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. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    It's inherently a negative, because those variables have to be assembled and sustained for the QB to perform in that manner. If what I've been saying here about Tannehill here is true, hell, it's taken seven years of his career to accomplish that! If he'd have gone to a team other than Tennessee, it may have never happened!
     
    Last edited: Sep 28, 2020
  2. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    Pay attention here tonight to how Lamar Jackson fares in a game situation in which his passing volume has to increase to account for what's happening on the scoreboard. He tends to fall apart in those situations.
     
  3. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Two problems with this:

    1) Run probability is dependent on point differential, and we all know that passing efficiency increases when you're ahead, so you'd have to first remove the effect of point differential on Y/A before comparing.

    2) Correlations don't add (talking specifically about that 0.72 + 0.26 ~ 1.0).

    Geometrically, correlations are actually cosines (which also go from -1 to 1) of angles between vectors in an N-dimensional space, where N = number of data points. Specifically, if you have N data points from both T = Tannehill and R = run probability, then the cosine of the angle between T and R (which are vectors in that space) equals the correlation between T and R. Now, imagine two known correlations: one between T and R, and one between H = Henry and R. There is a whole range of possible angles between T and H given the angles between T and R, and H and R. That's why correlations don't add.
     
    The Guy likes this.
  4. AGuyNamedAlex

    AGuyNamedAlex Well-Known Member

    3,582
    2,579
    113
    Sep 12, 2015
    If you could prove to me that every playoff, not superbowl winner, but playoff participating QB was successful in any circumstance on a regular basis over a ten to fifteen year period I'd accept that assertion.

    To the end paragraph:

    Its impossible disprove or prove that X wouldnt have happened if Y didnt happen in relation to human society on every level.

    I could just as easily make the claim if he was drafted by Andy Reid and Mahomes by Joe Philbin, both their careers may be very different.

    You can say its untrue until you get blue in the face, but it's not a claim that can be tested. You cant tell me what would or wouldnt have happened in his career if he signed somewhere else unless you've ascended to some sort of omnipotent godhood.
     
    resnor likes this.
  5. Fin-O

    Fin-O Initiated Club Member

    11,375
    11,392
    113
    Sep 28, 2015
    Isn’t hard to see what Adam Gase does to young Qb’s......I judged him hard and what I perceive as fair while in a Phins uniform and out. But one of the biggest reasons Tannehill wasn’t very good in Miami, can be correlated to Gase, who is flat out trash.
     
    Silverphin and resnor like this.
  6. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    Do you really think so on a season basis? These data are from the 2019 season.

    Right, wasn't implying they could be added, only that it was interesting how Tannehill's correlation is just as strong as Henry's is weak, like a mirror image of each other.
     
  7. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    No what I mean is that the more stars a QB needs to align in his surroundings for him to perform well, the less likely those stars are to align. In that sense any QB's need for particular surroundings is an inherent negative, to the degree to which those surroundings are unlikely to be assembled and sustained.

    If I have a QB who can thrive individually on any average NFL team with an average head coach, obviously that's better than having a QB who needs all-pros at every other position in the offense, a defense that can keep games close, and a HoF head coach.

    Just because a QB has certain needs to perform well individually doesn't mean those needs will ever be fulfilled at the NFL level. And in many cases that isn't the fault of the QB's surroundings -- it's the fault of the QB for having needs that are so unlikely to be fulfilled. If he were a better player himself, he wouldn't have so many needs!
     
  8. AGuyNamedAlex

    AGuyNamedAlex Well-Known Member

    3,582
    2,579
    113
    Sep 12, 2015
    I understand entirely what you mean and agree in general with the premise.

    It's the details of the current situation I somewhat disagree with.

    For instance I dont claim any of the statistics you use are incorrect or invalid. My claim is simply that there are multiple interpretations to most of them and a few at minimum are too general.
     
  9. resnor

    resnor Derp Sherpa

    16,329
    9,874
    113
    Nov 25, 2007
    New Hampshire
    My issue is simply this: to win in the playoffs, and to win a Super Bowl, you need more than average players around a superstar QB. So, honestly, it’s kind of a moot point whether or not Tannehill can carry a bad team on his own, since even if he could, he wouldn’t be winning a Super Bowl.
     
  10. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Absolutely. Efficiency as a function of point differential (for the entire league) is fairly consistent from season to season because you're dealing with passing attempts, and there are a ton of passing attempts in each season.

    You can even see in Tannehill's 2019 splits the effect of playing with a lead:
    https://www.pro-football-reference.com/players/T/TannRy00/splits/2019/

    For Tannehill in 2019, Y/A was 10.1 when playing with a lead while it was 8.6 when trailing, even though the "leading" one was with only 58 passing attempts so there are reliability issues.

