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

A long-winded post comparing Miamit to good teams.

Discussion in 'Miami Dolphins Forum' started by DolphinGreg, Sep 16, 2015.

  1. DolphinGreg

    DolphinGreg Season Ticket Holder Club Member

    5,227
    6,527
    113
    Dec 7, 2014
    Cbrad,

    Many thanks for doing the additional data collection! I’d be happy for us to continue to look into this. This weekend I’ve got some friends flying in though so I’ll probably not be on the forum.

    Anyhow…

    You probably already realized this, but it bears being said for the sake of clarity that we need to be careful with averages. The averages amongst offensive and defensive rankings will naturally move towards the middle (most likely the upper middle) in sufficiently large data sets.

    The challenge is to build a model smart enough to tell the difference between a team that is great at one thing while being bad at the others and a team that is average at all things. One of those teams may be more successful than the other.

    In other words, while there are two ways I can choose to dodge an oncoming train, taking the average of the two would be unwise! ;)

    For example, in the 22 SB teams, the average ranking for passing, rushing, pass defense, and run defense were 11[SUP]th[/SUP], 15[SUP]th[/SUP], 13[SUP]th[/SUP] and 10[SUP]th[/SUP], respectively. We see this natural clustering towards the upper-middle in all stats. While those are averages, they may not be great predictors because (1) you probably need to be ranked higher than 10[SUP]th[/SUP] in your best category (21 of 22 teams were) and (2) you probably don’t need to be ranked top-15 in all categories. For example, only 3 of the 22 teams managed to be top-15 in all four categories. Ironically, the 2006 Bears and the 2010 Steelers (who both lost) were 2 of those 3 teams.

    So, there is apparently value not in trying to be slightly above average in everything but rather in standing out significantly in one or two categories. At least that was true of SB teams. It’s almost as though great success really does come through specialization as opposed to balance.

    So, we need to be careful that while Miami may resemble the average, that average may not exist in reality and may just be some meaningless middle ground between the two or more opposing clusters which each represent specialized (but proven) paths to success. I suspect that if we include all Play-off teams that we’ll probably see more “average” teams, but amongst SB teams it was less apparent—so that may be some later predictor of Play-off success, idk.

    I’m not saying that’s proven, it’s just something we need to be aware of. For example, this effect gave me pause as to how to read into the opening stat-lines I used in my initial post.

    Even when I showed the average Super Bowl winning QB threw for 3,900 yards, what I may have neglected to mention was that the standard deviation was nearly 600 yards. That’s quite a bit. The reality is a lot of guys were throwing for 3,500 yards and several were throwing for around 4,500 yards. The guys that threw for 3,500 yards were on teams led mostly by defense. The guys who threw for around 4,500 were generally on teams that featured great passing offenses. So while Tannehill does in fact resemble the average, the average apparently doesn’t actually get to the SB very often.

    In that case the average is kind of lying. While Tannehill’s numbers are nearly identical to the average, the average wasn’t a good predictor. What my initial post really showed was that Tannehill’s stats are in an awkward middle ground. They’re better than the numbers put up by guys we consider to have been benefitted by great defense (Wilson, Eli, Roethlisberger, Kaepernick, Grossman, etc.) while they weren’t really as good as the guys we think of as having carried their teams (Peyton, Rodgers, Brees, Brady, etc.).

    Doubting that Tannehill’s future lies with the latter group, it’s easy to see why I suggested I thought that Miami needed to get serious about building its defense. I think a lot of people would actually agree that Tannehill is an awkward, middle-ground kind of QB.




    Now, you asked a very good question. If passing is such a great predictor of success (i.e. 50% of Super Bowl teams were top-10 in passing), why does rush defense seem more important than pass defense?

    For example, amongst SB teams, great pass defense showed up 8 times while great run defense showed up 13 times. In fact, only 2 teams had a great pass defense without having a great run defense, suggesting that run defense may in fact be a prerequisite of sorts for great pass defense.

