I agree that the defense will get better in the seasons to come. I think they get depth in linebackers and secondary positions. Then the defense improves dramatically. But that's just my opinion.

Sent from my XT1254 using Tapatalk

 

Which stats is he referring to? I think the most important is points allowed per game. Vance Joseph was at Houston as defensive backs coach from 2011-2013.

In 2011 Houston's defense was ranked 4th in the league in points allowed per game and they had a 10-6 record with first place in the division. In 2012, Houston's defense was ranked 9th in the league in points allowed per game, had a 12-4 record and won the division. The ONE year Houston had bad stats was 2013 where they were 25th in points allowed per game. That was the year they went 2-14!

Same with some other stats like rushing yards allowed per game: 4th in 2011, 7th in 2012 and 23rd in 2013. It's only with passing stats where I see something they were good at in 2013: they were ranked 3rd that year.

Point is.. in general the better you are in major statistical categories the better your team will do. It's why correlations between such stats and wins is positive. So yeah I do think it's a real concern that we are ranked 3rd WORST in run defense!.. that better not be the reason we lose a playoff game!

 

Gase mentioned it too...it is the following:

1. Defense Red Zone TDs Allowed; we're 16th at 55.32%.
2. Scoring; we're 10th at 23 ppg.
3. 3rd Down Conversions Allowed; we're 1st at 34.8%.

 

Well.. I guess that's when a statistician might help. Yards gained/allowed is very weakly correlated to wins (<0.15 from 1990-2011) so it shouldn't be a "major statistical category" in Joseph's mind.

 

Any idea how the 3 stats they're focusing on correlates to wins?

 

Oh those are good stats for sure.

I don't know precisely for #1 and #2, but let's be clear: you get a perfect correlation to wins if you look at points scored minus points allowed, and #2 is half that equation so that's got to be really high. Red zone scoring is really high at over 0.6 but that's all scoring and not just TD's, so I guess TD's is a bit lower.

3rd down rate depending on year is in the 0.4 range. All good stats.

 

Thanks!

 

I've never understood the "yards allowed or gained" being a major stat. Points for and against are what matters the most.

 

Yeah.

In general, other than points for/against, efficiency stats are far better at predicting wins than volume stats. So "yards" and "attempts" each have low correlations, but "yards per attempt" say for passing is over 0.4, etc.. And stats like passer rating and ANY/A are just weighted efficiency stats (correlations well over 0.5 and often 0.6). So efficiency in general matters most.

 

Yeah if the context was the importance of "yards" vs. those other stats I agree with him. Here's a link with a direct quote:
http://www.miamidolphins.com/news/a...e-Joseph/4219764e-5367-42e9-ad18-24062a0a59b5

"If I had two stats after points allowed, it would be third downs and takeaways. Those are direct reflections of keeping the point total down. It all comes back to getting off the field. Can I have red zone defense, too, because that’s so important in keeping points off the board."

..

INT rate and I think you mentioned sack rate are all around -0.3 correlation so including those are fine too.

One thing I don't understand is whether you can't also target yards per passing attempt because statistically speaking playoff teams tend to be above average in passing Y/A both on offense and defense. That's something that matters a lot more than most of the non-scoring stats he's mentioning and really helps distinguish playoff vs. non-playoff teams.

 

So if you were to use those figures you talk about to determine how Miami looks, as far as if they're a playoff team or not, what would they show you?

 

I've tried a bunch of combinations of what helps predict playoff teams best, but nothing so far works as well as passer rating ranking vs. points allowed ranking so for now that's what I use:
https://i.imgsafe.org/701c8038e8.png

The data is from 2015 and as you can see that line almost perfectly separates playoff vs. non-playoff teams. In 2016 the Dolphins are #9 in passer rating ranking and #14 in points allowed ranking. Since the x-axis is for defense, plug in x = 14 into the equation for the dotted line, y = -2.15x + 39.6, that separates playoff/non-playoff teams.

You get y = 9.5 for required passer rating ranking, and we're just below that! So the Dolphins in 2016 lie almost exactly on the line that separates playoff and non-playoff teams from the best predictor I can find, at least based on 2015 stats. Kind of fits with where we perceive the team to be.

 

Wish I could answer your question.. I don't know what causes passing Y/A to go up (though increasing completion percentage helps because there are fewer 0's you average over) which is why I asked the question myself of how one could design a defense to truly increase passing efficiency without necessarily increasing/decreasing overall passing yards.

I will say this to the "skewed" question. Whatever skew there may be for playing from behind, the fact you get a good correlation between passing Y/A for offense/defense and making the playoffs means any such skew isn't dominating whatever it is that most increases Y/A.

For reference.. here are the stats that got me looking at that:
https://i.imgsafe.org/705d0921a1.png
https://i.imgsafe.org/705b632498.png

Green circles are playoff teams, red circle is the SB winner and blue diamond is the league average. You can see how each playoff or non-playoff team fares in passing Y/A for offense and defense through these graphs:
https://i.imgsafe.org/7074b7ab31.png
https://i.imgsafe.org/7078d558d1.png

The upper left quadrant is where the most efficient teams are on both offense and defense, and you can clearly see with the first graph for playoff teams that they tend to be in the upper left quadrant.