    Of course, it's not the magnitude that counts here since you're making an argument about correlation — the magnitude difference could be small like 0.3 or 0.4 in a given year for the entire league but if the trend is highly consistent the correlation could be large. Either way, the effect of point differential has to be removed first before you can make your argument.
     
  11. AGuyNamedAlex

    AGuyNamedAlex Well-Known Member

    3,582
    2,579
    113
    Sep 12, 2015
    To me this debate is only boiling down to:

    "Tannehill is fine, but wouldnt you want a better QB like Mahomes?"

    Which, sure, but that's not so simple or realistic.
     
    resnor likes this.
  12. DolphinGreg

    DolphinGreg Season Ticket Holder Club Member

    5,227
    6,527
    113
    Dec 7, 2014
    I think you're going for something like what researchers would call "robustness," which is just to say that the output of a given thing (e.g. a QB) achieves some quantifiable level of stability despite undesirable values in certain input variables (e.g Franchise, GM, HC, Roster, OC, etc.).

    https://en.wikipedia.org/wiki/Robustness_(computer_science)

    I think we can summarize by just saying "robustness to failure."
     
    The Guy likes this.
  13. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Meanwhile.. we have the 1st breakout of COVID-19 in the NFL with 3 Titans players and 5 team personnel members testing positive for COVID-19, leading to the closing of facilities until Saturday. No decision yet on the Steelers game.
    https://www.espn.com/nfl/story/_/id...ility-multiple-players-test-positive-covid-19

    They better not cancel the games. Otherwise, the same Tannehill stats arguments will continue well into 2021.
     
    FinFaninBuffalo and KeyFin like this.
  14. AGuyNamedAlex

    AGuyNamedAlex Well-Known Member

    3,582
    2,579
    113
    Sep 12, 2015
    He is, and I accept that notion personally. I just dont believe you can make 100% factual statements based on the data he is providing without encountering multiple interpretations.
     
  15. KeyFin

    KeyFin Well-Known Member

    10,488
    12,821
    113
    Nov 1, 2009
    I disagree to an extent. While that statement is 100% true, run probability is dependent on being able to actually run the football. We've struggled there for a few seasons now (hovering around last in the league) because we haven't had the maulers up front like this season. If you can establish the run early, then you keep running the football and creating opportunities thru play-action.

    Simply put, a more balanced offense creates more opportunity to get away from those standard running and passing downs in all phases of the game. We lost a decent bit last year with the lead in the 4th when we couldn't run the football effectively...which made those passing downs more frequent when they shouldn't have been. I see that trend changing for 2020 and beyond.
     
    resnor likes this.
  16. KeyFin

    KeyFin Well-Known Member

    10,488
    12,821
    113
    Nov 1, 2009
    I agree because this will set a precedent on how the NFL handles this going forward- canceling the game is a MASSIVE mistake! They honestly should have given the Titans an automatic bye-week and then had them play Pitt later in the season...but that would have required giving all teams the same bye-week to start so that flexibility was built in.

    Hopefully this doesn't become "a thing" since it can derail the entire season for the league. Someone had to test positive first though; really hoping the NFL has a plan to circumvent this from creating a ripple effect. Or if they do have to cancel, then I'd hope that they would just give both teams a tie to stick to the 16-game format. It's very unfair though that the Steelers are practicing and the Titans aren't.
     
    cbrad likes this.
  17. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Actually that's not true. The average correlation between rushing Y/C and passing Y/A is a minuscule 0.16 throughout NFL history. And all kinds of rushing stats, from rushing efficiency to rushing attempts to prior rushing success, do not correlate well with increased play action efficiency. It's just one of those football "myths" people hold onto that's not borne out by statistical analysis. Here's a graph from FO that shows that well (not everything FO does is bad lol):

    [​IMG]

    Furthermore, I've shown that, on average, almost the entire difference in rushing attempts between winning and losing teams can be explained through point differential in the 4th quarter. So it's not just that The Guy's argument included the effect of point differential as a confound that needs to be removed, it's actually the primary confound. Remove that, and you have little left to explain, on average of course. Individual teams might require more of a scalpel, but that scalpel has to be based on sound statistical arguments.

    Remember, this isn't about what affects run probability per se, it's about what affects BOTH run probability and passing efficiency. And the primary confound that needs removal here is point differential.
     
