Separate names with a comma.
Discussion in 'Miami Dolphins Forum' started by bbqpitlover, Oct 16, 2019.
It’s not accurate according to your definition. According to mine, it is.
All of that is exactly what I’ve been exploring throughout this thread. There is a strong relationship between Henry’s efficiency and the percentage of the total offensive workload allocated to Tannehill, as well as between Henry’s efficiency and Tannehill’s passer rating.
Obviously however if you use number of pass attempts and not percentage of the offensive workload in those calculations, you fail to control for the variation game to game in total number of offensive plays. You get a misrepresentation of the amount of run-pass balance involved from game to game.
The Titans for example just ran 78 offensive plays Monday night, whereas the average number of single-game offensive plays in in the league in 2019 was 63. Using numbers of pass attempts does nothing to consider that variation, in terms of run-pass balance.
Right, but your definition does not agree with the majority. For example, this article pretty clearly puts high volume at 40 attempts and above:
Or this one, where they are looking at passing volume vs efficiency, looking at the argument that QBs who throw less often have better efficiency.
So you're free to have whatever definition you want. However, it's incredibly narcissistic to believe that your definition is the correct one because you came up with it. Getting annoyed because people don't agree with your made up definition is silly.
I didn’t say anything about my definition’s being the correct one. I simply said it’s mine and that I’ve used it and measured it consistently throughout the thread. There has been no re-definition of anything as you suggested above. I also suggested you do some research based on your definition. To each his own.
People have been disagreeing with your definition all throughout the thread.
And that’s fine. Ask them how they will determine run-pass balance from game to game when there is variation in the total number of offensive plays from game to game. See what their response is.
You're conflating those. Volume doesn't necessarily have anything to do with balance. Balance is simply the percents of run/pass. Volume is the number of times each one occurred, regardless of percent of total. Hypothetical, unrealistic game to illustrate, a team runs 200 plays, and the QB dropped back on 40 of them, and was never sacked. Your definition says that's not a high volume game. But it's still high volume for the QB, as he's throwing a ton.
I see what you’re saying, but I would consider the concept I’ve illustrated here to be a better representation of the degree to which the quarterback experiences himself as carrying the offensive load.
If there are 200 offensive plays and the team is choosing to feature the running back on 160 of them, obviously the quarterback is going to experience himself as playing second fiddle to the running back. I don’t think he would come out of such a game saying to himself “man, I just passed the ball a ton.” The running back on the other hand is going to be looking for the nearest ER.
Well, I think he'd still feel like he threw a ton, even though the back is looking for the ER. Lol
Now, he probably wouldn't feel as though he was working the hardest, but it would still be high volume.
The thing is that the way he is using volume would be correct in a scientific sense in a lot of situations.
That's just not what it means in football though. Literally noone goes "wow, 65% of the plays were passes, high volume" if you threw the ball 20 times.
But they should, because that game must’ve unfolded in such a way that it was placed in the quarterback’s hands to a good bit greater degree than average in the league. That team had a good bit less offensive balance in that game than the norm in the league.
Lots of the people in this thread have touted the value of offensive balance, but apparently nobody wants to actually measure it.
Pretty good for some people thats considering his performance vs Broncos "okay"
This volume argument is pretty silly. Look, this is the bottom line...for anyone with COMMON SENSE...
Every NFL team has a quarterback and either a premier running back, or running back by committee. If you’re quarterback throws the ball twice as much as your premier running back or twice as much as your committee of running backs, that’s a high volume passing game. Everything is nonsense.
According to my statistics class I took many many moons ago, a chess game has millions of possible opening moves. I play chess and had to call my professor out. I had to correct him and let him know there’s only 20 possible opening moves.
Stats people have a difficult time seeing the trees for the forest
It certainly would, because that would indicate that the percentage of pass dropbacks was approximately 67%, to roughly 33% running plays. However, that’s not what happened Monday night, where the Titans had 78 total offensive plays and 44 of them were pass dropbacks. That percentage of pass dropbacks is 56.4%, which is below the league average of 58.8%.