Point is.. one can collect stats to show passing Y/A on offense and defense is a stat you want to improve on, but how to actually improve it? That's beyond my knowledge and maybe someone who knows the football side more can help answer that.

 

y = -2.15x + 39.6 is the equation of a line. It basically says the slope of the line is -2.15 and the intercept is 39.6.

The interpretation: -2.15 means that for each defensive rank you lose (say you go from defensive rank #12 to #13 on the x-axis).. you have to increase passer rating rank by 2.15 to remain a borderline playoff team.

The intercept of 39.6 just tells you where the line crosses the y-axis. It doesn't really have a meaningful interpretation because the line crosses the y-axis when x=0, but x=0 is meaningless because you can't be ranked 0 on defense (0 rank is better than a rank of 1 haha). But IF you could have x=0 then you can be a playoff team with the impossible passer rating rank of 39.6 (impossible because there are only 32 teams in the league!).

 

Cool stuff.

 

I think he's done a fine job with the hand he's been dealt.

Going into this season, we all knew that our LB's and CB's would struggle. That was before we lost a pro bowl safety, another starting safety and 2 starting LB's.

I'm optimistic about the future seeing guys like Lippett and Howard develop. Alonso has played better than anyone expected and Maxwell has been Serviceable. Watching guys step in for injured players and see them elevate their game tells me someone is doing a good job of coaching.

Another offseason and hopefully drafting or binging in a good LB in the draft or FA, seeing our corners develop and actually being healthy should make this defense top half to third in the league, if not better.

If the OL can stay healthy next year, run the ball, protect Tanny and our defense plays better, there's no reason we can't compete for the division and playoffs.

 

Here's a question I've been kicking around. What do you think of Football Outsiders' DVOA stuff? They currently have Miami's Defense at 17th; 14th vs pass, 19th vs run (NE at 16th; 24th vs pass, 5th vs run) heading into this game, and it seems to me their info strikes a bit closer to reality than how sites like ESPN and others rank defenses.

 

The short answer is I'd treat it like any proprietary "stat": until FO is willing to be totally transparent and show their methodology, it can't be justified as a good method for doing what they're trying to do which is to adjust raw stats for down, distance, game situation and opponent, etc..

In other words, it's no better/worse than ESPN's QBR which I think you've seen rightfully panned on these boards. Any method that isn't transparent should be immediately discarded as a candidate for accurately adjusting for the factors they're trying to adjust for. The one justification for ESPN's QBR and FO's DVOA is that whatever they are doing, it turns out to predict things better (not by that much though) than simple, standard stats like passer rating or just using wins (e.g. DVOA does predict future wins a bit better than just using past wins, etc...)

So for the question you're asking (is it closer to reality) the answer is it's not even worth considering for that purpose.

To be clear though, the specific reasons (based on what is known publicly) for panning QBR and DVOA are different. QBR starts off well by asking how a given play in a given situation changes "team expected points added". That is.. how that play changed the average net points your team will gain/lose on the next scoring drive. The problem with QBR is that it then tries to divide up credit among players (e.g. a QB that just hands off the ball to a RB should get less credit than if he passes). In principle good idea, but how you do that in practice is the devil in the details, and as you can see in this conference where ESPN gives some idea of how they do it, they basically make arbitrary methodological choices for different situations:
http://www.sloansportsconference.com/content/total-qbr-what-espn-analytics-learned/

It's that arbitrariness that would be immediately ripped apart if they published their methodology.

Same thing with DVOA. Look at the description on FO's website:
http://www.footballoutsiders.com/info/methods

The one thing that should stand out to you is they don't describe their methodology!! They give you nuggets like this:

"A successful play is worth one point; an unsuccessful play, zero points with fractional points in between (e.g., eight yards on third-and-10 is worth 0.54 “success points”). Extra points are awarded for big plays, gradually increasing to three points for 10 yards (assuming those yards result in a first down), four points for 20 yards, and five points for 40 yards or more. Losing three or more yards is -1 point. Interceptions occurring on fourth down during the last two minutes of a game incur no penalty whatsoever, but all others average -6 points..."

Now.. it's not that FO just pulled those numbers out of thin air.. yet essentially that's what they're doing because those aren't numbers that are inferred from data with no additional assumptions made. Specific choices on what the anchor point should be (so relative to what situation are you computing change in expected points, etc..) or how you interpolate between different values, or even whether you use actual data to adjust any analysis for the probability of that play occurring in the first place (to my knowledge none of these methods do that even though they should) all affect those final numbers.

So, in summary.. both ESPN's QBR and FO's DVOA are better described as "models" not "stats" because they aren't directly inferred from data (includes all kinds of essentially arbitrary assumptions) and should not even be considered for accurately describing things until they publish their methodology. The one thing they ARE useful for is for prediction purposes. Just don't attach much meaning to the rankings.

 

Silly Brad...you can't "see" transparency!

Kidding, of course...thanks for the info.