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

New Coach and blow it up

Discussion in 'Miami Dolphins Forum' started by my 2 cents, Oct 31, 2018.

  1. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    A lot of stuff in that post!

    Well.. first of all a lot of science is similar to what you're seeing me do here with sports stats. You gather all kinds of data in real world conditions and start to see which component stats are correlated and/or good predictors of other stats, which correlations could be argued to be causations, and/or build models to test (some skip this step).

    So I guess first thing I'd point out is that no serious scientist reports correlations without considering causal relations. So the stuff about kneeling in the 4th quarter is a bit of a strawman. And it was through statistical analysis (a lot of which I've done here) that we saw the relation between point differential and rushing attempts, which btw only occurs from the 4th quarter, not the 2nd half per se (3rd quarter point differential has almost no effect on rushing attempts). So again.. it's through this kind of statistical analysis that you even have that kind of data to talk about, and it helps filter out hypotheses you might otherwise have.

    The second thing I'd point out is that you can't talk about how well stat X predicts stat Y without considering sample size. So it's kind of irrelevant to say passer rating isn't "reliable" when sample size is small, like over 3 games. That's obvious. I've repeatedly said you should at very minimum only look at passer rating over 150+ passing attempts. So don't ignore sample size in these arguments.

    Third, you're right that in some years we've had average QB'ing, measured by passer rating, while PPG has been way below average. The correlation between passer rating and points scored per game (PSPG) across NFL history is pretty darn high at 0.7777. Square that and you get 0.605, or 60.5% of the variance in PSPG is predicted by passer rating. Well.. that leaves 40% you can't account for. So sure it's not a perfect predictor, but pointing out cases where it didn't work doesn't change the overall predictive power of that stat.

    So all these stats ARE useful, even when over smaller sample sizes or in specific cases they may not seem to be good predictors. What you have to do is to see what else is going on in addition to what is being captured by passer rating (for example). I know some will say "injuries" is one example, but it's really not for this specific case because that affects passer rating too (by how much we don't know).

    Point is.. it's merely opinion as to how much injuries will affect key stats, be it passer rating, PPG or expected win%.
     
    Last edited: Nov 18, 2018
    Irishman, Surfs Up 99 and KeyFin like this.
  2. adamprez2003

    adamprez2003 Senior Member

    37,392
    14,745
    113
    Nov 27, 2007
    new york ciity
    thats just it, passer rating isnt a good measure of individual ability. QB ratings are offense ratings not quarterback ratings

    here is a breakdown of injuries by football outsiders AGL = (Adjusted Games Lost)

    The top 14 teams in AGL included 10 playoff teams, though the Super Bowl participants were close to average at 13th (Eagles) and 14th (Patriots). Super Bowl MVP quarterback Nick Foles was the Backup of the Year after what he did in the two championship games, but the Eagles also did a really fine job of overcoming the losses of Darren Sproles (13.0 AGL), Jason Peters (9.0 AGL), and Jordan Hicks (9.7 AGL) in addition to Carson Wentz(3.0 AGL). Danny Amendola did his best Julian Edelman (16.0 AGL) impersonation all year, but the Patriots could never find an impact play on defense in the big game without Dont'a Hightower (10.9 AGL) available. Malcolm Butler was available, but… (too soon?)

    The Saints (24th) were the only team ranked 20th or lower in AGL to make the playoffs last year.
    Long playoff streaks ended for the Seahawks (five years) and Packers (eight years). We know Green Bay has made the playoffs before with worse AGL rankings, but Aaron Rodgers was too serious a loss for Green Bay to overcome. The Seahawks had to deal with big injuries to Cliff Avril (12.0 AGL), Richard Sherman (8.1 AGL), Kam Chancellor (7.3 AGL), and they couldn't keep a running back upright all year. Seattle last ranked in the twenties in AGL in 2011 (27th), which was also the last time the team failed to make the playoffs.

    https://www.footballoutsiders.com/stat-analysis/2018/2017-adjusted-games-lost

    as you can see only one team made the playoffs ranked 20th or lower and 10 made the playoffs ranked 14th or higher. The Dolphins have been in the 20s in 2016 and 2017 and will be in 2018. Now these are stats that are useful and thats because they deal with the team and not the individual
     
    Surfs Up 99 likes this.
  3. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    First of all good job looking up the AGL. It's really important that we look at the entire league when talking about any stat.