You’re the guy who touts offensive balance as supremely important. All we’re doing here is measuring it with precision.
No Interceptions? That held the game in check. Tom had a little trouble in Tampa those turnovers are killer.
No one cares if you measure the balance. It's you're application that it means low volume. Literally no one has a problem with measuring the workload.
Is there a real game you can point to where the distinction between volume and the concept I have used has any meaning?
It's worse than just one game. The graph I showed you plotting correlations between passing attempts and passer rating (per QB season) from 1978-2019 demonstrated that your "low volume" hypothesis was incorrect with Tannehill. That is, Tannehill might be slightly worse than average, but he still straddles average.
However, your percent workload hypothesis is still alive, though as I pointed out "percent workload" is probably more an effect of high rushing Y/C than a cause of high passer rating (it's hard to argue percent workload is the cause because that would imply the coach could cause an increase in passer rating just by lowering percent workload).
In other words, this isn't just semantics. It's the result of your "low volume" hypothesis not working out. And you can't just say this is what you've been arguing the whole time because you actually did start with attempts. What you're arguing now has to be considered a separate argument, and it's best to be precise about that by saying "percent workload".
All I know is that in 2019, 1) the Titans had a run-pass balance that was virtually two standard deviations below league average in being imbalanced toward the run game, 2) there was a -0.64 correlation between Derrick Henry's YPC and the Tannehill's percent workload (the better Henry ran, the less Tannehill dropped back to pass), 3) there was a 0.64 correlation between Derrick Henry's YPC and Tannehill's passer rating, which was more than two standard deviations above league average, and 4) there was no relationship between Tannehill's YPA and the Titans' run-pass balance.
So, the relationship between Tannehill and Henry was noteworthy in several ways, and when we're talking about a QB who had a great year "out of nowhere," that possible situational advantage has to serve as a hypothesis to explain his performance while being something to monitor going forward.
In essence, Tannehill may have been experiencing a very low degree of difficulty in 2019.
The only point relevant to that regarding the Titans' first game this season is that the game wasn't as distinctive as it may have appeared from the Titans' run-pass balance in 2019. Tannehill's 43 pass attempts may have appeared to be "high-volume," though within the context of 78 total offensive plays, the run-pass balance involved was nonetheless slightly below the league average of 58.8% plays involving pass dropbacks.
In other words, Tannehill didn't go out there and play a game like Patrick Mahomes or Drew Brees regularly play, in terms of run-pass balance. The Titans were still slightly imbalanced toward the run (compared to the league norm) as they usually are, despite how "43 pass attempts" appears in print. Tannehill's degree of difficulty, at least in terms of "percent workload" as we've begun to call it, didn't vault upward beyond the league norm.
Moreover, Tannehill's passer rating was 97.9, while Derrick Henry's YPC was 3.7, and so the correlation between those variables, across the 15 games Tannehill has played for the Titans, remains an extremely strong (compared to the league norm) 0.65. The better Derrick Henry runs, the better Tannehill plays. The worse Henry runs, the worse Tannehill plays.
Yeah, definitely compelling stats. #1 suggests the Titans thought Henry gave them a greater advantage than most RB's, #2 and #4 suggest they didn't think Tannehill would provide as much of an extra advantage as Henry (otherwise they would have relied on him more when he did well), while #3 suggests there was some kind of synergy between Henry and Tannehill that went well beyond what is normal for RBs and QBs.
One thing those stats don't show is why Henry had higher Y/C than with Mariota. That's where Tannehill's influence does come in. But those stats definitely suggest that coaches thought they'd gain a greater advantage relying on Henry when he did well than on Tannehill when he did well.
