when a QB goes from one team to the next. The problem with using that as evidence it's the QB and not the environment is that the QB usually doesn't switch teams that often. I think if you see the QB start for maybe 4+ teams with 150+ passing attempts you can start to get the kind of data you're asking for.

Incidentally, our own QB Fitzgerald is maybe the best example, having started for 8 teams!! For 6 of those teams he's had 150+ attempts even after excluding all games where he had fewer than 10 attempts. The statistical test to use is an ANOVA. This is what you get:


All game by game ratings were adjusted to 2018 values and the p-value is 0.2738 which is greater than 0.05 so the null hypothesis that the average rating was the same across all teams is not rejected.

So while for most QB's I think it's hard to tease apart QB vs. environment, I think in Fitzgerald's case I'd be quite confident in saying that the (adjusted) ratings are due primarily to the QB and not the environment.

 

And even in Young's case you had a guy who was second in the Heisman Trophy voting, and went from a team that was 6-10 the year before showed up, as well as 4-11 the year after he left, to a team with perhaps the most innovative pass offense in the history of the league. How often is a QB 1) with that kind of pedigree in individual ability, 2) going to be able to make a jump in surroundings to that degree?

Again, we're talking about low-probability events here.

 

Well you also have to add in that Steve's USFL stats weren’t all that great, although better than Tampa. After 4 years of pro-football he looked like, at best, an average QB.

But because of Steve Young and some other QBs, another example is Jim Plunkett, who turned their careers in new lights there are still many teams willing to take the gamble.

 

Your judging the best dual threat in the game on his right arm without any calculations to your ratings relative to what he does with his legs relative to the competition..

Nice try though DG, but ultimately time wasted.

 

When is this nonsense gonna stop..

There is no rating that includes everything (impact and production using both skills) so arguing about passer rating is stupid actually.

 

Yeah there's no metric that allows you to evaluate the QB's full abilities in isolation as well as his impact on the game.

Still.. the reason it's worth digging deeper into stats like passer rating is because its correlation with win% is a whopping 0.633 over NFL history which translates to 40.07% of the variation in win% explained by this passing stat. Forget calling it a QB stat for a moment and just note how impressive it is that combining only 4 things: passing Y/A, Comp%, passing TD% and INT% explains 40% of win%!

And given that it's pretty obvious that the QB contributes greatly to this stat (even if no one knows exactly how much) it's worth looking at how this passing metric changes when you account for defense or strength of opposition etc.. Or at least I think it's useful lol (obviously!).