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

Ryan Tannehill best games compilation

Discussion in 'Miami Dolphins Forum' started by Brasfin, May 15, 2016.

  1. Stringer Bell

    Stringer Bell Post Hard, Post Often Club Member

    44,356
    22,480
    113
    Mar 22, 2008
    The legal definition of gambling specifically lies in the presence of skill. This is precisely what is going on with daily fantasy sites.
     
  2. Stringer Bell

    Stringer Bell Post Hard, Post Often Club Member

    44,356
    22,480
    113
    Mar 22, 2008
    Knowing the odds is not necessarily assessing them. If the odds are inherent to the rules of the game itself, then there is no assessment necessary. The odds of flipping a coin are inherent to the game itself.
     
  3. roy_miami

    roy_miami Well-Known Member

    1,385
    560
    113
    Oct 11, 2013
    Yeah, like the word predict...

    [h=2]Simple Definition of predict[/h]

    • : to say that (something) will or might happen in the future
     
  4. resnor

    resnor Derp Sherpa

    16,329
    9,874
    113
    Nov 25, 2007
    New Hampshire
    That's a very simple definition, and ignores the connotation of the word "predict."
     
  5. Finster

    Finster Finsterious Finologist

    3,087
    2,038
    113
    Jul 27, 2013
    As I said in my comment, if we are taking out wagers, because when wagering, it's gambling, no matter what the risk, if Bill Gates bets $5, that's gambling.

    That is one definition.

    There is another definition.

    My point was that normal driving is not gambling, because there is no high risk, in a non wager situation, gambling involves high risk.
     
  6. resnor

    resnor Derp Sherpa

    16,329
    9,874
    113
    Nov 25, 2007
    New Hampshire
    You're going way of what was being discussed. That is, using patterns to predict something, and if something is predictable, it therefore isn't random.

    You are now just making predictions, without basis in anything. Again when discussing say the lottery, a prediction, based on data, would be buy X number of tickets, and you will win. Your bet with me on powerball, for instance, if money were no object, I imagine there's a number of tickets I could buy that would make it likely for me to win. A prediction would tell me how many to buy.
     
  7. resnor

    resnor Derp Sherpa

    16,329
    9,874
    113
    Nov 25, 2007
    New Hampshire
    Of course there's high risk. Every time you're in your car driving there is risk of bodily injury or death.
     
  8. Stringer Bell

    Stringer Bell Post Hard, Post Often Club Member

    44,356
    22,480
    113
    Mar 22, 2008
    There really should not be much dispute over what constitutes gambling. There are well defined legal tests for this:

    https://en.wikipedia.org/wiki/Dominant_Factor_Test
     
  9. Finster

    Finster Finsterious Finologist

    3,087
    2,038
    113
    Jul 27, 2013
    That is a statutory definition however, and statutory definitions are sometimes different than the actual definition, as they are pertaining only to legal matters.

    In the actual definition, any wagering of any type is by definition, gambling.
     
    resnor likes this.
  10. Stringer Bell

    Stringer Bell Post Hard, Post Often Club Member

    44,356
    22,480
    113
    Mar 22, 2008
    it becomes a matter of semantics at this point, and dependent on what sources you are looking at. For example, this definition of gambling:

    ...specifically identifies games of chance.

    Legal, mathematical, and technical definitions are generally going to remove the level of ambiguity that exists with language and colloquial definitions.
     
  11. Finster

    Finster Finsterious Finologist

    3,087
    2,038
    113
    Jul 27, 2013
    The direct synonym for gambling, is betting, as the definition you provided says,
    without applying any conditions, to bet, is to gamble.
     
  12. Stringer Bell

    Stringer Bell Post Hard, Post Often Club Member

    44,356
    22,480
    113
    Mar 22, 2008
    Sure, but you still have the threshold of a game of chance.
     
  13. Pauly

    Pauly Season Ticket Holder

    3,696
    3,743
    113
    Nov 29, 2007
    I am so used to seeing 3x3 matrixes for analysing game theory I just divided by nine, when I should have divided by eight.

