Outliers, YPC, and the Cowboys running game

[Corso]

Sorry this is incredibly flawed.

If you want to glean something meaningful take the top 10 runs away from EVERY runner and then compare them.

Even the guys averaging pedestrian numbers had some long runs.

You boys understand the difference between average and median, right? There is a very important distinction between these two math terms and people often don't get it.

Yes it would be interesting to compare the MEDIAN runs of all backs in the league. I don't know if that stat exists.

It exists somewhere and it is useless for everyone except for the few trying to crane their neck to make some wild point.

 

[Beast_from_East]

Sorry this is incredibly flawed.

If you want to glean something meaningful take the top 10 runs away from EVERY runner and then compare them.

Even the guys averaging pedestrian numbers had some long runs.

You boys understand the difference between average and median, right? There is a very important distinction between these two math terms and people often don't get it.

Yes it would be interesting to compare the MEDIAN runs of all backs in the league. I don't know if that stat exists.

Good point bro.

You cant cherry pick runs to eliminate from some runners and not all runners. That would highly skew the data and you are correct, with skewed data you cannot use the mean and standard deviation to analyze a distribution. You have to use the median and IQR to get a better sense. You also need to analyze at least 30 RBs because you don't know if the distribution is normally distributed or left-tail/right tail skewed. You would then need to run your hypothesis test and compare the p-values you get before determining if your results are significant or not.

Or at least that is what they told me when I took stats at UT.

 

[Idgit]

The issue is that you cant just cherry pick which runs to throw out because when games are so close now days with many decided by a FG or less, those 7 plays might account for 2 extra wins a season. We won the East by 2 games last season, so you might be talking that 7 plays is what separates the division winners from division losers.

So yes, maybe 6 or 7 plays separates the top backs from the average backs but those 6 or 7 plays contribute directly to more wins, that is why the top backs get paid so much more than average backs. Otherwise, why would any team pay big money for a RB if they are all just about the same?

NFL teams are not stupid, there is a reason backs like Forte, Foster, Peterson, and Murray make huge dollars while other backs make peanuts.

On the other side of the equation is the ability to break the big play is what separates the average backs from the good ones. How many big plays did the average running backs make?

You say 7 and 9 plays, but is it that they broke 7 and 9 *more* big plays than the average back?

Not trying to discredit as I don't know their actual stats or the stats of the average running back. It would be interesting to look though. Something clearly set them apart. Even if it was just 7 or 9 plays respectively.

That's sort of what I was getting at, actually. Those 7-9 plays are important, because any big play is important. But it's a lot easier to get explosive plays like that out of the passing game where you only have to rely on a mistake or a trip from one player, which sort of masks the relative importance of having a statistically more explosive runner. Yes, it's valuable to have more of those big plays. But at ~$6.5M/year of differential, maybe you look at getting a more-likely-to-be-just average guy in the 2nd or 3rd round and hope you get lucky.

 

[Beast_from_East]

That's a good point, except for the fact that NFL teams with their "professionals" muck up all the time in player acquisition.
All the time. Almost every year. By almost every team. I'm excited to watch how this upcoming season unfolds.

But we are not talking about "player acquisition" here, we are talking about what separates the top backs in the league from the average backs in the league. Backs like Arian Foster, Matt Forte, Demarco Murray, Adrian Peterson, ect....have already proven their status as top backs in the league, be it with awards, probowls, all pro lists, ect....

So there is no debate on mucking up player acquisition here, the players have already proved they are among the best in the league. The issue is what makes them the best, is it the 6 or 7 exception runs that other backs cant make or is it a body of work over a season that makes them elite?

 

[Idgit]

Ok Idgit, go look at all of the 1-2 yard runs for that crucial first down and compare them to the other NFL running backs. I will await your answer in a week or so

It is those "dirty" yards that worry me, not the rest though.

That's where rushing effectiveness normally does affect the outcomes of games: short yardage and goal line. I'm actually not sure how we ranked last year in that regard. I imagine we were more effective than most at it. And, yes, that and pass protection is where I think we really struggle with the loss of Murray.

But I don't think that we're going to be seeing teams defending us a lot different otherwise for the vast bulk of the rushing plays we're going to be calling. We'll be doing everything else effectively enough relative to other teams that it's not going to matter.

