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

Titans to start Ryan Tannehill

Discussion in 'Miami Dolphins Forum' started by bbqpitlover, Oct 16, 2019.

  1. KeyFin

    KeyFin Well-Known Member

    6,324
    8,279
    113
    Nov 1, 2009
    Looking back, I still have mixed feelings on Gase because he had the right strategy for RT....his team just couldn't execute. Now we're seeing the same thing in NY.
     
  2. Fin D

    Fin D Sigh Club Member

    72,110
    43,521
    113
    Nov 27, 2007
    For me, I just think its such basic common sense what to do with Thill, that giving Gase props for that is like firing a chef because they like to take their pants off to cook then giving props to the next chef simply because he doesn't do that.

    EDIT: Make no mistake, I bought in hook, line and sinker on Philbin and Gase. I was so, so, so, so, so very wrong.
     
    Tin Indian, Dorfdad, Pauly and 3 others like this.
  3. KeyFin

    KeyFin Well-Known Member

    6,324
    8,279
    113
    Nov 1, 2009
    From Yahoo Sports-

    For what it's worth, Tannehill moved up to 13th in ESPN's QBR. He's 26th in Total EPA, which is their way of saying he sucks in the clutch. I have a feeling a part of their final grade comes from that EPA stat across the QB's career...which would explain why he's playing great but ranked 13th.
     
    Last edited: Dec 13, 2019
    The Guy and resnor like this.
  4. AGuyNamedAlex

    AGuyNamedAlex Well-Known Member

    1,672
    1,442
    113
    Sep 12, 2015
    Honestly, I dont think Gase is the worst in the world.

    I also dont think hes the best and I'm glad hes gone, but I'm not going to pretend he had a lot to work with. All the excuses we give Tannehill (except obviously Gase himself) to some extent also apply to Gase.
     
    Irishman likes this.
  5. The Guy

    The Guy Well-Known Member

    1,580
    826
    113
    Oct 1, 2018
    EPA doesn’t have a clutch factor built into it; you might be thinking of win probability added. Here’s a description of EPA:

    http://insidethepylon.com/football-...9/10/25/glossary-entry-expected-points-added/

    Tannehill could very well be so low in the league in Total EPA because he hasn’t played the whole season. I would be interested to see whether they are measuring EPA with him there, or EPA per play. He would likely be considerably higher in EPA per play.
     
    KeyFin likes this.
  6. Fin D

    Fin D Sigh Club Member

    72,110
    43,521
    113
    Nov 27, 2007
    I disagree.

    We were a playoff caliber team while Thill was healthy under Gase. But Gase lost the team in 2017. That's because he's terrible.
     
    Dorfdad likes this.
  7. PhinFan1968

    PhinFan1968 To 2020, and BEYOND! Club Member

    ESPN is garbage.

    3 4th quarter game-winning drives in 7 games, 6-1? Sure...
     
  8. cbrad

    cbrad . Club Member

    7,371
    8,619
    113
    Dec 21, 2014
    I would seriously suggest people stop quoting ESPN's QBR altogether. It's the anti-valid statistics approach. That thing has so many subjective assumptions built into it that no one – not even the creators – have an intuition for what's going on.

    For example.. do you guys know who has the greatest single game QBR ever? Charlie Batch with a 99.9 QBR who went 12 for 17 with 3 TD and 2 INT's in 2010 against Tampa:
    https://www.espn.com/nfl/player/gamelog/_/id/1490/type/nfl/year/2010

    What did ESPN do once people pointed out this crap? They removed Batch's game from the list of all time best QBR:
    http://www.espn.com/nfl/qbr/_/year/0/type/alltime-game

    How? They instituted a totally subjective new requirement that you can't have 17 passing attempts or less lol. Note that Batch still has a 99.9 QBR so they didn't change the formula, they just didn't like what their algorithm outputted and "secretly" removed that game (and others) by adding an arbitrary threshold in for attempts. That's just dishonest.