    The Guy and KeyFin like this.
  18. FinFaninBuffalo

    FinFaninBuffalo Well-Known Member

    2,474
    2,954
    113
    Dec 13, 2007
    Tannehill's fault
     
    resnor, KeyFin and cbrad like this.
  19. KeyFin

    KeyFin Well-Known Member

    10,488
    12,821
    113
    Nov 1, 2009
    Well, maybe I shouldn't have put the "play action" part in there since that was a very small part of the larger point. If your team can't run the football efficiently, then you're much less likely to run the football in the 2nd half of the game. So while I agree with you that rushing yards are often based on point differential in general, it's equally based on your strengths and the opponent's weaknesses.
     
  20. KeyFin

    KeyFin Well-Known Member

    10,488
    12,821
    113
    Nov 1, 2009
    Are we back to that stance already? I figured we were still blaming that evil "B-word" Carol Baskins who fed her husband to the tigers!
     
    resnor likes this.
  21. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Yeah, but the context here (The Guy's argument) isn't about what affects rushing attempts or Y/C. This is about what affects passing Y/A. So you have to think of something that affects BOTH rushing attempts/efficiency and passing efficiency.

    I'd agree with you if we were only talking about what affects rushing Y/C, but that's not the argument here.
     
  22. KeyFin

    KeyFin Well-Known Member

    10,488
    12,821
    113
    Nov 1, 2009
    Aah, okay. The Guy has had me blocked for awhile now so I don't see his posts. I'd still stick with the same argument though that the better you can run the football, the more opportunities you have to get receivers open since the linebackers have to stay at home.

    I don't think that's as easy to prove statistically though because literally every down is situational football based on down/ distance/ the score/ time on the clock and what you're doing successfully. I don't think it's always helpful to generalize that stuff to spot trends when the play call itself is based on what's seen in the moment.

    I mean, if a coach doesn't know for sure what he'll call ahead of time if it's 3rd and 2 with 6 minutes left in the game and he's down 3, then how useful are stats for analyzing that? Sure, it tells us what coaches do in that situation after the fact, but that doesn't mean they made the right call or if what they called can be replicated. It's literally a one-time event that's independent of every other play.

    For instance, Gase was NOTORIOUS for running the ball on 3rd and long because he said that it's the highest percentage play statistically...yet we ALMOST NEVER made a 1st down. So how do those stats help when he was clearly focused on getting the punter more room on 4th down?
     
  23. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    That's the intuition, but it's not borne out by the data, assuming more opportunities => higher passing efficiency. Remember, if it's true in enough cases for situational football, it will show up as a trend.
     
    KeyFin likes this.
  24. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    OK so here are the YPA figures for the five QBs noted earlier, as a function of run probability. These figures are exclusively for weeks 7 through 17 in the NFL in 2019, with teams leading or trailing by 8 or fewer points (point differential between -8 and +8).

    The data are grouped into the following categories:

    1st & 10 = run probability of 54%
    2nd & 4-6 = 46%
    2nd & 7-9 = 32%
    2nd & 10+ = 31%
    3rd & 1-3 = 44%
    3rd & 4-6 = 11%
    3rd & 7-9 = 7%
    3rd & 4th & 1-3 = 13%

    Here is the graph of passing efficiency (YPA) for Tannehill, Jackson, Brees, Wilson, and Mahomes, under the above conditions -- YPA is along the Y-axis (vertically along the left-hand side), while run probability is along the X-axis (horizontally along the bottom):

    YPA2019.jpg
    The correlations between YPA and run probability for the QBs above are as follows:

    Tannehill = 0.75
    Jackson = 0.67
    Brees = 0.08
    Wilson = -0.46
    Mahomes = -0.71

    Tannehill and Jackson's performance varied strongly as a function of run/pass probability, and they thrived in situations in which there was high run probability. As run probability increased, so did their performance. As pass probability increased, their performance decreased.

    Brees's performance did not vary as a function of run/pass probability (i.e., his near zero correlation).

    Wilson's performance varied moderately as a function of run/pass probability, and he exhibited improved performance when pass probability was higher (i.e., the negative correlation). As pass probability increased, so did Wilson's performance.

    Mahomes's performance varied strongly as a function of run/pass probability, and he thrived in situations in which there was high pass probability. As pass probability increased, so did Mahomes's performance, and strongly.

    So, when the whole world knew Patrick Mahomes was going to pass, and when opposing defenses were presumably defending against the pass, he rose to the occasion and played better under those conditions.

    Conversely, when the whole world knew Ryan Tannehill and Lamar Jackson were going to pass, and opposing defenses were presumably defending against the pass, their performance plummeted.

