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

Doughty love

Discussion in 'Miami Dolphins Forum' started by dolphin25, Jul 31, 2016.

  1. ckparrothead

    ckparrothead Draft Forum Moderator Luxury Box

    79,599
    159,162
    113
    Dec 1, 2007
    If you say so but that's not what I had heard or read.
     
    Fin D likes this.
  2. ckparrothead

    ckparrothead Draft Forum Moderator Luxury Box

    79,599
    159,162
    113
    Dec 1, 2007
    Can't find any tape of that distance throw. Can however find tape of Cardale Jones slinging the football 74-75 yards as the crow flies (once you account for horizontal distance) about a year and a half ago.

    [video=youtube;IANqp8X2DJI]https://www.youtube.com/watch?v=IANqp8X2DJI[/video]

    Gonna say I kind of doubt that Doughty's distance throw matched that.
     
  3. rdhstlr23

    rdhstlr23 Season Ticket Holder Club Member

    14,074
    11,142
    113
    Dec 2, 2007
    Chicago, IL
    The fact we have 11 pages dedicated to our 7th round pick and current 3rd string QB says a lot about our fan-base.

    I hope Doughty makes it. Outside of him beating the living hell out of Marshall and ruining our 2014 season, it was really fun to watch him at WKU. With that said, I just hope everyone is OK should Brandon Doughty not make it in the NFL. We seem, as most fan-bases do, to have crazy love affairs with our UDFA/late round 3rd string QBs.
     
    LI phinfan likes this.
  4. ckparrothead

    ckparrothead Draft Forum Moderator Luxury Box

    79,599
    159,162
    113
    Dec 1, 2007
    That's true of any team that hasn't had a franchise quarterback in more than a decade.
     
    Fin-O likes this.
  5. SlyFoxxx

    SlyFoxxx Sorry for Party Rocking

    95
    40
    18
    Jul 29, 2011
    Hard to take the conversation serious when distance thrown is the measure used for arm strength. I don't think one NFL representative would look at distance thrown as a top 5 "arm strength" quality, at least after Kyle Boller happened. But to each their own I guess.

    I might have missed it in all the other nonsense but Dolphin25, if Doughty has no obvious flaws like athleticism and arm strength, why do you think he went in the 7th?
     
    resnor, LI phinfan and Fin D like this.
  6. ckparrothead

    ckparrothead Draft Forum Moderator Luxury Box

    79,599
    159,162
    113
    Dec 1, 2007
    Rigged!!!

    Sorry couldn't resist. But I agree distance throw isn't the measure of arm strength. Velocity at a range of distances, velocity from a range of platforms and throwing circumstances, those are the measures. Consistency is also an important measure because some guys don't get the same heat on the ball every time even over the same distance. That's why I brought up two throws Doughty had in college that were virtually the same throw, virtually the same distance, one had a "D" quality to it and the other a nice solid "B" as far as velocity goes.
     
    resnor and Fin D like this.
  7. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Not disagreeing because I don't know this stuff, but one should remember from physics that velocity at the time of release is directly related to the distance any object travels. Only other considerations are outside of the control of the QB and deal with acceleration due to gravity and air resistance.

    So reading this as a disinterested 3rd party observer, it's somewhat funny to me because "velocity at time of release" and "distance thrown" will be highly correlated with each other over many trials. I guess if you had to choose I'd say "velocity at time of release" is better if you want a measure of the person's ability independent of external influences, but if you want a measure of how well whatever the QB is doing translates to the football field, I personally would think a combination of "distance thrown" and "time to impact" would make more sense.

    But like I said.. I don't know this stuff so I'm not pushing a position (except what physics says).
     
    dolphin25 likes this.
  8. ckparrothead

    ckparrothead Draft Forum Moderator Luxury Box

    79,599
    159,162
    113
    Dec 1, 2007
    Velocity problems in a player can actually be pretty varied, complex. It's not as simple as, "he's got a weak arm". His 20 and 30 yard throws (as the crow flies) could have enough heat on them but his 40 yard throws end up with higher lift and they lose energy. That's pretty common actually. There's a heat on most NFL 30 yard throw that some guys can get on their 40 yard throws before they hit the cliff in average velocity. But many guys if they're throwing 40 yards forget it, there's a massive drop-off in average velocity from their 30-35 yard throw to the 40-45 yard throw. That's where the biggest average velocity cliff is located, somewhere around the 35 to 45 yard range. I'd say every QB has one. I don't think any of them have a completely smooth, linear average velocity graph. The steepness of that cliff, as well as its location, can be big indicators of a guy's arm talent.