    And if you really do want to assign credibility to ESPN's QBR, fine.. let's start by pointing out that Tannehill had the 2nd lowest QBR in 2018 among starters, just above Josh Rosen. Who here really thinks Tannehill was worse than Blake Bortles? ESPN does:
    https://www.pro-football-reference.com/years/2018/passing.htm

    In any case, if you're interested in what that propriety rating includes, the best source I've found is this:
    http://www.sloansportsconference.com/content/total-qbr-what-espn-analytics-learned/

    First they start with team expected points added (this part's totally fine and is in fact a good starting point), and then the crap begins: they divide credit amongst players based on all kinds of hidden assumptions, adjust the rating based on everything people would want to adjust for (guys.. it's not WHAT you want to adjust for but whether you can do it!!) and then assign a clutch weight to get the final rating.

    btw.. did I mention that their QBR spans 10,000+ lines of code. Traditional passer rating = 1 line of code. What the hell do you think is in those 10,000+ lines of code? That's right.. hidden assumptions that not even the creators have an intuition about (hello Batch!).

    So as far as I'm concerned, anyone who quotes ESPN's QBR or FO's DVOA (same crap about "adjusting" for things without ever showing how.. meaning it's a bunch of subjective assumptions built in) doesn't care about valid statistical analysis and is demonstrating they prefer deceptive statistical analysis that IS objective (it is a formula) but codifies tons of unknown subjective assumptions.

    Like I said.. just stop quoting that stuff.
     
    Last edited: Dec 13, 2019
    Tin Indian, Irishman, Hoops and 6 others like this.
  9. KeyFin

    KeyFin Well-Known Member

    6,324
    8,279
    113
    Nov 1, 2009
    Just so you know, I quote QBR each week more to mock ESPN than anything- Tannehill is clearly better than the 13th overall QB this season; even people who couldn't stand him in Miami would admit that. I keep quoting it to see if it catches up to reality at all....which I don't think it will.

    The Charlie Batch story is priceless though! Freaking idiots at ESPN.....
     
    Tin Indian, Irishman, Pauly and 2 others like this.
  10. AGuyNamedAlex

    AGuyNamedAlex Well-Known Member

    1,672
    1,442
    113
    Sep 12, 2015
    I dont think we ever had a playoff caliber team honestly. I consider our playoff season an incredible fluke based on our talent level.

    I also give him a semi-pass for the Cutler year.

    You're seeing that now, everyone was shipped off and honestly, we are barely worse off minus Tannehill.

    We cant say Tannehill didnt have the surrounding talent AND Gase had the talent. It's a contradiction.
     
    Irishman and Fin-O like this.
  11. Fin D

    Fin D Sigh Club Member

    72,110
    43,521
    113
    Nov 27, 2007
    I don't understand the bold.

    My stance is this:

    Tannehill never had decent WRs until Landry.
    Tannehill never really had even an average line.
    When Gase took over, Thill had decent offensive skill position talent around him but he still didn't have an oline.
    Gase was a big reason to blame for not improving the oline. He did, however, commit to the run (begrudgingly) and let Thill audible.
    Doing so, made the team playoff worthy...until Thill went down. Then Gase couldn't hold the team together because he doesn't really know how to lead. He also had no idea how to spot defensive coordinator talent....and he's probably a coke head.
     
    PhinFan1968 and resnor like this.
  12. PhinFan1968

    PhinFan1968 To 2020, and BEYOND! Club Member

    His biggest flaw IMO.
     
