1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

Ryan Tannehill

Discussion in 'Other NFL' started by bbqpitlover, Oct 16, 2019.

Ryan Tannehill is...

  1. A terrible QB

    0 vote(s)
    0.0%
  2. A below average QB

    4 vote(s)
    5.7%
  3. An average QB

    7 vote(s)
    10.0%
  4. An above average QB

    39 vote(s)
    55.7%
  5. An elite QB

    16 vote(s)
    22.9%
  6. The GOAT.

    4 vote(s)
    5.7%
  1. resnor

    resnor Derp Sherpa

    16,329
    9,874
    113
    Nov 25, 2007
    New Hampshire
    Z scores don't win games.
     
  2. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    You're not addressing the inherent and overwhelming weakness of the eye test as it's traditionally understood and practiced -- you can't possibly determine a league norm or the degree of deviation from it on the basis of it. That is quite a low ceiling on what non-systematic personal observations are capable of. Statistics have weaknesses as well, but that low a ceiling isn't one of them.
     
  3. Cashvillesent

    Cashvillesent A female Tannehill fan

    770
    641
    93
    Dec 8, 2019
    My God man. Do you think you know more than GMs or even coaches that played this game for tons of years???? I doubt anyone evaulates the game like you DO.... NO ONE IN THE LEAGUE THAT GETS PAID MILLIONS TO DO JUST THAT.
     
  4. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    Definitely not. I don't think I know more about the game than GMs and coaches.

    What I think is that, generally speaking, with the aid of statistics, I know more about the game than a nobody on a message board -- i.e., someone like you.
     
    mooseguts likes this.
  5. FinFaninBuffalo

    FinFaninBuffalo Well-Known Member

    2,474
    2,954
    113
    Dec 13, 2007
    Are you being purposefully obtuse? Everyone agrees that statistics and personal observations are necessary to better understand what is happening. In addition, since none of us can watch every team the same amount, observations of others, professional analyses, and good old football common sense are all necessary.

    Using the wrong statistic or misusing a statistic is worse than using none at all.
     
    Irishman likes this.
  6. FinFaninBuffalo

    FinFaninBuffalo Well-Known Member

    2,474
    2,954
    113
    Dec 13, 2007
    It is called a straw man argument. He pretends that you said you should only use the eye test and then proceeds to argue against that rather than what you said. Classic distraction technique.
     
    Irishman, resnor and Cashvillesent like this.
  7. Etrius24

    Etrius24 Well-Known Member

    682
    685
    93
    Mar 4, 2020

    Guy

    When Foles is healthy he has been very good.

    The problem is that he only has 48 starts for his career. It is not that he cannot play

    Post Scriptum, I know about websites that have career stats like ESPN or Football Reference. But thank you.
     
  8. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    Return to the original context. What we were talking about is the functioning of offensive lines, using pass block win rate.

    Let's say for the sake of argument pass block win rate isn't completely valid. What do you propose we use to determine a league norm in pass blocking effectiveness and a degree of deviation of any team from it?

    Whatever that is, how would you compare its validity to the validity of pass block win rate?
     
  9. resnor

    resnor Derp Sherpa

    16,329
    9,874
    113
    Nov 25, 2007
    New Hampshire
    That's a silly thing for you to think, based on what you have presented here over the years
     
    Cashvillesent likes this.
  10. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    I would think the same thing about you for example. The same statistics would give you more knowledge about the game, generally speaking, than a nobody on a message board such as myself.
     
  11. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    Another quarterback statistic people might find interesting -- it's called Total Points:
    https://www.footballoutsiders.com/stat-analysis/2018/introducing-total-points

    Here are data from 2019: https://www.sisdatahub.com/leaderboards/QB

    Here are the league z-scores for total points per pass attempt in 2019:

    Jackson 2.01
    Mahomes 1.57
    Brees 1.52
    Rodgers 1.39
    Carr 1.38
    Cousins 1.32
    Wilson 1.16
    Brissett 1.01
    Prescott 0.94
    Tannehill 0.85
    Stafford 0.76
    Bridgewater 0.73
    Garoppalo 0.70
    Fitzpatrick 0.44
    Ryan 0.32
    Watson 0.31
    Rivers 0.17
    Murray 0.10
    Brady 0.05
    Goff -0.15
    Wentz -0.40
    Haskins -0.45
    Lock -0.46
    Minshew -0.52
    Blough -0.60
    J. Allen -0.67
    Darnold -0.71
    Winston -0.93
    Keenum -0.96
    Rudolph -0.97
    Mayfield -1.04
    Mariota -1.07
    Dalton -1.11
    Trubisky -1.14
    Flacco -1.15
    Jones -1.26
    Hodges -1.31
    K. Allen -1.80
     