    Correct numbers are
    57.6% and 66.5%
     
  14. Finster

    Finster Finsterious Finologist

    3,087
    2,038
    113
    Jul 27, 2013
    The semicolon indicates separation of the 2, so games of chance doesn't apply to bet, as there are many things that are bet on that are not games of chance.
     
    resnor likes this.
  15. roy_miami

    roy_miami Well-Known Member

    1,385
    560
    113
    Oct 11, 2013
    Pretty sure the word prediction does not fit there. I think probability calculation fits better?
     
    Finster likes this.
  16. Pauly

    Pauly Season Ticket Holder

    3,696
    3,743
    113
    Nov 29, 2007
    If we look at performance as explained by game theory.

    Outcome = Skill x % of optimum strategy +/- variance.

    Over sufficient iterations of a game variance will approach zero. (A novice at Monopoly can beat the world champion in one game, but the world champion will win 999/1000).

    So the ceiling of a player is if they choose the optimum strategy 100% of the time. So in football the ceiling of an OC's abiilty in the passing game can be measured by the QB's passer rating (or other measure of performance)

    In unopposed games (eg the Tax System) where the player chooses a strategy that is then applied according to the rules then the optimum strategy can be selected by straight mathematical analysis. For example in a game of Monopoly with low skill players getting all 4 railroads is commonly described as an optimum early game strategy because it provides the steadiest income stream.

    In opposed games (where the result is affected by your opponent's strategies as well as the rules). So the optimum strategy changes according to opponent. Analysis of tournament result of games like Monopoly and Backgammon show that the ability to select the optimum strategy according to opponent is a stable factor.

    So with the Sherman and Lazor years (32 games each) the sample is large enough to remove variance as significant factor for analysis. Also I think it is fair to assume both coaches were trying to choose the optimum strategy 100% of the time. The NFL averages show that being able to obtain an even split is the standard performance. Therefore if there is a differential according to split the explanation, according to game theory, is that one player is choosing as sub optimal strategy in that situation.

    My conclusion of the data is that Sherman's offense had Tannehill as maybe an 85 rated passer at its optimum when you adjust for RT17s rookie year. Under Lazor Tannehill was a 100 rated when the optimum strategy was being applied. With RT17's 80 rating when behind being (mostly) due to sub optimal strategy applied in that situation.

    I totally agree with the run/pass ratio that game situation being heavily dependent on game situation. I don't think that there is particular magic number for what is the optimum run/pass ratio should be. For me it was a convenient measure of when Lazor became predictable to opposing defenses.

    I think the offense design theory has problems because the yards per attempt and yards per catch numbers drop when Miami were behind (also refelected in the rushing less than normal stats\). The bottom 8 games of the rushing less than usual are the games where Miami were on average behind the most points for the longest part of the game. In that situation you would expect deeper passes being called yet yards per catch was going down.
     
  17. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    So, I agree coaches are trying to do what's best (optimal), but that doesn't mean they employ optimal strategies, meaning you can't conclude they were suboptimal in only one condition. If there's a differential, it could be because the player (in game theory) chose a suboptimal strategy relative to average in only one condition, or it could be because that player chose a closer-to-optimal strategy relative to average in the other. No way to tell.

    This is why I think a better theory that fits your data is the more parsimonious one that doesn't make assumptions about how optimal strategies are. Just assume Lazor chose an offense for Tannehill that emphasized lots of high-percentage (shorter) passes. Suppose he was right that such an offense would benefit Tannehill in situations where you don't need to take many risks. That right there explains all your data. The big increase in YPA under Lazor when tied/leading is explained by that, and the decrease in YPA when trailing is also (this is where you need to take risks and Lazor didn't).

    My point is there's a simpler explanation here that doesn't require any assumptions about the predictability of Lazor's offense, which is really hard to test anyway.
     
