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

Take Away His 'X' Best Carries and He's Average

Discussion in 'Miami Dolphins Forum' started by DevilFin13, Apr 3, 2015.

  1. Tannephins

    Tannephins Banned

    1,818
    572
    0
    Dec 23, 2014
    In that situation you could stop there. No other analysis would be needed.

    True, but we don't have that situation here. Here, we have variation.

    Is that variation a function of how many times Miller carries the ball? Not when the outlier Lions game is omitted.
     
  2. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    No, the point is the method doesn't give you the correct answer in principle. You need a method so that if each individual game correlation is X, then the final result of your analysis should be X.
     
  3. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    btw Tannephins, what you did answers a different question. It answers whether # of carries changes the relationship we're interested in. Remember, we're interested in how Miller performs as a function of # of carries per game, NOT whether how he performs as a function of # of carries per game changes as a function of # of carries per game.
     
  4. Fin D

    Fin D Sigh

    72,252
    43,684
    113
    Nov 27, 2007
    No, I'm saying that the discussed assumption (that Miller just wears down/is less effective after X amount of carries) wasn't accounting for the fact that teams run more if they have the lead in the second half. Also, Miller is not the only RB on the team that would get carries in the 3rd and/or 4th quarters.
     
  5. Tannephins

    Tannephins Banned

    1,818
    572
    0
    Dec 23, 2014
    The problem is that you're mixing two different analyses here. The first one is the correlation between the number of the rush and the yards gained on it, and the second is whether there is a correlation between the total number of rushes and that correlation.

    If there is no variation among the correlations in the first analysis, there is no need to do the second one. There is a need to do the second one only if there is variation in the first, and if there is variation in the first, we can look at what that variation is or is not a function of.
     
  6. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    The bolded part is all I responded to, and in the case of the Dolphins there's (weak) evidence that's not true. Also, whether Miller is the only RB or not doesn't matter as long as all other things are assumed to be equal (Miller can be assumed to take X% of the carries).
     
  7. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    It doesn't matter if there's no variation among the correlations in the first analysis. The result still must make sense. You can't say a stat is an unbiased estimate only if you cherry pick the data. Also, as I pointed out in post #123, you're also asking an entire different question than the one this debate is about.
     
  8. Fin D

    Fin D Sigh

    72,252
    43,684
    113
    Nov 27, 2007
    It does matter. You said his carries weren't significantly more or less in the second half. If the team is running more, whether Miller or Thomas or Player X is in, that causes the defense to expect the run and stop it better on the whole. So his runs and everyone else's runs may be less effective and lower the YPC.

    This is one of things where you're too buried in stats and not considering anything else.
     
  9. Tannephins

    Tannephins Banned

    1,818
    572
    0
    Dec 23, 2014
    Why wouldn't that be the question of interest, with the hypothesis being that he performs either better or worse on later versus earlier carries as a function of his total number of carries?
     
  10. Tannephins

    Tannephins Banned

    1,818
    572
    0
    Dec 23, 2014
    I'm not sure if you're understanding the first analysis that was done. Here's an example from one game in 2014:

    [TABLE="width: 192"]
    [TR]
    [TD="class: xl66, width: 64, bgcolor: transparent"]Patriots
    [/TD]
    [TD="class: xl64, width: 64, bgcolor: transparent"]1
    [/TD]
    [TD="class: xl65, width: 64, bgcolor: transparent"]13:05
    [/TD]
    [TD="class: xl64, width: 64, bgcolor: transparent"]1
    [/TD]
    [/TR]
    [TR]
    [TD="class: xl66, width: 64, bgcolor: transparent"]Patriots
    [/TD]
    [TD="class: xl64, width: 64, bgcolor: transparent"]2
    [/TD]
    [TD="class: xl65, width: 64, bgcolor: transparent"]12:28
    [/TD]
    [TD="class: xl64, width: 64, bgcolor: transparent"]2
    [/TD]
    [/TR]
    [TR]
    [TD="class: xl66, width: 64, bgcolor: transparent"]Patriots
    [/TD]
    [TD="class: xl64, width: 64, bgcolor: transparent"]3
    [/TD]
    [TD="class: xl65, width: 64, bgcolor: transparent"]13:52
    [/TD]
    [TD="class: xl64, width: 64, bgcolor: transparent"]1
    [/TD]
    [/TR]
    [TR]
    [TD="class: xl66, width: 64, bgcolor: transparent"]Patriots
    [/TD]
    [TD="class: xl64, width: 64, bgcolor: transparent"]4
    [/TD]
    [TD="class: xl65, width: 64, bgcolor: transparent"]14:28
    [/TD]
    [TD="class: xl64, width: 64, bgcolor: transparent"]6
    [/TD]
    [/TR]
    [TR]
    [TD="class: xl66, width: 64, bgcolor: transparent"]Patriots
    [/TD]
    [TD="class: xl64, width: 64, bgcolor: transparent"]5
    [/TD]
    [TD="class: xl65, width: 64, bgcolor: transparent"]7:22
    [/TD]
    [TD="class: xl64, width: 64, bgcolor: transparent"]13
    [/TD]
    [/TR]
    [TR]
    [TD="class: xl66, width: 64, bgcolor: transparent"]Patriots
    [/TD]
    [TD="class: xl64, width: 64, bgcolor: transparent"]6
    [/TD]
    [TD="class: xl65, width: 64, bgcolor: transparent"]4:38
    [/TD]
    [TD="class: xl64, width: 64, bgcolor: transparent"]13
    [/TD]
    [/TR]
    [TR]
    [TD="class: xl66, width: 64, bgcolor: transparent"]Patriots
    [/TD]
    [TD="class: xl64, width: 64, bgcolor: transparent"]7
    [/TD]
    [TD="class: xl65, width: 64, bgcolor: transparent"]4:14
    [/TD]
    [TD="class: xl64, width: 64, bgcolor: transparent"]5
    [/TD]
    [/TR]
    [TR]
    [TD="class: xl66, width: 64, bgcolor: transparent"]Patriots
    [/TD]
    [TD="class: xl64, width: 64, bgcolor: transparent"]8
    [/TD]
    [TD="class: xl65, width: 64, bgcolor: transparent"]9:27
    [/TD]
    [TD="class: xl64, width: 64, bgcolor: transparent"]9
    [/TD]
    [/TR]
    [TR]
    [TD="class: xl66, width: 64, bgcolor: transparent"]Patriots
    [/TD]
    [TD="class: xl64, width: 64, bgcolor: transparent"]9
    [/TD]
    [TD="class: xl65, width: 64, bgcolor: transparent"]8:54
    [/TD]
    [TD="class: xl64, width: 64, bgcolor: transparent"]1
    [/TD]
    [/TR]
    [TR]
    [TD="class: xl66, width: 64, bgcolor: transparent"]Patriots
    [/TD]
    [TD="class: xl64, width: 64, bgcolor: transparent"]10
    [/TD]
    [TD="class: xl65, width: 64, bgcolor: transparent"]7:37
    [/TD]
    [TD="class: xl64, width: 64, bgcolor: transparent"]6
    [/TD]
    [/TR]
    [TR]
    [TD="class: xl66, width: 64, bgcolor: transparent"]Patriots
    [/TD]
    [TD="class: xl64, width: 64, bgcolor: transparent"]11
    [/TD]
    [TD="class: xl65, width: 64, bgcolor: transparent"]7:04
    [/TD]
    [TD="class: xl64, width: 64, bgcolor: transparent"]2
    [/TD]
    [/TR]
    [/TABLE]

    The correlation between the number of the rush and the yards gained on it in that game was 0.12.
     
  11. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    What? First of all, one can test a claim without adjusting for every possible thing that theoretically could matter. I did that. There's no question whatsoever that I performed a test of your claim, though as I admitted up front it's not very strong evidence against it.