    Fin-O likes this.
  13. KeyFin

    KeyFin Well-Known Member

    6,324
    8,279
    113
    Nov 1, 2009
    Those last two Gase seasons we had virtually the entire starting offensive line on injured reserve. It wasn't just that they didn't have the right guys...our starters were "okay". The problem was that we were paper thin at depth and it was a massive drop-off to the reserve players.
     
    resnor and Irishman like this.
  14. cuchulainn

    cuchulainn Táin Bó Cúailnge Club Member

    11,026
    9,510
    113
    Sep 7, 2012
    Hattiesburg, MS
    The Guy likes this.
  15. cbrad

    cbrad . Club Member

    7,371
    8,619
    113
    Dec 21, 2014
    Interesting.

    Two things stated in those graphs are worth mentioning because if you don't see/understand them those graphs can be misleading:

    1) That first graph says "250-dropback average", meaning that the last data point in that spike with Tennessee includes the 203 passing attempts with Tennessee AND the last 47 attempts with Miami. In other words, that spike with Tennessee is NOT his performance with Tennessee alone, it's a mixture of his performance at the end of 2018 with Miami and 2019 with Tennessee. If it was only Tennessee it would be WAY higher than 2014 or 2016 so people shouldn't think that graph is somehow suggesting he did similarly well in 2014 or 2016.

    2) That second graph, in case people don't see it, is a mixture of 3 things: PFF's subjective grades, EPA per dropback, and draft position prior (meaning you automatically start off with a higher grade if you were drafted higher). The "Bayesian updating" part just means that after each new data point they change the likelihood of each hypothesis being correct based on a mathematics called Bayesian inference. The important thing to note here is that the reason Tannehill is mostly between 50th and 75th percentiles is not because of EPA per dropback but because of his PFF grades (which are subjective). You can see this clearly in the 2nd graph in that link:
    https://www.pff.com/news/nfl-titans-scheme-has-made-ryan-tannehill-a-top-10-quarterback

    The only thing I don't quite understand is how they define a dropback. There are over 3500+ dropbacks in that first graph, yet Tannehill has only 3114 passing attempts and 272 sacks. Adding those up won't get you there. They also say he has 256 dropbacks this season, yet Tannehill has 203 passing attempts plus 24 sacks, which is also short. So what else is PFF including in a "dropback"?

    Finally, the bolded quote from Kevin Cole in cuchulainn's post, "I found a strong fit between Tannehill, his surrounding talent, and the Titans' passing scheme that has me believing" isn't shown or suggested to us by these graphs. There must be something else Kevin Cole wrote in that article (can't see the whole thing for free) that is the basis for that statement.
     
  16. KeyFin

    KeyFin Well-Known Member

    6,324
    8,279
    113
    Nov 1, 2009
    Maybe they gave him 29 bonus dropbacks for the way he dropped back and knocked the be-jesus out of that defensive lineman that intercepted the ball!
     
    PhinFan1968, Pauly and cbrad like this.
  17. AGuyNamedAlex

    AGuyNamedAlex Well-Known Member

    1,672
    1,442
    113
    Sep 12, 2015
    What it means is a team has talent to win or it doesnt. There cant be a combination where Tannehill doesn't have the talent surrounding him to succeed on offense as a QB but Gase has the talent on offense to win games. I think a lot of our conservative play calling was from knowing that our line was awful and couldnt hold a pocket if the defense knew we had to pass.

    I do think there are better ways to mitigate pressure than just drinking and dunking btw, but I at least understand a bit why it happened.

    I cant speak to how much input he had into those talent problems as well, our GM situation was weird.

    To me he is painfully average. Maybe a little below, but I think people here also think hes a lot worse than he really is.
     
    Last edited: Dec 14, 2019
  18. The Guy

    The Guy Well-Known Member

    1,580
    826
    113
    Oct 1, 2018
    And here is the conclusion even in an article as favorable to Tannehill as the one outlined above, entitled "The Titans' Scheme has Made Ryan Tannehill a Top-10 Quarterback":
    https://www.pff.com/news/nfl-titans-scheme-has-made-ryan-tannehill-a-top-10-quarterback
     
  19. The_Dark_Knight

    The_Dark_Knight Defender of the Truth

    8,623
    6,986
    113
    Nov 24, 2007
    Melbourne, FL
    Huh? My point was that only 2 Heisman Trophy winning QBs have won the Super Bowl...