  12. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    Same as the above statistic, this time league z-scores for total points per pass dropback in 2019:

    Jackson 1.81
    Brees 1.78
    Mahomes 1.67
    Carr 1.45
    Rodgers 1.40
    Cousins 1.37
    Prescott 1.04
    Wilson 0.98
    Brissett 0.95
    Stafford 0.79
    Bridgewater 0.72
    Tannehill 0.70
    Garoppalo 0.70
    Fitzpatrick 0.36
    Ryan 0.32
    Rivers 0.25
    Watson 0.19
    Brady 0.18
    Murray 0.06
    Goff -0.06
    Wentz -0.39
    Lock -0.45
    Haskins -0.55
    Minshew -0.58
    Blough -0.59
    Darnold -0.69
    J. Allen -0.72
    Keenum -0.91
    Winston -0.92
    Rudolph -0.94
    Mayfield -1.02
    Dalton -1.08
    Mariota -1.13
    Trubisky -1.13
    Flacco -1.14
    Jones -1.26
    Hodges -1.31
    K. Allen -1.78
     
  13. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    As usual, the most important quote is this (in the article in which they are introducing the concept!!):
    Football Outsiders already has a black box rating system called DVOA and now they're introducing another black box system call Total Points lol. Best to throw this one in the trash can too. Under no circumstances should one accept lack of transparency.

    You yourself posted an example of how one can approach the division of credit problem through purely statistical means with that article you linked to in the Statistical methods for football thread:
    https://www.degruyter.com/view/journals/jqas/15/3/article-p163.xml

    There, they estimate different intercepts for the best-fitting lines for different positions from the data and use that as the basis for the different weights on each position. It's transparent, reproducible, and if there were no interaction effects that would be a good solution (of course, interaction effects mean that the model isn't the correct one, but it's an approach others can build on and improve).

    In contrast, Football Outsiders repeatedly demonstrate they have no interest in designing a statistical model of football. They're just in the business of conning people who don't think critically with their "trust us we've got the solution" approach. So I'd stay away from that site in general. Even their "statistics" like "QB pressure" are non-transparent so who knows what they really represent.
     
    resnor and The Guy like this.
  14. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    With you completely. I'd love to be able to calculate WAR using the article you referenced (https://www.degruyter.com/view/journals/jqas/15/3/article-p163.xml), but unfortunately trying to do so landed me in a labyrinth of stuff that is completely new to me.
     
  15. FinFaninBuffalo

    FinFaninBuffalo Well-Known Member

    2,474
    2,954
    113
    Dec 13, 2007
    There is no single stat that can give you what you want. PBWR, time to throw, YPA, pressure %, # of players kept in to block, # of pass rushers necessary to get pressure, research into how the offensive is functioning, professional independent written reviews, and good old watching games or film are all necessary to get an accurate evaluation.

    You just cannot quantify it the way you want.
     
  16. resnor

    resnor Derp Sherpa

    16,329
    9,874
    113
    Nov 25, 2007
    New Hampshire
    You've clearly shown a lack of understanding for the game, to the point that you argue that zscores win games.
     
  17. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    The alternative you're proposing (the highlighted portion above) wouldn't permit a high degree of precision (if any) in determining the degree to which a team deviated from the league norm. You'd need to measure those variables with precision for every team to accomplish that, and I don't see how that could be done.

    The approach you're advocating for is the equivalent of a case study of one team, and even if that were done with accuracy and precision, we'd still be unable to determine how that team functions in relation to the average team in the league.
     
  18. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    If I told you the number of points a team scored in a game was three z-scores above the league average, would you think it probably won the game?
     
  19. resnor

    resnor Derp Sherpa

    16,329
    9,874
    113
    Nov 25, 2007
    New Hampshire
    That's a strawman. It doesn't matter what I think it means. Zscores don't win games. They're something invented by statisticians.
     
    Cashvillesent, Hooligan and Etrius24 like this.
  20. resnor

    resnor Derp Sherpa

    16,329
    9,874
    113
    Nov 25, 2007
    New Hampshire
    And let me also add, The guy, that it doesn't matter anyway. The zscores don't tell you that the opposing starting QB was out with the flu that game, and their starting safety left halfway through with a broken toe. Of course, anyone who watched the game wouldn't need zscores, which measures something after it happens, to tell them that the opposing team was gonna get destroyed.