    In my opinion this suggests a strong dependence of Ryan Tannehill on Derrick Henry and the Titans' run game in 2019.
     
    Last edited: Sep 29, 2020
  25. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    No, for me it's "Tannehill is fine, as long as he has Derrick Henry."

    If Derrick Henry isn't around for whatever reason, Tannehill isn't fine.
     
  26. AGuyNamedAlex

    AGuyNamedAlex Well-Known Member

    3,582
    2,579
    113
    Sep 12, 2015
    Again, I dont think the numbers you are using actually say that.

    He plays within an offensive system that is built to do something. It doesnt mean in another system with good players he wouldnt be able to thrive. If the team is built to run why do you think a QB should have success in long yardage situations? They lack the receiving talent to win those matchups consistently and Tannehill isnt Lamar Jackson with his legs.

    You're only basing this off of our dumpster fire of a team but to me that's meaningless.

    Show me where Mahomes succeeds with Harline and Bess as his 1-2 combo without other decent weapons for example. I'm not saying he doesnt or even at a higher level, but you're clinically insane if you actually believe his stat line and the general impression of his abilities wouldnt be different.

    In the end I'm not even saying you're inherently wrong, I'm saying the numbers are too generalized around team play to mean much.
     
    FinFaninBuffalo and resnor like this.
  27. resnor

    resnor Derp Sherpa

    16,329
    9,874
    113
    Nov 25, 2007
    New Hampshire
    I couldn't agree more with this.
     
  28. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    No team in its right mind would be "built to run" if it had one of the league's best passers, and if it did it would change how its built immediately to capitalize on its quarterback's talents. This is a passing league and passing wins games. The only teams that are "built to run" are ones that don't have QBs who can win through the air.

    So if the Titans are "built to run" and they haven't changed that to suit Tannehill's skill set, that's as much an indictment of Tannehill as anything else.
     
  29. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Yeah, I don't think you understood how important my point was about conditioning on point differential.

    Based on an old play-by-play database I have from 2009-2017 (I'm switching over to a new one that includes 1999-2019 data, but the coding for that isn't ready yet), here are the run probabilities and passing Y/A as a function of point differential when each point differential bin is 5 wide and centered at [-15, -10, -5, 0, 5, 10, 15]:

    Run probabilities: [30.62% 33.30% 40.19% 45.69% 47.82% 51.76% 54.56%]
    Passing Y/A: [7.0022 7.1461 7.1542 7.1306 7.3878 7.5763 7.4886]

    The correlation between passing Y/A and run probability when you condition on point differential is 0.8779!!!

    That's massive! On average, that alone will explain 0.8779^2 = 77% of the variation in the stats you're seeing. It also means Tannehill probably straddles league average for QB's if you don't condition on point differential, as it's expected that the correlation is highly positive for a QB that often played with a favorable point differential during 2019.

    As to why Mahomes and Wilson have negative correlations? All you need for that is for those QB's to be way above average in executing comebacks repeatedly during the game (not just 4th quarter), and we all know those QB's are very good at that. So if anything, when you look only at QB's on winning teams like that, what those stats show is the ability to execute comebacks, not necessarily dependence on the run game.

    To show dependence on the run game you need to first remove the effect of point differential, and you'd do that the same way you "adjust" stats by year, except you adjust run probabilities for each game situation by point differential bin, and passing yards by point differential bin for each passing attempt.
     
    The Guy likes this.
  30. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    So help me understand how isolating a relatively small point differential (like the -8 to +8 I used above) and looking at the relationship between YPA and run probability doesn't accomplish what you're saying. My thinking is that the variation in run probability as a function of point differential would be reduced to virtually nil if we're talking about a game situation in which teams are up or down by a single score.
     
  31. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Just take the middle 3 values in those stats. Those go from -7.5 to +7.5 point differential, although it's quite coarse resolution at that point with only 3 bins: (-7.5, -2.5), (-2.5, +2.5) and (+2.5, +7.5).

    The correlation JUST within that -7.5 to +7.5 range is 0.6557. In other words, the effect of point differential is so huge that it matters even if point differential is just a few points, at least around zero = tied ballgame.
     
    The Guy likes this.
  32. The_Dark_Knight

    The_Dark_Knight Defender of the Truth

    11,817
    10,321
    113
    Nov 24, 2007
    Rockledge, FL
    Yea, because a 67.3% completion percentage, averaging 7.8 yards per attempt...competing 809 yards, 6TDs and 1 Int absolutely SUCKS, doesn't it?