    That you can't see it's logically speaking a test is not my problem.
     
  12. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    It's like the difference between a 1st and 2nd derivative. We're interested in the slope, not the change in the slope. The problem with pointing out the change in the slope is it doesn't tell you what the slope is.
     
  13. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Yes, assuming your calculations are correct, I know what a correlation is.
     
  14. Fin D

    Fin D Sigh

    72,252
    43,684
    113
    Nov 27, 2007
    No you didn't.

    Let's say Miller runs 5 times a quarter. The team can still call 8 runs in each of the first & second quarters and 15 run plays in the 3rd and 20 in the third.

    Your "test" doesn't account for that. You aren't considering that the increased runs by Miller and the other RBs in the second half would cause the opposing defense to better stop the run.

    And, as I said, you aren't seeing that because you watch football on MS Excel.
     
  15. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Yeah, I could just repeat what I already said. You're just pointing out a possibility, with no evidence for it btw. Could a more detailed analysis invalidate the simple one I performed? Sure. So? I assumed all else was held constant, which is something you can do btw. and still test a hypothesis.

    So, I assumed that whatever distribution of carries Miller has as a % of the total number is constant over the game. Is it true? In aggregate, who knows. Point is, I performed a valid test of your hypothesis. And your comment about watching football on MS Excel (a program I hate) means nothing since you are only going to be able to provide evidence about % of carries by Miller as a function of game time through STATS!
     
  16. jdang307

    jdang307 Season Ticket Holder Club Member

    39,159
    21,798
    113
    Nov 29, 2007
    San Diego
    All this debate between posters on a forum, meanwhile in the real world a reporter asked why he got weaker as a runner as the carries piled up. Instead of denying there was a problem he said he doesn't know why, all but admitting that the stats are accurate.

    So instead of taking the stats at face value, and Miller's word at face value, we want to talk about correlation. Well alright then.

    What's curious is they gave Moreno 24 carries in game 1, but Miller never reached 20, and only 4 games with more than 15. Games of 18, 16, 19 and 19.

    Now, that might be due to circumstances but I think he can, and should be a 20 carry back, not a 15 carry back.For whatever reason this coach and two coordinators haven't agreed yet.
     
  17. resnor

    resnor Derp Sherpa

    16,329
    9,874
    113
    Nov 25, 2007
    New Hampshire
    I don't think him saying "I don't have any idea" had really any meaning whatsoever. Maybe he disagreed with the question, but didn't care to get in an argument.

    Does looking at carries 11-20 have a negative effect when he rarely gets to 20? What I seen to recall was that in the second half of games, when the run game stalled on a couple runs, they just completely seemed to abandon it. Hard to get your numbers up with 3 or 4 carries in a half.
     
  18. jdang307

    jdang307 Season Ticket Holder Club Member

    39,159
    21,798
    113
    Nov 29, 2007
    San Diego
    Isn't that the crux of this whole debate ;)
     
  19. resnor

    resnor Derp Sherpa

    16,329
    9,874
    113
    Nov 25, 2007
    New Hampshire
    No...the crux is the statement that he gets worse as he gets more carries. The run game stalling, in my opinion, is different. Send out a patchwork line, and you get a run game that stalls when the line can't block in crucial situations late in games when defenses know what's coming.
     
  20. jdang307

    jdang307 Season Ticket Holder Club Member

    39,159
    21,798
    113
    Nov 29, 2007
    San Diego
    But Lamar Miller this year was very good on 3rd and 1/2, and 4th and 1/2. Too small of a sample size? Maybe as it's just a handful of chances but it was much better than last year.
     
  21. resnor

    resnor Derp Sherpa

    16,329
    9,874
    113
    Nov 25, 2007
    New Hampshire
    I think you'd find that late in games, well, second half of games, that the run game as a whole, struggled on 3rd and 1/2.
     