    And that's before you even get into stuff like how much comes off his passes when he has to throw on the run or without stepping in, etc.
     
    resnor likes this.
  9. ckparrothead

    ckparrothead Draft Forum Moderator Luxury Box

    79,599
    159,162
    113
    Dec 1, 2007
    You've got to take into account things like ball spin, the axis of the ball's spin, and arc. It's a pretty complex subject. It's not just velocity at release plus extraneous factors.
     
    resnor, Fin D and cbrad like this.
  10. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    That makes sense. But then why use velocity at release as a measure of arm strength if even more QB-dependent factors are important in where and when the ball lands? To me this would weaken the argument for using velocity at release. Isn't how far the ball travels and how long it took to get there in different situations ultimately more important?
     
    dolphin25 likes this.
  11. ckparrothead

    ckparrothead Draft Forum Moderator Luxury Box

    79,599
    159,162
    113
    Dec 1, 2007
    Here are just SOME of the physics concepts that go into an NFL pass.

    https://www.youtube.com/watch?v=c6EguTZFK5s

    Particularly note that it's not just ball spin which for a Drew Brees is at 600 RPMs, but also the "wobble". Surprisingly, a little bit of wobble is actually perfect for keeping the ball on target and immune from wind gusts. The perfect amount of wobble is 3 wobbles per 5 ball spins.
     
    cbrad likes this.
  12. ckparrothead

    ckparrothead Draft Forum Moderator Luxury Box

    79,599
    159,162
    113
    Dec 1, 2007
    I personally didn't use velocity at release. If you read what I posted above I prefer to evaluate average velocity at different distances, building a graph based on that, and if I have enough data (and time) locating the location and steepness of what I've personally deemed to be a cliff where your average velocity drops a lot more steeply. It's usually somewhere between 35 and 45 yards.
     
    Fin D likes this.
  13. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Well.. whether you use velocity at release or average velocity at different points in the flight path, the question is what is the ultimate objective? It should be where and when the ball travels. If you can directly measure that, that should be the preferred measure.

    Unless the objective is something else. I mean.. who cares about flight path if the ball is on target at the right time?
     
  14. Finster

    Finster Finsterious Finologist

    3,087
    2,038
    113
    Jul 27, 2013
    Velocity and distance can be 2 separate measures of arm strength, not always though, and not completely separate ever.

    There is no such thing as a guy who can throw the ball 70 yards that has a weak arm, but there are guys that can't throw the ball 70 yds but do have great velocity.

    So it's really best said that distance isn't the only measure of arm strength.
     
    dolphin25 likes this.
  15. ckparrothead

    ckparrothead Draft Forum Moderator Luxury Box

    79,599
    159,162
    113
    Dec 1, 2007
    You seem to be confused about the objective altogether. You care about the quality of a player's arm strength because you're evaluating his ability to execute passes in the future, which hasn't happened yet. Graphically representing a guy's average velocities over a number of distance ranges will help you to evaluate his ability which in turn helps you to predict his future performance and general ability to execute NFL passes.

    As for caring about the flight path, you care about that because you also want to know what areas you can coach the player into improving. If the player is angling his ball wrong and you can improve his process mechanically, you should aim to find that out.
     
    resnor and Fin D like this.
  16. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    The 2nd argument about why you'd care about flight path I'll accept, but I don't yet agree with your first argument.

    You have more than enough trials to map out the space of how well the QB can throw on target at the required time in different situations, so interpolation among the actual distance + time measures should be the most accurate way of predicting what he'll do, not some intermediate measure of average velocity. Unless I'm missing something.
     
  17. ckparrothead

    ckparrothead Draft Forum Moderator Luxury Box

    79,599
    159,162
    113
    Dec 1, 2007
    I think you overestimate both the amount and quality of the data. The reason you do velocities into ranges is because you want to be able to average a number of them. Throw velocity is not consistent even if it's the same quarterback and the same range and the same circumstances. If we go with your approach and have time data instead of average velocity data, then the time data points are only able to be mathematically treated with one another if the distances are exactly identical, which is never the case (hence what I said about overestimating the amount of data).

    You won't have five 100 foot throws to where you can average the times. That would be nice, I suppose. If you had five of them you could average together, say 1.31 seconds (52.0 mph), 1.28 seconds (53.3 mph), 1.29 seconds (52.9 mph), 1.30 seconds (52.4 mph) and 1.32 seconds (51.7 mph) then you'd get 1.30 seconds (52.4 mph) and feel good about that.

