Toxic Differential - Bills #1

[hondo in seattle]

I really wasn't trying to paint Billick as unintelligent, I was just trying to say his grasp on statistics probably isn't as strong as he thinks (I could be wrong). I've worked with many accomplished people that I consider to be intelligent (physicians, engineers, even CFOs paid to analyze numbers) and I am always a bit taken aback when I discover they don't fully grasp basic probability. Is there a correlation between successful people and aptitude with probability? In my experience, not a strong one (I'd estimate an r-squared of .3 )

 

 

 

I don't, however I can give you an example of a guy talking out of both sides of his mouth. Yesterday Billick told us the Bills were atop his toxic differential list (predicting success) and today he predicts us to miss the playoffs. Maybe I'm missing something?

 

That is interesting. He says the Toxic Differential is predictive and then predicts that the team with the best Toxic Differential in NFL won't even make the playoffs.

[thewildrabbit]

Not really sold on this metric, personally, but at the very least it's interesting to look at. As we saw a couple years ago, the type of extreme turnover differential we've been experiencing isn't really sustainable, generally speaking.

I agree, as lady luck often has a say in what happens with that oblong spheroid. Spiller on IR, and Fredex injured has to have an adverse effect. Then opposing teams will surely try and double Sammy off the line, and roll coverage's his way. Not only that its not every game that the Bills will face a QB as lame as Geno Smith who keeps throwing pick after pick.

 

I believe Edwards Arm made a thread after the win against the Patriots in 2011, as Brady threw 4 INT's with the Bills win. EA talked about how a win like that with so many turnovers is unsustainable, he turned out to be correct. I see the same thing this year if the Bills get some things fixed.

 

It is great to see that turnover differential so good, as it means the Bills are getting lots of takeaways, and not turning it over. This next half of the season is no cakewalk as six out of the next eight teams have winning records. Then once the snow hits they had better be able to run the ball, or its game over.

[Kirby Jackson]

I agree, as lady luck often has a say in what happens with that oblong spheroid. Spiller on IR, and Fredex injured has to have an adverse effect. Then opposing teams will surely try and double Sammy off the line, and roll coverage's his way. Not only that its not every game that the Bills will face a QB as lame as Geno Smith who keeps throwing pick after pick.

 

I believe Edwards Arm made a thread after the win against the Patriots in 2011, as Brady threw 4 INT's with the Bills win. EA talked about how a win like that with so many turnovers is unsustainable, he turned out to be correct. I see the same thing this year if the Bills get some things fixed.

 

It is great to see that turnover differential so good, as it means the Bills are getting lots of takeaways, and not turning it over. This next half of the season is no cakewalk as six out of the next eight teams have winning records. Then once the snow hits they had better be able to run the ball, or its game over.

I think that there is a lot of truth to this. Brown has had a history of fumble issues (knock on wood). He will be seeing more carries and they play better QBs in the 2nd half. I still think that the Bills will score pretty well in the toxic differential but they won't continue to grow at the rate that they are.

[OCinBuffalo]

Interesting stat; thanks for posting. That said, I recommend reading this: http://sportsworld.n...cal-revolution/ . Straight from the horse's mouth (he's not criticizing this stat, of course).

Essentially, this is what I've been saying regarding every analytics based thread post here. And, the notion that Bill James created anything with regard to advanced statistics is patently false. He merely re-applied what has been going on in management consulting/WW2 since the 1940s. If you want to know where all of this began look up Robert McNamara/bombing of Japan. The Air Force was responsible for all sorts of analytic and workflow concepts. Hell, IDEF was still in use in the late 1990s, and it may still be(but it shouldn't). And, the fact that Viet Nam went the way it did: merely validates what Bill James/many of us who work in this field say on a daily basis. Over-reliance on analytics is dangerous(hence the complete FAIL of the "body count" metric in Viet Nam used by McNamara). Also, use of analytics by the untrained/those with average intelligence? Like handing automatic weapons to children.

Super Bowl coaches aren't exactly known for their mathematical acumen. Plus, there's a reason Billick is in the booth and Marvin Lewis is on the sidelines, but that's going off topic.