  22. jdang307

    jdang307 Season Ticket Holder Club Member

    39,159
    21,798
    113
    Nov 29, 2007
    San Diego
    His aggregate numbers looked okay but yes I'd have to see the splits.
     
  23. Brasfin

    Brasfin Well-Known Member

    2,435
    1,672
    113
    Apr 27, 2013
    Brazil
    Completely agree with this. When we needed to run the ball late in games and keep the drives going to seal the win, we couldn't.
     
  24. jdang307

    jdang307 Season Ticket Holder Club Member

    39,159
    21,798
    113
    Nov 29, 2007
    San Diego
    So ... You guys are saying Miller got worse later in the games. ;)
     
  25. Fin D

    Fin D Sigh

    72,252
    43,684
    113
    Nov 27, 2007
    You haven't provided evidence either because your numbers aren't taking anything being discussed into account. You're misunderstanding the argument and arguing based on that misunderstanding.

    There's no basis for that assumption. You literally just made up that distribution of carries across quarters would be constant.

    I'm not an anti stat person so you're last declaration is pointless. My problem with your stance is that you look at nothing but stats. My problem with other people's arguments is that they look at nothing other than tape. They need to be used together. Conventional wisdom (like knowing teams try to run out the clock when they have the lead) is every bit as important to making an evaluation as factoring up YPC.
     
  26. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Fin D, this is what you said in the post I responded to:

    I think some are forgetting that if we have the lead in the 3rd/4th quarter, we're going to be running more and the defense knows it, so the YPC averages should go down and it could have nothing to do with Miller "wearing down".

    That's all I responded to. Now, don't tell me my numbers aren't taking anything discussed into account. They clearly do and they (indirectly) test whether it's true that the Dolphins run more in the 4th quarter if we have a positive or negative point differential at the start of the 4th quarter. There's no misunderstanding involved.


    You're worried about the assumption that all else is held constant, for example distribution of carries. In math, statistics and science, if you assume you know nothing about something except for the set of possibilities, a uniform distribution (probability of every possibility is the same) is adopted. Now technically speaking you're right that this is "made up", so why is it adopted universally? One of the best justifications is based on information theory: the uniform distribution maximizes entropy, which is a standard measure of uncertainty in systems. Either way, assuming the distributions are equal across quarters (or anything else) is standard in math and science IF we assume we know nothing about the distribution.


    Finally, you must read very few posts of mine if you think all I do is look at stats. I've been at the forefront of arguing against Tannephins' over-reliance on stats in multiple threads, as I think a bunch of other posters can attest.
     
  27. Fin D

    Fin D Sigh

    72,252
    43,684
    113
    Nov 27, 2007
    Jesus.....yes or no, are there other running backs on the team? If that is all you responded too, then your numbers (based on the false assumption that carries are a constant for one RB) don't actually deal with what I said.

    Its like I said most Americans like pizza and you countered with the perceived eating habits of one guy based on a assumption that he eats the same thing every day of the week.

    You are over relying on stats right now. You're actually trying to tell me we don't try to run the clock down with a lead, because science dictates that since we should know something about anything, we might as well make up a given RBs carries are constant from quarter to quarter. It is absurd.

    Just think it through. Even if Miller's carries are a constant 5 carries a quarter, the team can still run more than 5 attempts a quarter. If you understand that simple freaking premise, then you'd realize your argument is without a point.
     
  28. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Jesus back to you! Yes there are other RB's on the team. No, neither of us know what the distribution of carries is over quarters. Therefore, I assumed I didn't know and did a valid analysis of it.

    And what part of weak evidence do you not understand? I admitted it's only weak evidence, but at least it's evidence. I also said in my first post that your argument was common wisdom for NFL teams, but that there's evidence it's not true for the Dolphins.

    Do you even understand basic logic??
     
  29. Fin D

    Fin D Sigh

    72,252
    43,684
    113
    Nov 27, 2007
    Weak evidence is not reason enough for you to make an argument. I have decades of conventional wisdom from the NFL backing up my assertion that we, like other teams, try to run out the clock. That carries significantly more weight than you literally fabricating a foundation to build an argument off of that doesn't even really tackle my point.