    The problem is you're not going to have five 100 foot throws. You're going to have a 100 foot throw at 53.0 mph, a 95 foot throw at 51.5 mph, a 104 foot throw at 52.5 mph, etc. To me, because the varying quality of throws even in the same circumstances, that's why you convert the number into something distance-neutral before you treat them further.
     
    Fin D likes this.
  18. jdang307

    jdang307 Season Ticket Holder Club Member

    39,159
    21,798
    113
    Nov 29, 2007
    San Diego
    Yes! Thats the only reason I remember him.

    But a few other QBs have done the same. I think Aikman did as well.
     
    dolphin25 likes this.
  19. jdang307

    jdang307 Season Ticket Holder Club Member

    39,159
    21,798
    113
    Nov 29, 2007
    San Diego
    My head just asploded
     
  20. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    No, using the approach I'm suggesting you don't average anything before fitting, making it more accurate. For each throwing (or game) condition, the data is a function of two parameters: "time" and "distance". All you do now is fit a 2-D manifold (a surface) to the data. It's literally the best way of estimating what will happen for unknown "time + distance" points as long as you have sufficient data.

    The reason I suspect you have sufficient data is because QB's practice so many throws in so many conditions (and you can use game data too). Hard to argue there isn't enough data. And the quality of the fit will be reduced if you FIRST average mean velocities as you're suggesting. Best to use the raw data and then fit the best polynomial through it.

    So.. like I said, I don't know what the NFL currently uses, but IF the objective is to predict how well a QB can throw to a particular location at a particular time, then given the number of throws QB's make in practice, it seems based on the information given that what the NFL should do is what I'm suggesting. Software for doing that abounds. I personally use Matlab for most simple data analysis but there are tons of software for doing that.
     
  21. ckparrothead

    ckparrothead Draft Forum Moderator Luxury Box

    79,599
    159,162
    113
    Dec 1, 2007
    I think it's bad assumption that you can find a polynomial trend function to fit the raw data points like that. I see why you would want to but I don't think the human performance is accurately described by a polynomial function like that.

    And no you don't have good data on this because all you have is game film. The NFL can't have these guys throwing 200 balls in a practice session in order to populate this ideal data sample, especially since these teams don't exactly share information with one another, since the information is a source of competitive advantage.
     
    Fin D likes this.
  22. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    What? That first sentence is nonsense. This is just an exercise in predicting data points at unknown points based on known points. Once converted into numbers it's irrelevant whether the data was obtained from humans or not. So once you have the data, you simply use the best fitting method.

    And using a proxy like mean velocities only increases uncertainty, not just because you are first averaging (you shouldn't do that), but also because you introduce uncertainty between the relation of mean velocity and time + location of impact.

    Finally, the amount of data in both approaches is the same (because you're mapping mean velocities to time + distance), so you can't argue one method is better because of the amount of available data.
     
  23. rdhstlr23

    rdhstlr23 Season Ticket Holder Club Member

    14,074
    11,142
    113
    Dec 2, 2007
    Chicago, IL
    I'm surprised DJ hasn't hopped in yet. He can really give some detail into the specifics on the physical attributes associated to throwing strength, power, etc. He breaks it down pretty well.
    Hop in DJ?
     
  24. ckparrothead

    ckparrothead Draft Forum Moderator Luxury Box

    79,599
    159,162
    113
    Dec 1, 2007
    The amount of data available affects how well you can communicate the findings. The communication of the findings is important for comparability and so is the complexity of the presentation.

    I don't understand how my first sentence doesn't foot. You're taking inconsistent and unknowable human performance in inconsistent and unknowable atmospheric conditions, throwing inconsistent and unmeasured real distances due to arc choice, with the only knowns being horizontal distance and time, and you expect to find a polynomial equation that predicts the time measurement at unrepresented distances?

    And going back to comparability, what advantage does a polynomial function give you in comparing the graphs of different prospects? What useful information are you gleaning from something so complex and opaque? The unknowns in the exercise are already so great that simplification for the sake of comparability is more important than introducing uncertainty in the treatment of the data.

    I prefer my methods on this.
     