    Roger Staubach and Jim Plunkett
     
    Hoops likes this.
  20. PhinFan1968

    PhinFan1968 To 2020, and BEYOND! Club Member

    For the # of dropbacks, don't they count dropbacks even if a penalty erases the play? That may account for part of it.
     
  21. AGuyNamedAlex

    AGuyNamedAlex Well-Known Member

    1,672
    1,442
    113
    Sep 12, 2015
    If they do that it's kind of ridiculous to me.

    Half the time a QB sees a flag he knows is against the defense he will end up taking a massive deep shot he wouldnt have taken otherwise.

    I suppose their logic is if you include declined penalties you need to include accepted as well but ehhh...
     
  22. Irishman

    Irishman Well-Known Member

    327
    353
    63
    Oct 16, 2017
    High Point, NC
    I don't agree with you about Gase, therefore your stating "...it’s unanimous Adam Gase is just a garbage coach" is wrong.

    How wrong is still to be determined. Based on the unsupportable and overstated "unanimous" crap statement about Gase I just have to ask, do you work for CNN?
     
    Fin-O likes this.
  23. FinFaninBuffalo

    FinFaninBuffalo Well-Known Member

    444
    896
    93
    Dec 13, 2007
    Possibly scrambles and completed plays called back by penalty? At a minimum, they should include scrambles if they are going to include sacks.
     
  24. Mcduffie81

    Mcduffie81 Wildcat Club Member

    4,554
    3,131
    113
    Mar 23, 2008
    Lake Worth, Fl.
    Fin-O is one of the few here that wouldn’t be caught dead working for CNN.
     
    Fin-O and Irishman like this.
  25. cbrad

    cbrad . Club Member

    7,371
    8,619
    113
    Dec 21, 2014
    What makes it a bigger mystery is there are multiple sites that actually track dropbacks and they explicitly define them as "passing attempts + sacks - spikes removed". In other words, the purpose of distinguishing between "dropback" and "passing attempt" is to look at the number of times a QB actually drops back to pass, so that number is always LESS than passing attempts + sacks and doesn't include scrambles. And all official NFL passing stats remove whatever happened during a penalty (because those plays didn't officially occur).

    So whatever PFF is doing it's not based on the standard use of the term, which is bad for comparison purposes.
     
  26. cbrad

    cbrad . Club Member

    7,371
    8,619
    113
    Dec 21, 2014
    Yeah they're apparently not familiar with hypothesis testing. Otherwise they'd see that there's statistical evidence the conditions changed in Tennessee (either the surroundings changed or the QB changed or both) and that you really shouldn't base any prediction of future performance on the assumption the conditions didn't change!

    There's also a subtle hint they don't understand statistical analysis because they used Bayesian inference in a case where it's not really the right method. Bayesian inference is useful for predicting how the likelihoods of hypotheses change (or probabilities of events change) after a single extra event (ONE extra data point is obtained). You don't use that if you're just combining all data from the past, which is conceptually what they're trying to do in that second graph. You use MLE = maximum likelihood estimation.
     
    The Guy likes this.
  27. FinFaninBuffalo

    FinFaninBuffalo Well-Known Member

    444
    896
    93
    Dec 13, 2007
    I'm voting that the surroundings changed more. I'm basing that on the fact that he is playing in a different offensive scheme, with a different coaching staff, and different players. If Tannehill also playing better? Based on what I see, yes. He has been an accurate QB but he have been even more accurate this season. He has done well from play action but he has been even better this season.

    Clearly he won't maintain an average passer rating of 118 and a YPA of 9.8 for the rest of his career (or even the rest of the season, most likely), but, IMO, he has shown that he is a top 10 QB when given adequate surroundings.
     