    This is easily one of the more absurd arguments I've had on this board which is saying quite a bit since I get suckered into ridiculous arguments all the time.

    Now, if you bring actual numbers that show we don't run out the clock then fine, I was wrong, but until that happens the way you've decided to "disprove" or argue what i said is ridiculous.
     
  30. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Show me ONE TIME I said your hypothesis is wrong!

    Do you even understand that it's possible to have evidence FOR and evidence AGAINST a hypothesis at the same time???

    Get that in your head first before attacking my argument, which is that I provided evidence (albeit weak) your hypothesis is wrong for the Dolphins. Does that mean it is? Who knows, but as long as you keep claiming I didn't provide evidence I'm not respecting your thought process one bit.
     
  31. Fin D

    Fin D Sigh

    72,252
    43,684
    113
    Nov 27, 2007
    Then all you're doing is argue just to argue.
     
  32. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    No, I provided actual data as evidence, which is something you never did.
     
  33. Fin D

    Fin D Sigh

    72,252
    43,684
    113
    Nov 27, 2007
    Once again you provided crap data that was no more relevant to my point than if you had provided average wind speeds of hurricanes that never made landfall.

    Oh look, you used 2 commas in your last post. That's weak data, must mean I'm right.
     
  34. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    It's not crap data because it actually tests your hypothesis. ANY data that tests a hypothesis can't be "crap" relative to determining whether that hypothesis is correct or incorrect.
     
  35. resnor

    resnor Derp Sherpa

    16,329
    9,874
    113
    Nov 25, 2007
    New Hampshire
    This is ridiculous. Basing conclusions of Miller's performance off of data that includes runs by other backs is pointless. Using data that doesn't apply, to test a hypothesis, indeed means it's crap data. Miller having very few runs in the third/fourth quarters could just as easily be a result of the line not being able to block as it is that Miller gets worse on more carries. I'd even think it's more likely that the line degrading in known running situations is more to blame for Miller looking worse. Would also make sense why Miller wouldn't want to comment on it, saying "I have no idea", instead of throwing his line under the bus.
     
    Fin D likes this.
  36. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Wow.. can't believe the types of arguments here. So you'd say any analysis done on a "treatment" group and a "control" group is "crap" if we know that neither group is a perfect "treatment" or "control" group? That is, by taking samples, we find out that there are individuals in each group that we would ideally not want there.

    OK, so you and Fin D think that any analysis done on such data is just "crap". Sorry, man that's just stupid. You estimate what you can assuming you know only some subset of what you'd like to know (the rest is part of the "all else assumed to be equal" part).

    I guess you and Fin D think practically every study done with human subjects is thus "crap". Sorry, no respect for this way of thinking. And I won't argue it too much more either if you guys can't comprehend something that simple.
     
  37. Fin D

    Fin D Sigh

    72,252
    43,684
    113
    Nov 27, 2007
    Actually what is "stupid" as you put it, is comparing your admittedly weak analysis to other studies.

    Data without context is pointless. Data applied incorrectly to a hypothesis is crap. Stop lecturing me on what is or is not proper data, as you clearly have no idea.
     
  38. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    I applied it correctly. You just don't get it. I think it's best to end our argument here on this subject.
     
  39. Fin D

    Fin D Sigh

    72,252
    43,684
    113
    Nov 27, 2007
    lol, ok man.


    Unrelated to our argument.....

    I drank 4 bottles of water today so far. It reached 80 degrees here today. Based on that data, I'm a millionaire. Granted, my evidence is weak, but hey I provided data.
     
  40. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014

    If you can show that a millionaire is more likely than a non-millionaire to drink 4 bottles of water on a day where it reached 80 degrees, then I'd agree that would be weak evidence you're a millionaire. But can you show that?
     

Share This Page