    Pauly likes this.
  18. resnor

    resnor Derp Sherpa

    16,329
    9,874
    113
    Nov 25, 2007
    New Hampshire
    Well, truthfully, I don't entirely like the Powerball example for arguing prediction vs probability, but really, the prediction would be telling me, based on my purchasing history, and the probability of winning, how long I'll have to play the Powerball, until I actually win (again, given no time constraints). That's a prediction. Simply stating odds to me is not really predicting anything.
     
  19. resnor

    resnor Derp Sherpa

    16,329
    9,874
    113
    Nov 25, 2007
    New Hampshire
    Why couldn't the theory be that Lazor employed an offense that best fit all the offensive personnel, including the incredibly limited oline?

    Why does stuff always get looked at in a way to try to cast Tannehill in a negative light, when we know that the oline, for instance, was far, far worse than Tannehill. If an OC were adjusting an offensive scheme, it would seem far more likely to scheme around the weakest part.

    Further, if the OC believed the QB to be the weakest part, then why the heck would he have his QB throw more than most QBs in the league? It defies logic.
     
    Fin D likes this.
  20. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Seriously?? .. resnor I think you're a bit paranoid about people casting Tannehill in a negative light. I think it's clear Pauly and I are talking about the effect of Lazor. I suggest you try to take context into account and not just look at whether the word "Tannehill" is in there. I suggested nothing negative about him in case you can't see that.
     
    jdang307 likes this.
  21. resnor

    resnor Derp Sherpa

    16,329
    9,874
    113
    Nov 25, 2007
    New Hampshire
    Oh, I'm sorry. I assumed when talking about Lazor tailoring the offense for Tannehill, that you were taking about Tannehill. Not to mention, you aren't the biggest fan of Tannehill. LOL I was just trying to point out, that you don't have to look at the offense design as being about Tannehill.

    Regardless, it's just as likely, imo, that the short pass offense was designed to minimize the effect of the oline.
     
    cbrad likes this.
  22. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Yeah, I used "Tannehill" just like you'd say a pitcher won or lost a game. It's shorthand (and I'll use it that way again.. be warned!). My issues with Tannehill are well known by now but haven't been the topic of this conversation.
     
    Pauly and resnor like this.
  23. resnor

    resnor Derp Sherpa

    16,329
    9,874
    113
    Nov 25, 2007
    New Hampshire
    Yeah...if a pitcher gives up one run, and his team loses 1-0, I don't think anyone is going to hang the loss on that pitcher, even though it goes on his stats as a loss...
     
  24. Fin D

    Fin D Sigh

    72,252
    43,684
    113
    Nov 27, 2007
    Exactly.

    The argument is that Lazor wasn't an idiot and that Thill is the problem, but if Thill was the problem then that means Lazor made a crap Qb throw it way more than he should have....which makes him an idiot.
     
    resnor likes this.
  25. Pauly

    Pauly Season Ticket Holder

    3,696
    3,743
    113
    Nov 29, 2007
    My first bone of contention is that if the strategy remained the same and the personnel didn't vary, why did the outcomes vary? The NFL average indicate that most teams can keep similar performance, which makes Miami an outlier.

    Secondly, when we looked deeper at the Sherman years and the bottom tier QBs we saw that their efficiency (as meausured by passer rating) remained about the same, despite the stats showing that a different pass mix was called. This is consistent with you can only coach to your maximum ability. So changing the mix to higher reward/higher risk options didn't change the average outcomes, but it did change the variability of outcomes. So if Lazor changes to a low reward/low risk mix and still coaches to his maximum I would expect similar outcomes in terms of efficiency, but with lower variance.

    The simplest explanation, and something I have seen in many times in my years of playing board games, is cognitive dissonance. In this instance itl meant Lazor believing that his offense is optimum.
    When his offense does not deliver he can either admit his belief that his strategy in optimal is false or look for explanations that his strategy was optimal but there are other reasons (variance) that explain failure and he can believe his strategy is optimal.
    a - My strategy was good, but it was an exceptional opponent. My strategy doesn't need adjusting.
    b - My strategy was good, but implemented poorly by players. My strategy doesn't need adjusting.
    c - My strategy was good, but I was let down by the defense or special teams. My strategy doesn't need adjusting.
    d - We won, therefore my strategy was good and doesn't need adjusting.