    PhinFan1968 likes this.
  28. Cashvillesent

    Cashvillesent Well-Known Member

    278
    278
    63
    Dec 8, 2019
    Stats in football are pretty much useless.

    It never gives you an accurate way of evaulating a QB.

    Football might be the only sport that it is really a TEAM SPORT.
     
    Dol-Fan Dupree likes this.
  29. cbrad

    cbrad . Club Member

    7,371
    8,619
    113
    Dec 21, 2014
    Football is somewhere in the middle in terms of the degree to which you can identify individual contributions vs. the degree to which stats are affected by what the rest of the team does. The fact (American) football separates plays, separates offensive/defensive/ST units, and has many different positions where the rules aren't the same for each position makes it easier to apply stats than in more fluid sports like soccer or ice hockey where it's nearly impossible to do useful statistical analysis.

    Obviously football is a lot more complicated for statistical analysis than baseball, but it's by FAR not the only "team sport" on the planet lol.
     
  30. Fin D

    Fin D Sigh Club Member

    72,110
    43,521
    113
    Nov 27, 2007
    I'm sorry I wasn't clear. I meant I don't understand it relative to the discussion. As I said, when Gase was here, surrounding talent (outside of the oline) wasn't bad on offense. We were literally the hottest team in the league until Thill got hurt.

    Gase is average, on strategy and scheme. His biggest flaw int hat department is his ego, as it takes him too long to shift gears from something not working. The leader aspect of the job, he was woefully below average at. The team didn't really buy into him and it wasn't he sucked it up and started committing to the run did the team go along with him, because that's when we started winning. The following year, with no Thill, we weren't good and the only thing he had to get the team to follow him, which was winning, was gone.

    Gase failed because:

    - He was a **** leader. (Sorry, nowhere in the TOS does it say I'm not allowed to type *'s. Stop editing my posts.)
    - Thill got hurt.
    - He chose Burke to be DC.
    - He's likely a coke head.
     
    Last edited: Dec 14, 2019
    resnor and PhinFan1968 like this.
  31. PhinFan1968

    PhinFan1968 To 2020, and BEYOND! Club Member

    You watch the most successful coaches, and they don't just pass the other areas that aren't their specialty off to coordinators and forget about 'em...Gase does. Biggest leadership fail he has shown, to me. Just listen to Harbaugh, Tomlin, Belichick, etc. They come across as equally knowing all areas, even though you know they have a specialty.

    When you're the HC, you don't operate "Us vs. Them" within your team...that's up to the coordinators to have their internal competitions between the units. It's your TEAM moron...not your offense. He'll never be more than a good coordinator IMO.
     
    Last edited by a moderator: Dec 14, 2019
    Fin D and resnor like this.
  32. The Guy

    The Guy Well-Known Member

    1,580
    826
    113
    Oct 1, 2018
    That depends on sample size.

    Certainly the statistics from a single game could be highly inaccurate in determining the ability of a quarterback. But by the same token, you can’t very well say that Peyton Manning’s career statistics are meaningless in reflecting his ability. Obviously they mean something.
     
    cbrad likes this.
  33. The Guy

    The Guy Well-Known Member

    1,580
    826
    113
    Oct 1, 2018
    Given the data above that show the peak in the graph in 2014, similar to the one this year, and given my post I copied above, I'm still stuck wondering what the problem would be with comparing his most recent eight games to any other eight games in his career.

    I hear what you're saying above about how there isn't an identifiable difference in those previous subsets of games, but that implies that we need to use deductive rather than inductive reasoning in this instance.

    Why can't we use the data to induce that, because Tannehill's current eight-game performance isn't significantly different from those previous ones, whatever conditions that are causing his current performance aren't significantly different from the ones that existed during the eight games in both 2014 and 2016 (in my post above), despite that he was on different teams?

    Certainly there is overlap among teams in the NFL. Simply being on a different team doesn't necessarily cause a significant difference in one's surroundings of import.
     