Yeah, such stats help explain a decent portion, but nothing adds up to #1 in passer rating, at least not using "expected effects". It's like that 3-13 to 13-3 Manning record change where you can't explain ~3 wins through stats alone. I fully expect 2019 to be an anomaly even if Tannehill starts performing at a sustained above average level from here on out. #1 in passer rating is just REALLY hard (for everyone except Steve Young I guess).
Don't forget Denver bottled Henry up last year too, so this might not be the best game to base predictions on. Next two games are against the Jags and Minnesota, which should be a boon for the Titans offense. But then they face Pittsburgh, Buffalo and Houston before the bye. I think we'll have a good idea of where Tannehill is by that time.
Here's an interesting concept to apply with regard to the "degree of difficulty" proposition I've made regarding Tannehill's performance in 2019.
During the Titans/Broncos game the other night, of the 49 plays in which Tannehill was involved as a passer or a runner, seven of them (14%) involved a negative expected points value. The expected points per play for those plays was -0.53.
It would be interesting to compare the above two values (14% of plays and -0.53 expected points per play) to league norms, to determine the degree to which QBs are being asked to function "in a hole they need to dig their teams out of" so to speak.
And then of course it would be interesting to explore differences among QBs in how successful they are in digging their teams out of those holes. You'd easily have a measure of degree of difficulty and success as a function of it.
You want to use win probability as the independent variable, not expected points, since WP takes into account current score and time remaining while EP does not. The dependent variable would be win probability added.
However, you can't directly compare WPA as a function WP for two QB's without knowing "Expected Win Probability Added" (EWPA) which isn't something anyone reports. That is, EWPA almost certainly depends on WP, and the same is true with Expected EPA (EEPA) as a function of EP. You'd then want a z-score for each value of WPA as a function of WP and then report the average z-score per QB. One can definitely do this with play by play data, but it's a lot of work to set up the program.
I already showed him that he was all screwed up on any concerns over Tannehill in games that are truly high volume (e.g. threw a lot of passes) last season. In every case, the passing volume was lower in the first half but the offense was not working. In the second half of each of those game, they threw the ball more and his passer rating was significantly higher than the lower volume first half, every time.
Look up CPOE......... and YPA....... Tell me what they are and how they relate to what a QB has to do.
This is exactly what I said all of last year. I also showed that there has only been one time in the history of the league that a QB had consecutive seasons with a passer rating about 115. And in the second season, he didn't lead the league in passer rating....... so yeah, #1 is REALLY hard.
The problem with CPOE, as I've stated earlier here, is that Tannehill's performance with regard to it throughout his career suggests he benefited from situational advantages in 2019 rather than his individual ability.
If his individual ability was responsible for his performance with regard to CPOE in 2019, then his performance in Miami would've followed the same pattern with regard to CPOE, albeit with a lower expected completion percentage owing to poorer surroundings. His completion percentage over expectation in Miami would've been just as large, owing to his individual ability. He essentially would've been surmounting poorer surroundings in Miami with his ability.
Again however, that isn't what we see. What we see is that 2019 was markedly different in that regard, which suggests he was experiencing something specific to 2019 with regard to CPOE, and not something specific to his individual ability.
Jim Harbaugh led the league in passer rating in 1995. He likewise had an abbreviated season consisting of 12 starts.
Nick Foles led the league in passer rating in 2013. He likewise had an abbreviated season consisting of 10 starts.
Every one of the career average QBs noted earlier in the thread who produced 11-game stretches not unlike Tannehill's in 2019 (Rypien, Testaverde, Kramer, etc.) likely led the league in passer rating during those 11-game stretches.
Can't use unadjusted passer ratings for comparisons across NFL history. The league average passer rating has changed way too much for that. In 1978 the league average rating was 65, in the 1990's was in the mid-70's, in 2010 was 84.1 and today it's low 90's. z-scores are necessary for such comparisons and when you use z-scores you see a lot more consecutive seasons with 2019-adjusted ratings of 115+.