 

[Idgit]

But we are not talking about "player acquisition" here, we are talking about what separates the top backs in the league from the average backs in the league. Backs like Arian Foster, Matt Forte, Demarco Murray, Adrian Peterson, ect....have already proven their status as top backs in the league, be it with awards, probowls, all pro lists, ect....

So there is no debate on mucking up player acquisition here, the players have already proved they are among the best in the league. The issue is what makes them the best, is it the 6 or 7 exception runs that other backs cant make or is it a body of work over a season that makes them elite?

This issue for me, anyway, was really more of just how fine that line was. If you're distilling it down to just that small number of plays, and then trying to suss out whether it was the RB or the blocking or the threat of the passing game or the downfield WR blocking or some combination of the four, for just a very small handful of plays, it starts to get a little more difficult to guess what it what. It's sort of a different discussion than the 'Randle can't handle 400 carries' variety of debate we've been having.

 

[Corso]

But we are not talking about "player acquisition" here, we are talking about what separates the top backs in the league from the average backs in the league. Backs like Arian Foster, Matt Forte, Demarco Murray, Adrian Peterson, ect....have already proven their status as top backs in the league, be it with awards, probowls, all pro lists, ect....

So there is no debate on mucking up player acquisition here, the players have already proved they are among the best in the league. The issue is what makes them the best, is it the 6 or 7 exception runs that other backs cant make or is it a body of work over a season that makes them elite?

Then I tip my hat to you more erudite expeditioners into this realm. I obviously got the ground rules wrong on this thread and offered a skewed remark. I back away semi-gracefully.

 

[superonyx]

The problem with stats and removing carries and dissecting YPC is this all takes out the timing of the carries and the effective it has on the defenses. I believe what makes a guy like Peterson so effective isn't his average runs. It's the threat to break off a TD on any run. This forces the defense to bite on his play action fakes. The other side of Pererson (or a elite back) is their ability to punish the defense throughout the game.

I believe the biggest impact of a running back or running game is the way they can keep a defense on their heels and break down their confidence. Having a back convert a 3rd down into a first down is demoralizing for a defense. Having a back in the game that can both get the 1st down and also take any carry all the way is truly game and season changing.

It remains to be seen if we have anything that can truly have a game changing impact.

 

[Idgit]

To follow-up on this you can check out the final paragraph here:
http://smartfootball.blogspot.com/2009/04/median-yards-per-attempt.html

Adrian Peterson--if you take away all his top runs...and his less-than-good runs...and strip it down so you have the most likely probability...the one run smack in the middle of all outcomes....

What do you get? 3 yards.

This looks really interesting, too. Going to have to digest it at my leisure.

And, for the record, T-Ro, I obviously defer to your statistical analysis in threads like these. As I've been trying to say, it was more the size of the effect of the small number of big rushing plays on YPC that surprised me than anything else, given the tendency to measure the performance of backs by yards/carry.

 

[Corso]

The problem with stats and removing carries and dissecting YPC is this all takes out the timing of the carries and the effective it has on the defenses. I believe what makes a guy like Peterson so effective isn't his average runs. It's the threat to break off a TD on any run. This forces the defense to bite on his play action fakes. The other side of Pererson (or a elite back) is their ability to punish the defense throughout the game.

I believe the biggest impact of a running back or running game is the way they can keep a defense on their heels and break down their confidence. Having a back convert a 3rd down into a first down is demoralizing for a defense. Having a back in the game that can both get the 1st down and also take any carry all the way is truly game and season changing.

It remains to be seen if we have anything that can truly have a game changing impact.

Very good point. Every play is relevant to it's situation in the context of the entire game and is hardly, if ever graded upon it.

 

[Idgit]

The problem with stats and removing carries and dissecting YPC is this all takes out the timing of the carries and the effective it has on the defenses. I believe what makes a guy like Peterson so effective isn't his average runs. It's the threat to break off a TD on any run. This forces the defense to bite on his play action fakes. The other side of Pererson (or a elite back) is their ability to punish the defense throughout the game.

I believe the biggest impact of a running back or running game is the way they can keep a defense on their heels and break down their confidence. Having a back convert a 3rd down into a first down is demoralizing for a defense. Having a back in the game that can both get the 1st down and also take any carry all the way is truly game and season changing.

It remains to be seen if we have anything that can truly have a game changing impact.