    Some stuff you just know by watching the game. Zscore doesn't capture that. It tests every game as equal regardless of what other factors are in effect.
     
  21. FinFaninBuffalo

    FinFaninBuffalo Well-Known Member

    2,474
    2,954
    113
    Dec 13, 2007
    You are caught in the difference between accuracy and precision. You are looking for precision before you can even be sure that you are accurate. You want to precisely measure something without even knowing what that measurement represents. Most likely, you'll never be able to accomplish what you want.

    Most people only care about one team. Most people don't care about the precise degree to which a team deviates from the league norm. Whatever that is. The plain truth is that for most things, there is no such thing as a league norm.

    Take pass blocking (which is the most recent discussion). Every team in the league is not trying to block the same way, using the same techniques, with the same anticipated outcomes. You don't block for Tom Brady the same way you block for Lamar Jackson. Measuring the two lines against the same standard is not going to work. Also, the team PBWR measurement does not include only the 5 standard linemen. It includes extra linemen, TEs, RBs, H backs, and sometimes even WRs. Every team employs those extra blockers to different degrees.

    When Miami's OLs were particularly bad, they often kept in numerous extra blockers. Often only 2 or 3 receivers with running routes. In addition, they were often facing only 4 pass rushers. The PBWR in those instances is not the same thing as measure PBWR of 5 blockers against 4 or 5 pass rushers, yet that measure does not distinguish.

    Every team in the league is trying to balance keeping players in to block vs having multiple receiving options in the passing game. You want to keep as many blockers in as necessary but no more. That number varies by offense and opponent. Offenses will do what is necessary to limit the pressure on the QB without giving up too much. That is why PBWR doesn't varying by a huge amount, not because OL performance doesn't vary.
     
    Last edited: Apr 18, 2020
    Cashvillesent, resnor and The Guy like this.
  22. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    I think we can agree that single game wins aren't measured in z-scores. However, the underlying point was that they're measured in points, and those are statistics as well. So z-scores per se don't win games, but statistics (points) do.
     
  23. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    I can appreciate that.
     
    FinFaninBuffalo and resnor like this.
  24. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    Here's another rundown that seems more thorough -- see what you think:

    https://sportsinfosolutionsblog.com/2019/09/12/a-primer-on-total-points/
     
  25. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    deja vu quote:
    More specifically, throughout that link they tell you that credit is assigned based on how the player performs relative to expected. Nice statement but the entire validity of the approach lies in the details of how they do the calculations because that's what shows you precisely what assumptions are being made.

    For example, in that nflWAR paper they use linear mixed models so you automatically know they're not assuming the existence of certain types of interaction effects. What assumptions are being made here? We don't know. And looking through sports info solutions, it looks like they're very much in the business of selling you black box stats => reject.
     
    Last edited: Apr 18, 2020
    The Guy likes this.
  26. Etrius24

    Etrius24 Well-Known Member

    682
    685
    93
    Mar 4, 2020

    He lives and breathes by the strawman
     
  27. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    Why do you think they didn’t use a model with interaction effects in that paper?
     
  28. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    An article that bears on the recent discussion here:
    https://www.pff.com/news/pro-z-quar...pff-data-study-of-who-controls-pressure-rates
     
  29. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    Interesting article here about Andy Dalton, who again had a stellar single season in 2015 that stands in sharp contrast to the rest of career. This highlights what I've said here previously about the difficulty in sustaining or replicating the surroundings responsible for such performance.
    https://www.theringer.com/nfl/2019/...on-cincinnati-bengals-quarterback-litmus-test
     
  30. FinFaninBuffalo

    FinFaninBuffalo Well-Known Member

    2,474
    2,954
    113
    Dec 13, 2007
    This is exactly what I have been telling you. Teams will change their play calling to account for problems on the OL. Teams/QBs can control the pressure rate, but not without a cost (in some cases a big cost). Just as PBWR doesn't take into account the style of offense, number of blockers, and number of rushers, neither does pressure rate. From the article:

    That said, a team with virtually no protection can throw the ball quickly and gain yardage (although not optimally)

    How sub-optimal depends on what they are giving up (i.e. how quick they throw and how many blockers they need to keep in), the QB, and the receivers.

    One of the articles that was linked had this:

    A shortage of effective linemen has affected the way offenses function, and blocking struggles have been the worst offender in creating the lackluster product on display at times during the first half of the 2017 season. Scoring league-wide has dropped from an average of 22.8 points per game last season to 21.9 in the first half of this fall, and teams are scoring fewer touchdowns per game (2.38) than they have since 2006.