    This is particularly true in situations where a strategy is partially successful. The player will ascribe all the successes to the brilliance of their strategy and all the failure to different variances. If a strategy is mostly unsuccessful the player will go back and come up with a new strategy.

    The whispers we have coming out of the club house indicate that Lazor blamed Tannehill for the failures in his strategy. I would say this is a sign of cognitive dissonance because it was exactly the same QB running the offense when it worked successfully.
     
    resnor and cbrad like this.
  26. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Good question! My guess: Miami isn't that much of an outlier. You probably see teams like Miami and teams the opposite of Miami. It's just that the averages of each stat are about the same league-wide. One small piece of evidence for this is 2012-2013 Miami vs. 2014-2015 Miami. YPA is greater (as you showed) in exactly the opposite situations.

    Yeah, again I'd bet that if you look at the distribution you'll see quite a bit of variance. You showed averages, which were actually not that different according to situation for all groups, not just the bottom group. You're right, using passer rating the bottom tier showed the smallest difference, but using YPA the bottom tier showed the highest difference and the upper two tiers showed almost none. All in all, those differences are small anyway. So I think what that's masking is the variance. Sherman just happened to be one of those OC's where for him both stats were similar.

    Not sure you can use cognitive dissonance to explain much because it arguably applies to every (or almost every) coach and player.
     
    Finster likes this.
  27. Pauly

    Pauly Season Ticket Holder

    3,696
    3,743
    113
    Nov 29, 2007
    I should have also said cognitive dissonance doesn't explain selecting a sub optimal strategy in the first place, but it does explain persisting with that strategy over a long period of time despite mounting evidence that the strategy was sub optimal.

    Also my personal bias comes into play. I have seen many 'second tier' players in complex games persist with sub optimal strategies, when the best players are more flexible about tailoring their strategy to the current situation.
     
  28. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Yeah, humans have many behavioral biases. Confirmation bias, sunk costs, all kinds of things. So cognitive dissonance is just one property of being human (and in general even the best players have a lot of these types of biases).
     
    roy_miami likes this.
  29. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Also.. there is a question as to whether these stats are accurate or even the ones we should be using. You say all "starting QB's" so maybe it's different than the official averages:
    http://www.pro-football-reference.com/years/2015/splits.htm
    http://www.pro-football-reference.com/years/2014/splits.htm
    http://www.pro-football-reference.com/years/2013/splits.htm
    http://www.pro-football-reference.com/years/2012/splits.htm

    For those 4 years, the average passer rating in trailing situations is much lower than in tied/leading situations (which btw makes more sense). It averages about 7 points less. Not sure if you're compiling the stats for a select group of QB's on your own, but since these are pretty much official stats, we should probably use them.

    So Tannehill under both Sherman and Lazor produced numbers off from average. Sherman's was 7 points less than average difference (because passer ratings were tied), while Lazor's was 11 above average difference. Question now is what is the variance across the league. Here's the data for Tannehill:
    http://www.pro-football-reference.com/players/T/TannRy00/splits/2015/
    http://www.pro-football-reference.com/players/T/TannRy00/splits/2014/
    http://www.pro-football-reference.com/players/T/TannRy00/splits/2013/
    http://www.pro-football-reference.com/players/T/TannRy00/splits/2012/

    Obviously, this isn't separating by run/pass balance, but if that balance is more an artifact of game situation, it's useful to look at overall averages.
     
  30. Pauly

    Pauly Season Ticket Holder

    3,696
    3,743
    113
    Nov 29, 2007
    I used "starting QBs" (299+ attempts in my definition) because
    1) I didn't want the data being skewed by back-up QBs being used with cut down playbooks/reduced preparation time and non-QB passing [logic]; and
    2) I already had the data in my spreadsheet in a form where I could get some more detailed breakdowns [laziness].