Regardless, one can just look up the number of times each QB came in #1 in passer rating (which implicitly takes into account changes in league average). Since 1978 when the rule change made this far more difficult — the standard deviation in passer rating went way down, meaning that "great" QB's had a much harder time being "great" afterwards — you find that no one comes close to Steve Young.
#1 passer rating leaders since 1978:
Steve Young: 6 times (1991-1994, 1996-1997)
Peyton Manning: 3 times (2004-2006)
7 guys doing it twice:
Staubach* (1978, 1979)
Anderson* (1981, 1982)
Montana (1987, 1989)
Warner (1999, 2001)
Brady (2007, 2010)
Rodgers (2011, 2012)
Brees (2009, 2018)
* Both Staubach and Anderson had 2 more #1 passer ratings prior to the 1978 rule change, but those were statistically much easier to accomplish. If you don't care about rule changes, Steve Young is still #1 with 6 but Bart Starr had 5.
Either way, note that this is a who's who list of QB's. So yeah.. once #1 in passer rating is usually an anomaly, but do it twice and that really suggests you're a very good QB. If Tannehill makes that list I got a whole bunch of crow to swallow lol.
All irrelevant when assessing WHAT CPOE and YPA are and how they relate to what the QB had to do.
What did Tannehill have to do in Miami with regard to CPOE? Something different? No, and so why wasn't his CPOE just as large in Miami?
LOL........ You just can't do it..... answer directly. Pretend we're not talking about Tannehill.
Apply the concept to any QB. If a QB has one year of his seven-year career in which is CPOE is far higher than other seasons of his career, how do you explain that? Why wasn't the QB's CPOE roughly as high in every season, with the variation being only in the expected completion percentage?
Here's what Tannehill's career pattern with regard to CPOE should look like if you all's hypothesis about his ability and his surroundings is correct:
Tennessee: expected completion percentage = 62.2, actual completion percentage = 70.3 (which are the actual figures from 2019).
Miami: average expected completion percentage ~54.7 (owing to poorer surroundings), actual completion percentage ~62.8 (owing to his individual ability and indicating that his ability surmounted his surroundings).
That isn't what we see however. How do you explain that?
There's only one way to explain that: he was experiencing situational advantages in 2019.
Expected completion percentage is measuring one outcome through two different players, the QB and the WR.
In order for the QB to actually complete said pass, the receiver in question also needs to make an above average play in most if not all cases.
If you dont have receivers that are capable of making those plays consistently I dont see how it matters what his numbers in Miami were.
If his numbers in Tennessee continue the implication -could- be that perhaps in fact he always WAS making these plays whether the final number showed it or not.
You're great with numbers but you dont seem to have any sort of grasp for their implications or how they are impacted by or connected to other variables.
Also I'm not stating that IS the reason. I'm stating that you're again making definite claims of of information that be interpreted in multiple ways. .
What we're dealing with here is the difference in CPOE between Miami and Tennessee, not the conceptual underpinnings of CPOE. Your post is afield from what we're talking about.
LOL...... Let me ask another way.
If Tannehill had a "low degree of difficulty" in 2019, why was his expected completion % 8th lowest in the league? This takes the QB completely out of it and focuses exclusively the situational advantages on pass plays.
So your assertion is that he was making the same difficulty of throws in Miami as he was in Tennessee?
No, I asked directly about the conceptual underpinnings of CPOE. You ducked the question, twice. I'd prefer that you ignored my questions rather than answer them with something else.
Well there again, how can you consider expected completion percentage an accurate measure of degree of difficulty when his expected completion percentage in Miami wasn't significantly different than his expected completion percentage in Tennessee, despite that his surroundings in Miami were supposedly so much poorer?
We don't see a difference between Miami and Tennessee in expected completion percentage. We see a difference only in actual completion percentage.
From the Next Gen Site.... the source of the stat:
Expected Completion Percentage (xComp) gives an indication of the level of difficulty of a quarterback's throws.