 

First of all, is this metric actually predicting anything, or is it just telling you what happened? Let's say I told you, "Teams that score more than 40 points and commit no turnovers have a 99% chance of winning -- you'd be hard-pressed to find any controllable statistical combination that has quite the same rate of predictive success." You would probably sarcastically laugh, "No kidding." I haven't really predicted anything, I've simply told you teams that score a lot of points and don't turn the ball over win games. We all know that. Billick is essentially saying the same thing - get turnovers and then score (explosive plays). In the end it's not telling us anything that we don't already know.

 

The key word here is "controllable." Can you really control and predict turnovers for a game? For a month? For half a season? Seattle led the league in turnovers last year, this year they are struggling to stay in the black. Our 2011 Bills stormed to a 5-2 start primarily based on turnovers. When the turnovers dried up, well...you know the rest.

 

I think you can win the turnover battle consistently if you have the elitest of the elite QBs - a Rogers, Manning, or Brady. Is that really telling us anything? Get a top 5 QB and you will have success? Not exactly a revelation.

 

For the record I predict we will have a negative toxic differential for the month of December. Any guesses why?

This is right....almost. You're forgetting one thing: the relational aspect of the data. The data of offense and defense combining, and that combination's historical propensity to predict outcome. What's being said by Billick is that having a D that gives up few big plays, gets few big plays, and also having an O that generates few big plays....is less likely to succeed that other combinations. It's the combination of these things, adding the differentials of the O and D.

 

That is predictive.

 

In other words: the Ravens of old were so good at making big plays/generating turnovers on D, that the fact that they had Trent Dilfer at QB and won the SB.....was an outlier. Or maybe it wasn't? If the differential on D was enough to overcome the differential on O on that team....then perhaps the Trent Dilfer/SB winning Ravens follow the model as designed?

 

We could say that having a D with a high differential propensity tends to put the O in better position to make big plays....but...consider the field position of our last game. We didn't make most of our "explosive" plays when we started at the 50. No. The big plays came on drives we started on ~our own 20. Besides, when you get the ball at the other team's 10 yard line, you are, by definition, automatically prevented from making an "explosive" play, because the most yards you can gain....are 10. EDIT: So, was the Bills D the cause of the explosive plays by the Bills O, thus making this model not very predictive at all(is this circular reasoning)? Not really. The "explosive plays" by the Bills O stand on their own 2 feet as independent data, just like the Bills D "explosive plays"/turnovers stand alone as well. Combining these two sets of independent data = "toxic".

 

Thus, there has to be some correlative effect here, that is predictive. The question is, as you point out, how much prediction is occurring as a function of hindsight/circular reasoning? EDIT: I'd argue that throwing around confidence #s like 80% and 95%....when perhaps as much as 40-50% of the 80% may very well be due to a factor like a top 5 QB...is, as you say: silly.

I've worked with many accomplished people that I consider to be intelligent (physicians, engineers, even CFOs paid to analyze numbers) and I am always a bit taken aback when I discover they don't fully grasp basic probability. Is there a correlation between successful people and aptitude with probability? In my experience, not a strong one (I'd estimate an r-squared of .3 )

Forget grasp of probability. The correlation that deserves your attention is between successful people and their aptitude with their own business processes.

I don't, however I can give you an example of a guy talking out of both sides of his mouth. Yesterday Billick told us the Bills were atop his toxic differential list (predicting success) and today he predicts us to miss the playoffs. Maybe I'm missing something?

Nope. Almost the entire sports media claims expert status, but the only thing they are experts at? Playing save ass, bet-hedging, and click generation.

 

Best example: a certain Buffalo News columnist "reserving the right to change my opinion"....yet never being accountable for the original awful, and emphatic, opinion he has reserved the right to change.

 

Once again: MOST media is now about click-rate(wawrow NOT included). They will say anything, however speculative/wrong, for clicks/likes/RTs, and then play save-ass...hopefully for long enough until the next Ray Rice story breaks. Then it's on to their other favorite activity: hypocritically calling out stars/some politicians, not others/athletes for not living up to their public responsibility. As if the media themselves are holding up their end on public responsibility. They are a joke.

 

I won't be surprised to wake up one day and find out that the Bills/Seahawks were considering trading Kyle Orton for Russel Wilson straight up.....if LaConfora, some Seattle reporter, or the rest of the click-whores thought it was worth the clicks.

Edited October 31, 2014 by OCinBuffalo