    If great run defense were a perquisite for great pass defense though, how do we explain the anomalies we see? In 2006, the Colts were 2[SUP]nd[/SUP] against the pass but 32[SUP]nd[/SUP] against the run. You can’t be much more of an outlier than that! In 2010, the Packers were 5[SUP]th[/SUP] against the pass and 18[SUP]th[/SUP] against the run. Heck, in 2014, the Dolphins were 6[SUP]th[/SUP] against the pass and 24[SUP]th[/SUP] against the run!

    Based on that I’m inclined to believe that neither is a prerequisite for the other but we’d need a bit more info on that to be conclusive.

    On a side note, it’s interesting that the 2010 Packers resemble the Dolphins quite a bit. Rodgers’ 2010 wasn’t far off from Tannehill’s 2014. Why did the Packers win the SB beating one of the most well-equipped teams on my list (the 2010 Steelers) while Miami missed the Play-offs entirely!? That’s neither here nor there, but it’s interesting.

    I don’t have a statistical answer for you. There may be logical hypotheses for it but I think what we see is probably in some sense a remnant of the game and how it's structured. It could just be that there are fewer defense players great against the run than the pass making the talent pool more competitive, idk. I’ll have to ponder that.
     
    Finster likes this.
  2. Rock Sexton

    Rock Sexton Anti-Homer

    2,553
    1,793
    113
    Mar 14, 2015
    [​IMG]
     
    cbrad likes this.
  3. DolphinGreg

    DolphinGreg Season Ticket Holder Club Member

    5,227
    6,527
    113
    Dec 7, 2014
    It could be that pass defense is simply more to do with scheme than talent where as run defense is more to do with talent than scheme. In that scenario we'd expect to see the better teams be consistently good at stopping the run all year while they pass defense would get better as the games became more important. For example, we've all seen instances of Bill Belichick and Rex Ryan (great coordinators) find ways in big games of turning average pass defenses into impressive squads that somehow seem to up their competency--at least in terms of dealing with great QBs.

    Just a hypothesis, but it's at least logical.

    Another idea might be that if teams really viewed there being 2 paths to success, one in which they were carried by a great passing game and one in which they were carried by great defense and rushing, that we might expect great run and pass defenses to be relatively split, virtually mirroring the distribution of great run and pass offenses which they have to counter.

    As I've said, there were 11 of 22 teams that seemed to be passing oriented. There were 8 defenses that seemed to be great against the pass. That's somewhat close.
    There were 8 teams that featured great running games. There were 13 defenses that seemed to be great against the run. That's a little more lopsided.

    But, in this case, the theory might be stronger than the evidence. As Tannephins pointed out, using total yards is sort of the easy way out. For example, if a team has a bad rush defense, teams may not even have to throw against them thus leading to low numbers of pass yards given up. Miami might be an example of that late in the season. They were giving up so much on the ground, their 6th ranked pass defense may have been inflated a bit simply because teams weren't testing that element of their defense as much.

    Therefore lopsidedness in defense may be a critical sign of BS. For example, in 2006 the Colts were 32nd against the run. Undoubtedly teams just ran a lot against them. That surely helped them attain their elite pass defense numbers which ranked 2nd.

    Do we really believe that was true? Probably not.
     
  4. Piston Honda

    Piston Honda Well-Known Member

    7,853
    8,088
    113
    Sep 23, 2014
    Which is why stats are useless without context. Those Colts teams didn't stop the run, not couldn't stop the run. Their strategy for winning was that Peyton and the offense would score enough to force teams into throwing to keep pace. They built a defense focused on pressuring the QB and forcing TOs, and it worked. Now, some really smart guy will come along and say that stopping the run isn't important bc the Colts won 12 games a year without doing it and that guy would be right, while entirely missing the point.
     