    You do realize those are Marino-esque numbers don't you? At this pace, Tannehill will go 373/555 for 4315 yards, 32 TDs and 5 Ints. Just face it, Tannehill is a GREAT quarterback and you just can't stand it...because he's winning for another team; a team that actually has its schidt together as compare to what we had underneath the winless Wunderkind.
     
  33. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    Right now his passer rating is 0.56 standard deviations above the 2020 league average. That's nothing special. 71st percentile.

    In 2014, his third year in the league, his season passer rating was 0.42 standard deviations above that year's league average. 66th percentile. He's performing at about that level so far this year.

    Marino on the other hand had a season passer rating in 1984 that was 2.87 standard deviations above that year's league average. 99.8th percentile.

    Tannehill's 2020 performance is nowhere near the neighborhood of Marino's. There is nothing "Marino-esque" about his numbers.
     
    cbrad likes this.
  34. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Yeah, always have to adjust by era or you get these weird comparisons between Tannehill and Marino.

    One thing that's really interesting this year is not just the higher average passer rating of 96.3 (last year it was 90.4), but also the massive increase in standard deviation in passer rating, which is 17.15 right now and is the reason why a 105 rating doesn't have that high a z-score (at the moment).

    The average standard deviation in passer rating since 1978 has been about 11, while the average pre-1978 was 14.6. The highest ever was 16.57 in 1976, so the effect of COVID-19 (most likely, the effect of the lack of crowd noise) is not only leading to the highest average passer rating ever but also the highest standard deviation ever.

    In those weekly Tannehill updates for statistical significance I haven't yet included these effects because it's too early in the season (such stats do vary from week to week and aren't yet reliable enough for my liking). I'll start including them maybe around week 9 or so, but while the higher average rating might only change the result a little, a much higher standard deviation could affect the results quite a bit more. We'll see. Depends on whether this lasts or not, but that standard deviation is amazing. It's more than 50% higher than it's been in previous seasons.
     
    The Guy likes this.
  35. resnor

    resnor Derp Sherpa

    16,329
    9,874
    113
    Nov 25, 2007
    New Hampshire
    Listen, the average QB in the NFL isn't throwing for almost 4500 yards, 30+TDS and only 5 INTS.

    In trying to explain it away, you're completely missing that he's putting up phenomenal numbers.
     
    The_Dark_Knight likes this.
  36. The_Dark_Knight

    The_Dark_Knight Defender of the Truth

    11,817
    10,321
    113
    Nov 24, 2007
    Rockledge, FL
    My point exactly. These numbers guys are just looking at numbers and summoning it’s the end all be all.

    If OUR team was built like the Titans when Tannehill was here (which I was so hopeful with Ajayi) these guys would be singing praises about Tannehill and how we were winning games left and right.

    Instead of being so obsessed with quarterbacks, maybe they should put more focus on coaches. Reid has done awesomely with 2 different teams and 4 different starting quarterbacks. Belichick seems to be doing quite well with Brady’s successor. John Fox won in Carolina with Newton and with Manning in Denver.

    They can’t do that though. A coach isn’t a “player” (which is crap) and doesn’t have any measurable stats other than wins and losses
     
    Dol-Fan Dupree likes this.
  37. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    The average QB in 2020 right now projects to throw for roughly 3963 yards (nearly 4000 yards), 29 TDs, and 12 INTs, and the numbers you cited for Tannehill are about 0.56 standard deviations better than that, which again is nothing special. If everything continues as-is, there will be about 10 QBs in the league (nearly a third of them) who perform better than that.

    You guys are looking at 2020 numbers for Tannehill and putting them into a context from previous years, rather than this one.
     
  38. The_Dark_Knight

    The_Dark_Knight Defender of the Truth

    11,817
    10,321
    113
    Nov 24, 2007
    Rockledge, FL
    Ok hero, using your number of 3963, 25 teams in the NFL threw for less than that last season so please don’t try and sell snake oil. If you have 7 teams that are throw 4500+ yards and try to average total yards from every team and call that average, that’s as stupid as a football bat.

    And of the 7 teams that threw for over 4000 yards last season, 5 of them didn’t even make the playoffs.
     
    resnor likes this.
  39. The Guy

    The Guy Well-Known Member

    6,598
    3,323
    113
    Oct 1, 2018
    Again you're placing this year's numbers within the context of a previous season.
     
  40. The_Dark_Knight

    The_Dark_Knight Defender of the Truth

    11,817
    10,321
    113
    Nov 24, 2007
    Rockledge, FL
    Ok so this years “average” is going to be so vastly different from last years that the actual hard numbers I posted are completely erroneous and moot.

    You’re too much :sidelol:
     

Share This Page