    But like I said, if (adjusted) passer rating has a very high correlation with what most observers, from coaches, players to fans, think of are the best QB's in history once sample size is large, then that right there shows it DOES measure individual ability.

    Now.. obviously it's true that passer rating (or almost every stat in football) is technically a team stat. I'd describe passer rating as a "QB-led offense stat" since it is specific to both the QB and the offense. However, the logical (as opposed to empirical) reason passer rating ends up measuring QB ability is that some unknown X% of this team stat is due to the QB, meaning the team stat is to X% an individual stat. No one knows what X is, but with large enough sample size any value of X will be sufficient. What is "large enough" sample size? That's where you need the empirical evidence, and I think what I stated before about how adjusted passer rating correlates with who is considered the best QB's (over large sample size) demonstrates this unknown X is large enough for practical purposes.

    So yes.. technically passer rating is a team stat, but both in theory and in practice it does measure QB ability (with sufficient sample size).
     
    Surfs Up 99 likes this.
  4. KeyFin

    KeyFin Well-Known Member

    10,488
    12,821
    113
    Nov 1, 2009
    I agree with everything you said (and learned a few things).....BUUUUTTTT....I also have a massive problem with it. If a three-game sample is too small to make a prediction either way, and we haven't had the same offense on the field since that three-game stretch, then how can we predict anything useful for 2018?

    For instance, let's say we want to evaluate whether or not RT should be the starting QB in 2019. It's almost pointless looking at his past 36 games since we're skipping almost 3 years to do that. We need to judge him on what we see at the end of the season AND we're going to have to use that small sample size of the first three games plus whatever he plays down the stretch.

    You're saying its not enough...yet it has to be enough because that's all we have.

    Maybe there's not a mathematical formula that can deliver a "moneyball" type answer in this case, but surely there are more exact traits we can find in the numbers to make better predictions. I mean, if there's nothing we can use to predict how the next 6 games will turn out and nothing to evaluate small sample sizes, then what are we really doing that's helpful?
     
    resnor and Surfs Up 99 like this.
  5. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    The way I've argued that in the past is to show stats like adjusted passer rating as a function of years in the league, or age, or years since the first time the QB started 6+ games, etc.. I even separated QB's into two different "tiers", with one tier being similar to the first 3 years we saw from Tannehill, etc.. I did this before 2018 started and showed that on average QB's don't get better from year 6 onwards (until you wait long enough so that the only QB's left are the better QB's.. so from 10+ years the average is a bit better).

    I can post those graphs again if you want, but that's how you get the sample size necessary to get at least a decent estimate. Tannehill is an average QB overall.. statistically speaking he's unlikely to improve, so I'd make my decisions based on that.
     
    Pauly, Surfs Up 99 and KeyFin like this.
  6. adamprez2003

    adamprez2003 Senior Member

    37,392
    14,745
    113
    Nov 27, 2007
    new york ciity
    Coaches dont care about QB rating. That's for fans and media. Coaches watch film. You think belichik called Bert Jones the greatest QB he had ever seen because of his 78 QB rating? Or Bill Walsh saying Greg Cook because of his 87 rating?
     
    Last edited: Nov 18, 2018
  7. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Given how often coaches and GM's get it wrong (see the draft) I'm not sure how much credence you want to put in them w.r.t picking QB's. Regardless, no one is arguing you should ONLY use stats. I never said your opinion on injuries is "garbage". But mine isn't either. And there are many different paths to success, not just one.