  34. cbrad

    cbrad . Club Member

    7,371
    8,619
    113
    Dec 21, 2014
    Yeah what I was trying to convey in the post you quoted is that you have to be mindful of the reason a statistical test says two sets of ratings aren't significantly different from each other. The reason could be entirely due to small sample size in which case it's kind of meaningless to use the statistical test.

    For example, let's compare two sets of ratings: {70, 80} and {100, 110}. The t-test says the probability those two sets come from the same QB is 5.13% and is therefore not statistically significant.

    Now watch what happens when we increase sample size a bit: {70, 80, 70, 80} and {100, 110, 100, 110}. Now that same test says the probability those ratings come from the same QB is 0.0324%. Huge difference due only to sample size.

    Where is that coming from in the math? Look at this link:
    https://en.wikipedia.org/wiki/Student's_t-test#Assumptions

    That n in the denominator represents sample size, and it goes into the estimate of the standard error. When n is small, the SAME difference in means is much more likely to occur, which makes total sense: the less data you have the more uncertainty in your estimate.

    Point is.. you want to get to a sample size large enough with these statistical tests so that the result doesn't depend on adding one or two more data points lol. It shouldn't really change with 10+ more data points. Otherwise you're just talking about the effect of sample size. So if you were to make the argument with those 8-game stretches, I'd just simply point out the test is kind of meaningless in this case because of the small sample size and interpret the result by saying the uncertainty is too large to make a determination.

    Note however that if you do find statistical significance (as opposed to it not being significant) then it's another story.
     
  35. AGuyNamedAlex

    AGuyNamedAlex Well-Known Member

    1,672
    1,442
    113
    Sep 12, 2015
    I believe he failed because while he is capable of creating some great ideas he is too arrogant to admit when one of his ideas isnt working and leave it behind.

    If that helps clear my own position any.
     
    resnor likes this.
  36. Fin-O

    Fin-O Initiated Club Member

    9,885
    9,272
    113
    Sep 28, 2015
    I am Don Lemon.
     
    Irishman and PhinFan1968 like this.
  37. FinFaninBuffalo

    FinFaninBuffalo Well-Known Member

    444
    896
    93
    Dec 13, 2007
    I don't agree that his current 8 game performance isn't significantly different than previous ones.
     
    PhinFan1968 likes this.
  38. Fin-O

    Fin-O Initiated Club Member

    9,885
    9,272
    113
    Sep 28, 2015
    Or ANY of those laughably bias brainwashing shows. Not my style.
     
    Irishman and PhinFan1968 like this.
  39. cbrad

    cbrad . Club Member

    7,371
    8,619
    113
    Dec 21, 2014
    It is significantly different, but you have to use the right approach to show that. I described that approach in post #1516:
    So.. first you adjust all ratings to a common year (let's say 2019), then you look at the mean passer ratings of ALL 8-game stretches within a season (so we don't cross over from one season to the next). This is what the distribution of those 47 means looks like:
    [​IMG]

    Now you calculate the z-score of his 118.95 (that's the literal mean for the current 8-game stretch), and that's 3.07. Basically anything beyond z-score = 2 is statistically significant, so the current 8-game stretch is statistically significant.

    The Guy is trying to do a statistical test on two sets of 8 game stretches at a time, and while there are statistical tests for that, as I pointed out with such a small sample size the results are meaningless (or better: indeterminate). It's best in this case to simply compare to all 8-game stretches and calculate the z-score. So yes what Tannehill is doing in Tennessee is statistically significant even IF you treat it as just another 8-game stretch rather than looking at 88 ratings from Miami vs. 8 from Tennessee.
     
    The Guy likes this.
  40. PhinFan1968

    PhinFan1968 To 2020, and BEYOND! Club Member

    Wouldn't that be a "all other things being equal" type of comparison?
     

Share This Page