    Also I noticed something funny. When I did a simple average of QB ratings I got a 5 point drop from being in front to being behind, (NFL Tied/Ahead 94.9; NFL Behind 89.7) but when I re-calculated using the total data and the passer rating formula the number ended up even. So I'm assuming that the distribution of the data was the cause of the discrepancy. I wasn't sure how to handle it, so I preferred the recalculated passer rating as I thought it was more accurate to describe the league.
     
  31. Finster

    Finster Finsterious Finologist

    3,087
    2,038
    113
    Jul 27, 2013
    I don't think you can use NFL avgs to gauge any 1 team, because the avg is a conglomerate of all the teams, with all their variances, e.g., any individual stat avg is far off from the top or the bottom, like TD passes, 2015 avg is 27, 36 was the high, and 14 was the low for 16 game starters.

    We can compound that thought even further, lets call 26, 27 and 28 the mean, not one QB threw for any of those numbers, all were either above or below the mean, which is why historically, NFL avgs are only used as "checks" for teams, because there is always a wide variance from team to team.

    So I would bet that if you did all teams, there would then be a "group" in which Miami would fit.
     
    roy_miami likes this.
  32. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Yeah, taking the average of several QB's passer ratings isn't the same as combining all the data and then computing the passer rating. They are only the same when the number of attempts for each QB are identical. A simple example should illustrate:

    Suppose we simplify the passer rating formula to just completion percentage (what is true for this one component will be true for all others). Let's say QB #1 has 5 completions out of 8 attempts, and QB #2 has 7 completions out of 12.

    If passer rating is simply completion percentage, PR1 = 5/8 and PR2 = 7/12. Now, if you average those two, you get (1/2)*(5/8 + 7/12) = (1/2)*(15/24 + 14/24) = (1/2)*(29/24) = 0.6042. If instead you combine the data, you get (5+7)/(8+12) = 12/20 = 0.6, so a small difference, but still not the same.

    But try the same thing where attempts are the same and you get no difference. Say it's PR1 = 5/10 and PR2 = 7/10. Averaging you get (1/2)*(5+7)/10 = 0.6 and combining you get (5+7)/(10+10) = 0.6.

    So the question is which is better? It depends on the question you are asking. If you want to know the average passer rating of a single QB (that is, you randomly pick a QB), then you average the passer ratings. If instead you want to know the average passer rating over some period or in some condition is (note: NO mention of a randomly chosen QB), then you combine the data.

    So filtering by attempts or starts etc... is fine as long as you stick to saying something like "average for a starting QB". But if you want to talk about the average passer rating in a year, or the average passer rating when leading/tied vs. trailing (again note: no mention of average for a randomly chosen QB), then you use the stats I linked to.
     
  33. dolphin25

    dolphin25 Well-Known Member

    6,338
    2,400
    113
    Nov 22, 2014
    but, you could argue that the defenses adjusted to the QB's strengths and the QB could not overcome those. Thus the strategy was still good.
     
    Finster likes this.
  34. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    That's right. There are simply too many possibilities regarding the level of optimality of different strategies consistent with the data that we can't test. This is precisely why I prefer a more parsimonious explanation (that simply notes the effect of a high-percentage, pass-heavy offense with Tannehill) that doesn't even make reference to how optimal the strategy was.
     
    roy_miami likes this.
  35. rdhstlr23

    rdhstlr23 Season Ticket Holder Club Member

    14,074
    11,142
    113
    Dec 2, 2007
    Chicago, IL
    I have been Ryan Tannehill's staunchest critic, but how can you not watch those clips and come away with no hope?

    I'm all in on this coaching staff. I really believe that with this coaching, the draft picks they've made to put talent in front of him/around him, that Tannehill will finally take off. And when I say take off, I don't mean all these absurd numbers people put up 30+ TDs and 10 INTs, but I mean he will let the natural talent you see in those highlights flourish.

    I think/hope we'll see a guy more comfortable in the pocket, because there is time. I think/hope we'll see a guy with more creativity/ingenuity to make plays. And I also think you're going to see a gameplan week to week that brings about mismatches to attack a defense.