    I also doubt that coaches and GM's completely ignore stats, including passer rating. That would actually seem foolish to me. I'd predict (not sure if accurate) it's a lesser version of the effect Moneyball has had in baseball. Stats show you things you would have otherwise missed, they suggest doing X rather than Y, but in the end you use whatever information you think is useful and run with it.
     
    texanphinatic and Surfs Up 99 like this.
  8. Surfs Up 99

    Surfs Up 99 Team Flores & Team Tua

    1,950
    1,785
    113
    May 5, 2016
    I just want to say (again) nice discussion! I am learning a lot!! I have a question for ya'll. Is it safe to say a QB who makes those around him better (like Drew Brees) should see that reflected in their personal AND in the team's various offensive stats? Whereas, a QB (like Tannehill ) who usually doesn't make those around him better and more productive will probably see it reflected only in their personal stats, but not in the team's various offense stats? I hope my question makes sense. Forgive me if it doesn't.
     
    cbrad and KeyFin like this.
  9. KeyFin

    KeyFin Well-Known Member

    10,488
    12,821
    113
    Nov 1, 2009
    I think that's too complicated to answer. For instance, Landry had a billion catches last season for around 900 yards. Did Cutler make Landry better by going to him more often....or did Landry make the team worse by not doing more with the football on those billion catches? What would have happened if Cutler divided those passes up evenly to Grant, Stills, Parker and Fasano?

    Likewise, was Drake so effective because Landry caught a billion passes and everyone expected Landry to get the ball every down? If the answer is yes, then we have Landry to thank for Drake's success and Cutler to thank for Landry's. But how many of Drake's yards were due to fantastic line blocking or play action and how many were pure talent? How many were due to poor LB or DE reads that gave him those first four yards....we can keep going down the rabbit hole with new wrinkles and twists to look at.

    It's very tough to say whether RT makes a player better or worse because we only get to see what happened...not what could have happened with "x" variable (better/worse ball placement, a better/worse block, throwing the pass sooner/later, etc.). We DEFINITELY know RT makes Stills better because he's only missed him deep once or twice the entire year. But I'm not sure how we'd prove or disprove the rest.
     
    cbrad and Surfs Up 99 like this.
  10. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    As KeyFin pointed out it's hard to measure whether a QB makes those around him better. Here the fact that passer rating is technically a team stat really makes it hard to test because you don't want to treat it as an individual stat for the question you asked.

    I will say this though. You can look at the game-by-game correlations of passer rating to points scored for both Tannehill and Brees just to see how those are related (how much of points scored is or is not captured by passer rating for that particular QB).

    For Tannehill the average correlation per year (have to do it "per year" because otherwise passer rating inflation creeps in) is 0.6880 while for Brees it's much lower at 0.6045. However, it's lower for Brees primarily due to a single year: 2017 where unbelievably the correlation is negative at -0.2678 (meaning that the higher Brees' passer rating the fewer points NO scored!). Remove that outlier and Brees' passer rating has an average correlation of 0.6590 to points per game. It's also interesting that both those game-by-game correlations are lower than the 0.7777 I quoted earlier when you use team passer rating. So I don't know if both Tannehill and Brees are simply lower than average or if game-by-game correlations with starters is fundamentally different than team passer rating (could be either.. have to look into this further).

    How to interpret those correlations is another matter altogether and you guys and go after it if you want. I'd really like an explanation for 2017 Brees. That -0.2678 is just baffling. Here's Brees' stats in 2017:
    https://www.pro-football-reference.com/players/B/BreeDr00/gamelog/2017/

    Either way, it's a great question, just one that can't be answered until you have an independent metric for "the degree to which a QB improves his surroundings". We don't have such a stat yet.
     