  5. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Yeah Greg.. I'm almost at the point where I will do a more careful analysis (instead of just averages, etc...), specifically to look at the distributions of a whole bunch of stats. This question is just nagging me. I love how you started this analysis, but I want to really know what's going on here. The big question is whether there's an easy-to-download database so that I can just write a program to do the analysis instead of punching in numbers by hand.

    We'll see.. but damn I can't get this question out of my mind! haha!
     
  6. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    All true, but what makes this problem so interesting is there doesn't seem to be an obvious "context" that explains it.
     
  7. Finster

    Finster Finsterious Finologist

    3,087
    2,038
    113
    Jul 27, 2013
    That was a really good comment by Greg, I think the run D becomes more important late in the year when the weather is worse, and in the playoffs where all the teams are good, and the ability to prevent teams from converting on short yardage, and also putting teams in 3rd and 5+ situations.

    Stopping the run is about consistency, but stopping the pass can be done by sporadically disrupting the passing game via pressure, especially if you can regularly produce a passrush in certain situations via the blitz packages.

    The one very consistent thing among all these defensive stats however, is the level of QB play it takes to win the SB, high.
     
    djphinfan likes this.
  8. Tannephins

    Tannephins Banned

    1,818
    572
    0
    Dec 23, 2014
    How are you defining importance? Remember that if you're using team rankings in terms of rushing yards allowed, you have the confound of how many rushes the opposing teams had. Again, yards per carry is the better measure. I suspect that if you use yards per carry surrendered, that measure will pale in comparison to offensive and defensive YPA in terms of the variables most strongly related to winning.
     
  9. Rock Sexton

    Rock Sexton Anti-Homer

    2,553
    1,793
    113
    Mar 14, 2015
    The run D is very important late in the year when teams/players are worn down from the grind of the season.
     
    Finster likes this.
  10. Piston Honda

    Piston Honda Well-Known Member

    7,853
    8,088
    113
    Sep 23, 2014
    I agree, because context lies in the inner workings of each individual team. Very few teams have a Peyton, ARod, etc, those who don't have elite QBs have to play by more traditional means and methods.
     
  11. Piston Honda

    Piston Honda Well-Known Member

    7,853
    8,088
    113
    Sep 23, 2014
    The best teams imo are the ones who can A) play one style extremely well or B) play a multitude of styles fairly well. Look at the Pats last season. In the divisional round they threw 50+ passes and barely ran it a dozen times. The very next week they ran it 40 times and bludgeoned the Colts. Then in the Super Bowl they were pass happy again. The offense was effective all three weeks. They can adapt to their opponents and field conditions in a way that no other team can.
     
    Brasfin likes this.
  12. Kud_II

    Kud_II Realist Division

    3,662
    1,404
    113
    Oct 15, 2011
    Seneca, SC
    The title of this thread made me LOL literally. And not because I think we're a good team.
     
  13. KeyFin

    KeyFin Well-Known Member

    10,488
    12,821
    113
    Nov 1, 2009
    Question Greg- are these stats for the entire seasons of each Super Bowl appearance, or just the playoffs? Because I have a feeling that that's the key here to decoding the numbers and how they really matter in a trip to the Super Bowl.

    Why do I say that? Because you're taking averages over 19 games and trying to normalize a pattern, but we also know that bad teams have made it to the Super Bowl. For example, the Giants were an average team when they toppled New England in the Super Bowl a few years back, yet they did enough of something in the post season to get that opportunity. What was that one thing? Eli' passing? A solid run defense? I don't know...but I think factoring out their ranking in the post season would explain a lot. And when you clump all those stats together across ten years and 20 teams, those kinds of insights fade away into averages (like you mentioned earlier).

    I personally do not believe that there's a way to make a magic formula that would identify the actual "x factor" for Miami's success, but it seems like the best path would be to back up a few steps, show the four main rankings for each super bowl team, and then find an analysis there. Going back to Eli's match-up vs the Pats, where did each team rank in the four stats? Which stats proved to be the most important for victory? Did the winner do anything to negate their stats in that game or throughout the post-season (for example, a bad run team really stuffing the run)? I think that's where the true insight is.
     