    Fin D likes this.
  25. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Yeah, so regarding that first sentence. You can always find a polynomial fit to any data to minimize error between predicted and actual up to the theoretical maximum. The problem is actually that you can do that! That is, what you want to be wary of is arbitrarily increasing the number of terms in the polynomial just so you can "fit" the data as well as possible. Doing that is frowned upon in science because it means you're just using a fitting technique and not explaining anything.

    But.. you CAN fit the data. That's the reason I said it was nonsense (as in literally nonsense).

    Regarding communicating the results. That's as you point out very important and very easy with such a fit. It might look something like this:
    http://www.mathworks.com/help/examp...tsandPlotFitShowingExcludedDataExample_03.png

    Or it might look like something else but it'll be a surface that you can rotate around to show people how things look based on different parameters. You can also show slices through the data, holding one parameter fixed while varying the other.

    Point is, whenever possible you want to use the measures you're trying to predict (time + location of impact) and not some other intermediate measure where you add the uncertainty of the relation in addition to all the uncertainty present in the target data.
     
  26. ckparrothead

    ckparrothead Draft Forum Moderator Luxury Box

    79,599
    159,162
    113
    Dec 1, 2007
    Given that average velocity is a pure mathematical treatment of the two variables (time, distance)...why would you even consider it an intermediate measure that adds uncertainty to the relation? Especially if the velocity at distance communicates the relationship in so much more of an understandable and comparable way.

    No offense to your graph but nobody is going to look at it and say oh yeah that helps me evaluate this guy's arm strength. Whereas if I draw two line graphs representing average velocity in the y-axis and distance range in the x-axis, that would be much more intuitively understood.
     
    Fin D likes this.
  27. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    There are two problems with using average velocity. First, you have to try to predict time + location from average velocity, and as you pointed out earlier there are so many other factors like wobble and trajectory, or as I pointed out you have external factors like air resistance and gravity, that means you're inserting uncertainty into predicting the true time + location given the average velocities.

    The second problem is that you are losing information about the true time and distance information (note: that's different than the time + location of impact) when averaging. That is, you no longer are using the raw data when doing the fit. The fit will always work best when using the raw velocity data, not when you average at any location first.

    And yeah I understand that just showing that graph without further explanation isn't helpful for people not used to it, but you can take that graph and simplify it to ease communication. For example, over some range of distances it might be that a given QB's accuracy (if you choose to measure it that way) or time to impact changes linearly. Most people can understand linear relations. So one technique would be to divide up the space into parts easy to understand. Anyway.. point is communication is up to the person doing the communication and not the method of fitting.
     
    dolphin25 likes this.
  28. ckparrothead

    ckparrothead Draft Forum Moderator Luxury Box

    79,599
    159,162
    113
    Dec 1, 2007
    We're on different wavelengths here. Let's simplify this. You have the following ten data points. The x is distance and the y is time.

    Feet, Seconds
    60 , 0.79
    70 , 0.88
    80 , 1.01
    90 , 1.16
    100 , 1.31
    110 , 1.50
    120 , 1.70
    130 , 1.93
    140 , 2.17
    150 , 2.49

    You would choose to represent this graph exactly as above and find a polynomial best fit line to minimize the error. Incidentally the ability to expand the polynomials as far as you like in order to minimize the error is exactly what I was talking about when I said there's a real problem that way with comparability between quarterbacks, you're not actually communicating any information that is meaningful. But I digress. You'd take the above and create a polynomial function.

    I on the other hand would prefer to represent the data in this format, with the x being distance and the y being velocity. How does that insert uncertainty?

    Feet, MPH
    60 , 52
    70 , 54
    80 , 54
    90 , 53
    100 , 52
    110 , 50
    120 , 48
    130 , 46
    140 , 44
    150 , 41

    Now obviously for the reasons I mentioned you're not going to get those exact distance data points for every QB and so if you want to compare them I've put them into buckets. Something like...

    60 to 79 : 53.0
    80 to 99 : 53.5
    100 to 119 : 51.0
    120 to 139 : 47.0
    140+ : 42.5

    Which you could then compare to other quarterbacks in such a way as to be meaningful.

    Are you saying that rather than doing buckets we should be creating that polynomial function and using it to predict velocity at a certain distance and then compare those distances between players? If so my problem with that once again is that with so much inherent uncertainty involved, I question the predictive value of your polynomial especially if you're wary about creating too many terms. The inflections are hard to predict. It's one thing if you've got a ton of data but you never really do, when it comes to this. I think there could be enough of a gap in the data to miss an inflection point or two.
     