    Surfs Up 99 likes this.
  11. adamprez2003

    adamprez2003 Senior Member

    37,392
    14,745
    113
    Nov 27, 2007
    new york ciity
    i dont think QBs really make the players around them better other than if you have a very dangerous QB you are probably going to be playing more nickel and dime and therefore the RB should theoretically benefit. They will make players statistics better but not the player himself. i think offensive coordinators can make players better or worse however depending on scheme and what is asked of them. Jerry Rice and Joe Montana became HOF players in a west coast offense but would they have been hall of famers if they played in a Norv Turner vertical offense? I dont think so
     
    resnor and Surfs Up 99 like this.
  12. Surfs Up 99

    Surfs Up 99 Team Flores & Team Tua

    1,950
    1,785
    113
    May 5, 2016
    Thanks Key, Cbrad, and adamprez! I guess that is why we all love football. Any given Sunday, right?
     
    cbrad, adamprez2003 and KeyFin like this.
  13. KeyFin

    KeyFin Well-Known Member

    10,488
    12,821
    113
    Nov 1, 2009
    The first thing that jumps out at me is that New Orleans had 23 rushing TD's (and 23 passing TD's) in 2017. His biggest games statistically (week 3 and 8) had 239/149 and 286/101 in passing/rushing yards. His 5th best game statistically was against the Bills where he threw for 184 and the team rushed for 298. So they were a pretty balanced team that didn't rely heavily on Brees arm.

    Example- his 2nd biggest game in yardage was week 7 against GB...he threw for 324 yards and zero TD's....although they scored 26 points. He also had 2 TD's against Carolina in week 11 but threw for 375 yards....accounting for only 14 of the team's 31 points. These numbers tell me the Saints rely heavily on their running backs in the red zone and get solid production from them. In half of Brees games in 2017, he threw zero or 1 TD. The only time he tossed 3 TD's was in week 3 (which leaves seven 2-TD games).

    Brees worst game on the entire season was a 77.9.....in a game the team won 52-38. They only had 379 total yards on offense so there had to be multiple defense/special teams scores going on there. Their 2nd highest scoring day was against the Bills w/ that 298 yard rushing day. In these two games combined for almost 100 points, he threw 2 total TD's.

    Plain and simple, these stats tell me that NO is a balanced team that doesn't rely heavily on their QB to make things happen. And that's part of the reason why Brees is so darn productive.
     
    Pauly, Surfs Up 99 and resnor like this.
  14. adamprez2003

    adamprez2003 Senior Member

    37,392
    14,745
    113
    Nov 27, 2007
    new york ciity
    i should add that once in a while QBs can make offensive lines look good or bad. Jamarcus Russell made his line look worse than it did because he would often times not drop back the proper amount of steps. If the play called for a 5 step dropback he would drop 7 or 3. Well what happens in that case is the OTs are blocking like its a 5 step dropback and if the DE cuts inside at 3 they are taught to push him but let him since he's going to be short of the dropback and effectively taking himself out of the play. Well unfortunately they were pushing him into Russell and it was Russell's fault but every amateur blamed the oline. Conversely someone like Dan Marino had such a quick release that he could make an average Tackle look like a hall of famer statistically just by getting rid of the ball so fast
     
    resnor and Surfs Up 99 like this.
  15. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Well.. the only issues with that are: 1) we're actually talking about a negative correlation, and 2) all other years are between 0.4435 and 0.8252. For example, in 2018 (before today's game) they have 21 passing TD's and 17 rushing TD's (fairly balanced) yet the correlation between Brees' passer rating and points scored is 0.6368.

    Point #1 is actually the real issue. Having a lower correlation is one thing, but negative??? You can be as balanced a team as you want it's just hard to imagine how increasing any of completion%, TD%, Y/A or decreasing INT% would lead to scoring less. Doesn't matter how balanced a team you have that just shouldn't occur.

    And point #2 really asks you to argue why 2017 was the ONLY year Brees ever played with a "balanced offense". Hard to argue IMO. This negative correlation (point #1) is actually such a mystery that I'm more comfortable just saying it was a statistical anomaly and being done with it. I can't make sense of it any other way.
     