    Piston Honda likes this.
  14. Tannephins

    Tannephins Banned

    1,818
    572
    0
    Dec 23, 2014
    The dependent variable in any sort of analysis like this should simply be win percentage. Super Bowl appearances are a low base-rate phenomenon. What makes teams win. That's what we're looking for here.
     
  15. Piston Honda

    Piston Honda Well-Known Member

    7,853
    8,088
    113
    Sep 23, 2014
    Scoring more points than the other team makes teams win. The rest is all variable. You can look for commonalities till you're blue in the face but every team is different and the idea that there's a singular path to success is weak, at best.
     
    Fin D likes this.
  16. Fin D

    Fin D Sigh

    72,252
    43,684
    113
    Nov 27, 2007
    ....unless you Belicheat.
     
  17. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    OK, I just wasted (or maybe not?) 4-5 hours of my life to write code to parse all kinds of data from http://www.pro-football-reference.com/ which is an awesome site that allows you to download all kinds of football stats (look under "seasons" for offensive and defensive stats).

    Anyway, what I did was look at the distribution of Y/A (yards per attempt) for all playoff teams from 2002-2014 (reason for 2002 is that there were 31 teams in 2001 and that broke my code!). I looked at how many standard deviations away each playoff team was from the league mean (for that year) for all 4 conditions: offensive/defensive passing/rushing.

    Since the images were too big to upload here, I uploaded them to a free hosting website. Each green circle represents a playoff team, the big red circle is the Super Bowl winner, and the filled blue diamond is the league average (which is trivially at 0 in this graph because everything is being compared to the mean). You can see the results here:
    http://postimg.org/image/886p3gq8b/ (rushing Y/A)
    http://postimg.org/image/8d1tg0or3/ (passing Y/A)
    http://postimg.org/image/und4y7q31/ (opponent rushing Y/A)
    http://postimg.org/image/5vyp5vdmz/ (opponent passing Y/A)

    Basically, the result is this: when you look at Y/A, things are intuitive again. Rushing basically doesn't matter, as not only all playoff teams (green circles) are evenly distributed around the mean, but so are the SB winners (red circles). Passing matters a ton, as most of those green circles are well above the league average. So this part of the story jibes with what most of us know: it's a passing league.

    But look at the last two links there. Those are defensive stats. The rush defenses of playoff teams seem to be distributed more or less similarly to those of non-playoff teams. However.. the pass defenses of playoff teams are on average MUCH better than those of non-playoff teams (opponent Y/A for passing is well below the mean for almost all of them).

    So.. things are intuitive again. To become a playoff team you have to become more efficient against the PASS, not the run. This makes total sense because the greatest damage an offense can do is through passing, at least measured by Y/A.

    So Tannephins was right about Greg's result possibly being due to using NFL ranking instead of run/pass efficiency. Later on, I'll still follow up on Greg's idea of not just looking at averages (or even distributions) but looking at which combinations of offensive/defensive pass efficiency (measured by Y/A) best correlate with playoff performance. But that's for later... I spent too much time on this haha! (though I must say, the code I wrote is well suited for doing further analysis like this, so it won't take anywhere near as much time in the future).
     
    77FinFan likes this.
  18. KeyFin

    KeyFin Well-Known Member

    10,488
    12,821
    113
    Nov 1, 2009
    Okay, but the Super Bowl appearances are our "x"...what we're trying to solve for. That means its involved in every formula used here, and the "low base rate" means nothing. Nevertheless, two out of thirty-two make it to the big show every year and the OP was trying to gauge how they got there.

    Besides, if you go solely by win percentage, then you're looking at league standings and calling it a day. We already know that the team with the most wins had the best season for their division.
     