    Fin D likes this.
  29. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Ahh! Good you posted that. Yes, we were talking a bit past each other. I didn't know you still kept the distance information! After you pointed out the wobble stuff, etc.. I thought velocity data was being treated as one of many possible parameters one could use to predict time + distance in a model.

    OK, in that case there is no uncertainty you're inserting through modeling. So the only issue is the averaging and binning, and those aren't as problematic compared to the modeling uncertainty I thought was there.

    Averaging will generally not matter much if the variance divided by mean remains relatively constant so you can check that. Binning can introduce sample size issues, so you might want to make sure no bin has too few samples. Of course safest thing to do is not average or bin and just fit. Inflection problems you can deal with by choosing a polynomial with not too many terms.
     
    ckparrothead likes this.
  30. ckparrothead

    ckparrothead Draft Forum Moderator Luxury Box

    79,599
    159,162
    113
    Dec 1, 2007
    That's fair. Creating a polynomial of best fit and then using that function to create velocities at 20 yards, 25 yards, 30 yards, 35 yards...honestly I hadn't thought of that. I would have issues with the fit of the polynomial function but it might be cleaner than binning.
     
  31. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    You can plot how the mean squared error (or some other measure of variance of the data relative to the polynomial fit) changes as a function of number of terms in the polynomial. As soon as the error stops dropping too much more with an extra term you've probably got about as clean a fit as you'll find.
     
  32. dolphin25

    dolphin25 Well-Known Member

    6,338
    2,400
    113
    Nov 22, 2014
    But, YOU did not see. ........................
     
  33. dolphin25

    dolphin25 Well-Known Member

    6,338
    2,400
    113
    Nov 22, 2014
    AGAIN if you can read, I NEVER said he has the strongest arm, NOT have i ever said he isn't athletic, I did ask what defines athletic in terms of QB. Marino was slow, does that make him not athletic?

    How did Brady last till the 7th round? How did Kurt Warner go undrafted? What about Tony Romo?
     
    Finster likes this.
  34. dolphin25

    dolphin25 Well-Known Member

    6,338
    2,400
    113
    Nov 22, 2014
    I didn't say he didn't have *** his throw, I just said Doughty out threw him. I would think someone on a national event like that would want to compete at his best, I guess some do not. And, I would not want that guy on my team.
     
  35. Finster

    Finster Finsterious Finologist

    3,087
    2,038
    113
    Jul 27, 2013
    Yeah, Bill Walsh refuted the idea that Dan was not a mobile QB, his reasoning was that Dan was the best QB ever at moving around in the pocket, that his foot movement was sublime, so athleticism is not defined by speed.

    Usain Bolt is faster than Ashton Eaton, by a wide margin, but any notion that Bolt is a better athlete would be utterly preposterous.
     
    dolphin25 likes this.
  36. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Is it? Both are great athletes, but if I had to rank order athletes I'd for certain put Bolt ahead of Eaton because of his utter dominance.
     
    ckparrothead likes this.
  37. dolphin25

    dolphin25 Well-Known Member

    6,338
    2,400
    113
    Nov 22, 2014
    I think it is interesting discussion. What defines being athletic? It comes in many forms.
     
  38. jdang307

    jdang307 Season Ticket Holder Club Member

    39,159
    21,798
    113
    Nov 29, 2007
    San Diego
    Well first you have to be black. Then you are athletic even when you are not.

    Noemi you said hustle and high motor ...

    On a serious note I won't hold Doughty draft position against him. It's what he does now that matters. People are over drafted and under drafted.

    How did Jamarcus Russell go #1 despite many warning signs? Who drafted Romo?
     
    dolphin25 likes this.
  39. ckparrothead

    ckparrothead Draft Forum Moderator Luxury Box

    79,599
    159,162
    113
    Dec 1, 2007
    Draft position shouldn't be held against players but if their college careers justified their draft position then the college career and what they'd shown previously on tape should always be used as a context for further evaluation, IMO.

    The whole "athletic" thing is borderline ridiculous. If you're trying to communicate what kind of prospect Brandon Doughty is relative to other NFL quarterbacks, and you say, "athletic", then you made a mistake. Straight up. End of discussion.
     
    Fin D likes this.
  40. danmarino

    danmarino Tua is H1M! Club Member

    15,361
    20,989
    113
    Sep 4, 2014
    That wasn't because he had a "weak" arm. Leaf was prototypical.

    I'll take what Bill Polian said about Manning during his pro day than what anyone here says.


     
    Finster likes this.

Share This Page