    Surfs Up 99 likes this.
  16. Irishman

    Irishman Well-Known Member

    573
    532
    93
    Oct 16, 2017
    High Point, NC
    upload_2018-11-18_17-36-42.png
     
    Surfs Up 99 likes this.
  17. Irishman

    Irishman Well-Known Member

    573
    532
    93
    Oct 16, 2017
    High Point, NC
    This post is a breath of fresh (and rational) air at a point in a very difficult season for the Dolphins. Our team is either over performing due to circumstances or under performing due to circumstances. The end of the season should provide us with better information relative to cause and effect.
     
    Surfs Up 99 and cbrad like this.
  18. KeyFin

    KeyFin Well-Known Member

    10,488
    12,821
    113
    Nov 1, 2009
    Like I said, the games where Brees threw more were his lowest scoring games (personally, that is). His best statistical games were the ones where the running game dominated. In essence, the less he had to do, the more they scored. So while the TD's were balanced between run/pass, evidently the play on the field suggested they should have just kept running.

    I agree with you that it's very strange, especially since his QB rating on the year is 103, but I have a feeling that (1) they would pass into the red zone and then run it in the rest of the way and (2) Brees had very few deep ball TD's. I think you're right that it's an anomaly that doesn't mean anything, but your stats are trying to say that the team would have won just as much (or even more) with a different QB...even though that's not what our eyes tell us.
     
    resnor and Surfs Up 99 like this.
  19. Surfs Up 99

    Surfs Up 99 Team Flores & Team Tua

    1,950
    1,785
    113
    May 5, 2016
    Love that shirt! We just got a Golden Retriever puppy. One of the things I love about her, other than being one cute puppy, is that she doesn't bark very much. Here is a picture of Ginger that my wife took yesterday. :-)

    Ginger1.JPG
     
    KeyFin, resnor and Irishman like this.
  20. Irishman

    Irishman Well-Known Member

    573
    532
    93
    Oct 16, 2017
    High Point, NC
    I hope Ginger blends well into your family.

    We had basenjis for 35 years and really enjoyed them, when they weren't outsmarting us. I recommend obedience training as much for the opportunity to really relate to your dog and for the benefits everyone gets from a well trained pet.

    There are a lot of other dog related activities available and I hope you get a chance to look into them. The most impressive thing I ever saw was the performance of herding dogs.

    The most fun I ever had was letting my dog run in open field lure coursing. When my little Basenji (14" at the shoulder) blew most of the other dogs off the course it was impressive. It took a while for the Afghan people to get friendly. They felt bad when my little Basenji (who was roughly the size of an Afghan's head)would dramatically outrun the Afghan hounds. The two fastest dogs on the course was a Whippet named "Willie" and my Basenji who was named "Bagel".

    It's been a few decades since then. Thanks for getting me away from wondering about the Dolphins and remembering some of the good old days watching my dog fly across the fields. She could get up to about 35 MPH. That's basically covering 1/2 the distance between 2nd & 3rd base in a second. She covered that distance in a single stride at the rate of one double suspended stride per second.
     
    Surfs Up 99 likes this.
  21. AGuyNamedAlex

    AGuyNamedAlex Well-Known Member

    3,582
    2,579
    113
    Sep 12, 2015
    Ill be honest, at this point I don't know if Gase is a good or bad coach. Its likely he falls into the "competent" category.

    His first season we made the playoffs. Whether you consider it a fluke or not, it happened and it was with Gase at the helm. At the very least, we know a team CAN make the playoffs with Gase leading them.

    Year two was a disaster. I feel like almost anything would have been better than that Cutler circus. His passion wasn't in the gane and it showed. I at least have to put some of this on Gase, but everyone makes mistakes and you hope they learn.

    This year, in my eyes anyway, is much better than last but the injuries are just killing us. Its a lot worse to lose starting offensive players in waves IMO because that side of the ball requires much more chemistry. I'm not saying defense requires none, but its easier to plug and play guys IMO.

    In other words....I find it extremely hard to judge Gase based on results at this point.
     
    resnor and Irishman like this.

Share This Page