  19. adamprez2003

    adamprez2003 Senior Member

    37,392
    14,745
    113
    Nov 27, 2007
    new york ciity
    To win a Super Bowl you have to be at least competent in all phases of the game, running, passing, defending the run, defending the pass and special teams. The reason for that is the road through the playoffs will force you to win a at least one game where your strengths are taken away from you and you have to win the game with the weaker part of your team. Additionally, teams have to have some areas of dominance. On offense you better be able to either pass or throw exceptionally well and on defense either stop the pass or stop the run exceptionally well. The more dominant one area is (ie running the ball) the less talented the other side has to be to maintain effectiveness. Meaning if you have the best running game in the NFL, your passing game only has to be mediocre in talent to be effective. However if your running game is 5th best, you will probably have to be around 10th best passing to maintain the same level of effectiveness. Right now I dont see us as being dominant in any any area so we would have to be around 7th best in all the categories to win a super bowl which I dont see this year. I think its possible next year if we improve our offensive line dramatically and add a couple of pieces to the defense

    We do have a chance at the playoffs if our offensive line stays healthy and our dline does also but if they get hit with injuries it will be another year of waiting

    also the ravens team that won the super bowl recently was a completely different team from the beginning of the season to the end. Their offensive line didnt play well for the first hallf of the season and in the last four games of the regular season became dominant. Its why stats are a flawed way to look at the game. Too many variables

    for me the way to build a super bowl team is to have a dominant defensive line, a dominant offensive line and a top ten QB
     
  20. Tannephins

    Tannephins Banned

    1,818
    572
    0
    Dec 23, 2014
    Right, but by doing that, you're excluding the variation in winning associated with the poor teams in the league. What you want to do in this sort of analysis is show that the good teams do X, while the poor teams don't do X, or do X more poorly. When you're focusing on Super Bowl teams, you don't have any poor teams in the analysis.
     
  21. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Still stinging from the Jags loss.. but I said I'd finish the analysis, so here are the rest of the figures. They look at all combinations of offensive vs. defensive pass/run efficiency (in std's above or below the league mean) for all playoff or non-playoff teams from 2002-2014.

    The conclusion is that the biggest difference between playoff and non-playoff teams is that playoff teams tend to not only be more efficient than league average in offensive pass efficiency but also in defensive pass efficiency (measured in Y/A). No other combination is as strong.

    In the graphs below, a green circle represents a playoff team while a cyan circle is a non-playoff team (title of the graphs tells you too). What you need to look for is in which quadrant most of the playoff or non-playoff teams fall. Upper left quadrant is best because that represents high offensive efficiency (y-axis) and low opponent offensive efficiency (x-axis):

    http://postimg.org/image/61s9vowrx/ (passing offense vs. passing defense for playoff teams)
    http://postimg.org/image/gcxs92957/ (passing offense vs. passing defense for non-playoff teams)

    Note how for pass offense vs. pass defense, the playoff teams tend to be in the upper left quadrant (efficient in both offense and defense) while non-playoff teams tend not to be in the upper left quadrant. The other combinations don't show as strong a difference:

    http://postimg.org/image/dn03l7955/ (passing offense vs. rushing defense for playoff teams)
    http://postimg.org/image/tasb9tn5h/ (passing offense vs. rushing defense for non-playoff teams)

    http://postimg.org/image/tcbpxp0or/ (rushing offense vs. passing defense for playoff teams)
    http://postimg.org/image/akdeq4nal/ (rushing offense vs. passing defense for non-playoff teams)

    http://postimg.org/image/yle3nfeah/ (rushing offense vs. rushing defense for playoff teams)
    http://postimg.org/image/7mh9hwk6b/ (rushing offense vs. rushing defense for non-playoff teams)

    OK.. back to sulking about the Jags loss..
     