    Ryan Tannehill has been in this purgatory of putting up acceptable stats, but contributing to this team's inability to move above mediocrity.

    This is the year. And it may not be all him. Honestly, I can see him putting up the same stats that he did in 2014, or 2015 and this team still improves by 2-4 wins because this is a coaching staff I believe in. And you're damn right that myself, and most, will package it up as Tannehill "finally taking the next step".

    Best offseason move in years was hiring Adam Gase, and acquiring the level of coaching talent we did, sans Chris Foerster. But like in SF, he has too much talent to **** it up.
     
    resnor and cuchulainn like this.
  36. Pauly

    Pauly Season Ticket Holder

    3,696
    3,743
    113
    Nov 29, 2007
    If your strategy is rely on Plan A and hope the opponent does not adjust you don't have a strategy you have a hope. As Clauswitz put it No plan survives contact with the enemy.
     
  37. Pauly

    Pauly Season Ticket Holder

    3,696
    3,743
    113
    Nov 29, 2007
    I hadn't though that there would be such a difference between averages and re-calculating the rating.

    I have gone back to my original data and done a bit more poking around. Doing it properly this time.

    Tier 1 (top 8)
    Combined passer rating when ahead: 103.2
    Combined passer rating when behind: 98.0
    average passer rating when ahead: 103.3
    average passer rating when behind: 99.6
    Standard deviation ahead: 8.3
    Standard deviation behind: 8.0

    Tier 2 (middle 14)
    Combined passer rating when ahead: 93.4
    Combined passer rating when behind: 89.8
    average passer rating when ahead: 91.1
    average passer rating when behind: 90.4
    Standard deviation ahead: 11.5
    Standard deviation behind: 7.3

    Tier 3 (bottom 8 + Tannehill)
    Combined passer rating when ahead: 81.2
    Combined passer rating when behind: 83.5
    average passer rating when ahead: 81.5
    average passer rating when behind: 81.7
    Standard deviation ahead: 15.8
    Standard deviation behind: 11.8

    Combined NFL starters
    Combined passer rating when ahead: 94.9
    Combined passer rating when behind: 89.4
    average passer rating when ahead: 91.5
    average passer rating when behind: 90.3
    Median when ahead: 93.2
    Median when behind: 92.2
    Standard deviation ahead: 14.4
    Standard deviation behind: 11.0

    Because I think the NFL passer rating formula is really sensitive when there is a low number of attempts
    QBs with 150+ attempts in both categories (14)
    Combined passer rating when ahead: 97.3
    Combined passer rating when behind: 90.2
    average passer rating when ahead: 96.7
    average passer rating when behind: 91.1
    Standard deviation ahead: 9.6
    Standard deviation behind: 7.4
     
  38. Pauly

    Pauly Season Ticket Holder

    3,696
    3,743
    113
    Nov 29, 2007
    I had made a transposition error in this part
    Correct data is:
    NFL ahead
    4648 of 7279 for 54,310 yards, 374/151
    63.9%; 7.5 ypa; 94.9 rating
    NFL Behind
    4671 of 7463 for 54,556 yards, 341/188
    62.6%; 7.3 ypa; 89.4 rating

    My wife was telling me to get off the computer and do chores so I didn't double check before posting. Sorry about that chief.
     
    cbrad likes this.
  39. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    No problem! Yeah, it's always important to double check data.

    Anyway, for the future, I'd suggest continuing to filter by sufficient number of attempts as long as you talk about average passer rating for a QB. However, note that no such filtering is required (or justified) if you talk about average passer rating by condition.
     
  40. dolphin25

    dolphin25 Well-Known Member

    6,338
    2,400
    113
    Nov 22, 2014
    but if the QB cant complete that strategy, does that make the strategy bad? his first year the deep ball was in the offense, that was taken out because it could not be completed, short game implemented, defenses started sitting on short stuff, where else is there to go?
     

Share This Page