    This flies in the face of the notion that line play isn't important. Getting the right players on the line seems like a sure fire way to do well since other teams may not be so lucky. Because there seems to be too few to go around, getting good players for your line is even more important. The Dolphins have sucked at it and it showed.
     
  31. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Too many possible models (too many possible types of interaction effects), meaning that it would be hard to argue they used the right one.

    Of course, you can search over a large range of possible models to see which one predicts the data best. The problem is that you'll invariably end up with a large subset of models, many of them quite different, that predict the data similarly well. They'd have a hard time convincing anyone that the model that predicted the data best was the correct model for football.

    So instead they used a standard statistical technique that everyone agrees is a good starting point and allow people to build from there.
     
    The Guy likes this.
  32. resnor

    resnor Derp Sherpa

    16,329
    9,874
    113
    Nov 25, 2007
    New Hampshire
    While interesting, it's different than Tannehill playing on garbage teams for 7 years, then goes to a different team that is 2-4, and then having a crazy season. Dalton played for the same team prior to and after that season. Also, it's silly to act like Tannehill's first few years were not good. It's not like Tannehill had never shown the ability.
     
    FinFaninBuffalo likes this.
  33. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    All of that is certainly a possibility, but at this point we really don't know how bad Marcus Mariota is, and if the Titans' surroundings were elevating his performance to a roughly average level, while he was simultaneously a ball-and-chain on the team.

    In other words, at this point we can't really assume that Mariota is average and Tannehill is above average. It's possible Mariota is terrible and was being propped up to an average level by very good surroundings, while Tannehill was elevated from average to well above average by the same surroundings.

    The belief by many folks here is that the Dolphins' surroundings 2012-2018 were terrible and the Titans' surroundings in 2019 were much better but nothing special, but we don't know that to be the case with certainty at this point. The situation with Tannehill and the Titans could in fact be a facsimile of Dalton's career, where Tannehill gets but one good season out of his surroundings there, and they are unable to sustain or replicate those surroundings thereafter.
     
  34. FinFaninBuffalo

    FinFaninBuffalo Well-Known Member

    2,474
    2,954
    113
    Dec 13, 2007
    It seems really important to you that Tannehill fails next season.

    I'll remind you again, you've provided no evidence that Tannehill was in a particularly easy situation for throwing the ball. On the contrary, his CPOE and aggressiveness % indicate that his passing numbers far exceeded what should be expected. Depth of routes, tightness of coverage, and proximity of pass rushers all contribute to the CPOE. He had the 8th lowest expected completion % in the league. That does not show that he had it easy throwing the ball.
     
    resnor likes this.
  35. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    That's certainly a compelling argument. CPOE has been tracked since only 2016, the year after Dalton's single season that stood in such sharp contrast to the rest of his career, so we don't have that data for him. However, we do know this about him that season (from the article posted above):
    We also know that the z-score of Dalton's 2015 EPA per play, within the context of the past five NFL seasons, was 1.05, whereas Tannehill's in 2019 was 0.64.
     
  36. FinFaninBuffalo

    FinFaninBuffalo Well-Known Member

    2,474
    2,954
    113
    Dec 13, 2007
    What Dalton did and what Tannehill did are unrelated. My only point is that you have claimed multiple times that Tannehill enjoyed a "perfect storm" that made his job easier. Certainly, having a solid running game led to very good numbers across the offense, including points per drive, red zone efficiency, etc. But, in the passing game, Tannehill made his own perfect storm.
     
  37. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    This is a really good explanation of the CPOE statistic:

    http://www.nfl.com/news/story/0ap30...-stats-introduction-to-completion-probability
     
    FinFaninBuffalo likes this.
  38. FinFaninBuffalo

    FinFaninBuffalo Well-Known Member

    2,474
    2,954
    113
    Dec 13, 2007
    The Guy likes this.
  39. The_Dark_Knight

    The_Dark_Knight Defender of the Truth

    11,817
    10,321
    113
    Nov 24, 2007
    Rockledge, FL
    Just adding to your statement...

    If his claim is that Tannehill enjoyed a "perfect storm" that made his job easier...wouldn't that also mean that Heisman Trophy winning quarterback also enjoyed that "perfect storm"? :wink2:

    One sank and the other swam
     
    xphinfanx and Irishman like this.
  40. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    Apparently Ben Baldwin, one of the NFL writers who uses a lot of analytics in his pieces, combines CPOE and EPA per play into an index measure that he says is more predictive of future performance than either one alone, and more predictive of future performance than QBR. The formula for it is at the bottom of the image in this tweet:

     

Share This Page