    Tannephins likes this.
  22. Tannephins

    Tannephins Banned

    1,818
    572
    0
    Dec 23, 2014
    YPA differential. Notice when the Dolphins were holding their own against the Jaguars with regard to YPA, the game was tied. When the Dolphins' YPA fell from nearly 10 to 4.9 in the fourth quarter, and the Jaguars' YPA remained high, the game was decided. Now you might have some idea why I've criticized the play-calling so vehemently during that period of time. By the way, the correlation between YPA differential and win percentage in the NFL is nearly 0.90.
     
    cbrad likes this.
  23. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    Yeah, you were right when you first suggested the apparent paradox of going after a team's relative weakness was due to using NFL rankings instead of YPA. I must say.. after doing this analysis, I have more confidence in the power of YPA than just reading about others claiming that. Thanks dude!

    Having said that.. I'm not sure I'm totally on your side about the play-calling issue (like some others in that other thread, I'm not certain more run attempts would have helped), but that's a separate issue from the utility of YPA.
     
    resnor and Tannephins like this.
  24. Tannephins

    Tannephins Banned

    1,818
    572
    0
    Dec 23, 2014
    It's awfully hard to argue they would've hurt, when the team had a 4.9 YPA and punted four times on four possessions, with very little movement of the ball on those drives. The only direction there is up, and you have the previous three quarters of much better offensive functioning to consider, when there was far more balance.

    Not smart to disagree with that in my opinion. It's objective evidence versus conjecture about what "would have" happened, and the objective evidence is awfully hard to dispute with any solid footing in this particular case.

    Of course, everyone has his own mind, however, and just as soon as someone feels as though you're trying to change it (whether you really are or not), he digs his heels in and clings to a position that may be absurd, in the name of being "his own person." ;)

    On the one hand, we sure shouldn't be "sheep" and take everything someone else says as the gospel, but on the other we should be flexible and permeable enough to accommodate quality information. The stuff you provided in your post above is a very good example.
     
  25. Fin D

    Fin D Sigh

    72,252
    43,684
    113
    Nov 27, 2007
    Its early so forgive me if this is incredibly stupid.

    How would rushing more in the 4th increase YPA?
     
  26. resnor

    resnor Derp Sherpa

    16,327
    9,874
    113
    Nov 25, 2007
    New Hampshire
    Fin, he seems to think that the massive 1.7 ypc that the Dolphins were gaining in the first three quarters, and the 3 runs per quarter that they ran, somehow kept the defense honest, and allowed the passing game to be efficient. He is, of course, ignoring the loss of Albert and Cameron, among other things, as the primary drivers of the passing game becoming unsuccessful in the fourth quarter.
     
  27. Fin D

    Fin D Sigh

    72,252
    43,684
    113
    Nov 27, 2007
    I mean just from purely a math level how does running the ball (as poorly as we did) actually raise the YPA, mathematically? Wouldn't it lower it?
     
  28. cbrad

    cbrad .

    10,659
    12,657
    113
    Dec 21, 2014
    I think you have to take a more detailed look than just aggregate Y/A. Here's the play-by-play:
    http://espn.go.com/nfl/playbyplay?gameId=400791680

    Go through the 1st half plays vs. 2nd half plays. First thing you notice is the big penalties in the 2nd half. Practically every drive (except the last one with almost no time left) had a 10+ yard penalty or a sack, etc.. That wasn't true in the 1st half. So the conditions are different. With such penalties, it makes it less logical to run the ball in the first place. Also, note that those big 2nd half penalties came as often during a run play as a pass play (though arguably the sacks wouldn't have occurred had those been run plays). And the 3rd quarter Y/A stats are greatly influenced by that first drive that resulted in a TD.

    So while I agree a good goal is to maximize Y/A, and in theory a balanced attack should make it more difficult for the opponent to predict your next play, the big difference between what seemed to "work" earlier and what didn't later IMO are the penalties.

    So my position is that it's not clear more run attempts would have helped.. or hurt. Just hard to say in this case because the conditions logically favored the pass more in the 2nd half.
     

Share This Page