I usually discard the idea that teams try to tackle better RBs any harder than they try to tackle bad ones. They're defending formations and tendencies more than anything. For the really great players, they're probably keying on them more because it's a safer assumption he's going to get his no matter what, but play action for the Dallas Cowboys is going to be effective no matter who we've got in the lineup if we're committed to rushing the ball often.
I do agree that those tough conversions on the ground are demoralizing. Not sure that having the other guys demoralized more of the time necessarily correlates with winning more. When a better QB can then come in and erase the demoralizing points with big plays of his own, it doesn't matter all that much.

Short yardage and goal line, though, where you can extend drives or get points regardless of what the defense does, that matters. Which is why it also shows up in the victory correlation data.

 

[Beast_from_East]

That's sort of what I was getting at, actually. Those 7-9 plays are important, because any big play is important. But it's a lot easier to get explosive plays like that out of the passing game where you only have to rely on a mistake or a trip from one player, which sort of masks the relative importance of having a statistically more explosive runner. Yes, it's valuable to have more of those big plays. But at ~$6.5M/year of differential, maybe you look at getting a more-likely-to-be-just average guy in the 2nd or 3rd round and hope you get lucky.

It may be easier to get more explosive plays out of the passing game, but isn't that changing the offense?

I think that is what most people are concerned with when it comes to our running game and Murray. Romo had the fewest attempts of his career last season as we became a power run team. Murray gave us those 6 or 7 plays that made him an elite RB. Now if we switch RBs to somebody like McFadden, who does not give you those 6 or 7 special plays a season, then you have to get those 6 or 7 special plays from the passing game.

Hence, we become less of a power running team and more of a passing team in order to get those 6 or 7 special plays that we used to get out of the running game. So in effect, we are changing our offense from what made us successful last year and it might work and it might not work. I think that is what a lot of fans are concerned about.

 

[Beast_from_East]

Then I tip my hat to you more erudite expeditioners into this realm. I obviously got the ground rules wrong on this thread and offered a skewed remark. I back away semi-gracefully.

No problem bro, we are all just having a friendly discussion here...............all viewpoints are welcomed.

 

[Idgit]

It may be easier to get more explosive plays out of the passing game, but isn't that changing the offense?

I think that is what most people are concerned with when it comes to our running game and Murray. Romo had the fewest attempts of his career last season as we became a power run team. Murray gave us those 6 or 7 plays that made him an elite RB. Now if we switch RBs to somebody like McFadden, who does not give you those 6 or 7 special plays a season, then you have to get those 6 or 7 special plays from the passing game.

Hence, we become less of a power running team and more of a passing team in order to get those 6 or 7 special plays that we used to get out of the running game. So in effect, we are changing our offense from what made us successful last year and it might work and it might not work. I think that is what a lot of fans are concerned about.

I'm not talking specifically about the Cowboys in that regard. I mean in general, it gives a good indication of why passing effectively is more important than running effectively. Because the number of effective passing plays is going to swamp the number of effective rushing plays. It's a lot more important to have a statistically effective QB.

For the Cowboys in particular, I do think that's what people are worried about, and at least part of the point is that maybe they really shouldn't be. Even with a less than optimal stable of backs, we'll still probably be performing at or near the league average on the vast bulk of the rushing plays we're going to call. And we'll still have the big QB and the big play WRs and receivers we'll need if we have to rely on the passing game a little bit more in some games this season.

I don't think that does make us appreciably less of a power running team. If we're still rushing the ball more than half the time, we're still getting the benefit of our 5 OLs on their defense more often. We're still removing a lot of the passes or the must-win passing situations we were putting Tony in previously. We're talking about less than half of one big passing play per game here, and only that if we're not getting the big running plays instead form McFadden or Randle--both of whom actually have a reasonable amount of big play potential of their own.

 

[FuzzyLumpkins]

Removing outliers means you remove them from both ends of the spectrum and not just the inconvenient end. If you want to take away the long runs then you need to take away the TFL as well.

 

[Corso]

Removing outliers means you remove them from both ends of the spectrum and not just the inconvenient end. If you want to take away the long runs then you need to take away the TFL as well.

Like they do in Olympic scoring, but hopefully less corrupt thinking would be involved...

 

[Idgit]

Removing outliers means you remove them from both ends of the spectrum and not just the inconvenient end. If you want to take away the long runs then you need to take away the TFL as well.

We're really more counting the outliers here than we are removing them. Measuring the volatility of the statistic and how that might affect our play calling or productivity rather than trying to diminish anything Murray did. Those plays are gone, an they're big plays that win games no matter what. If we don't replace them, we're less effective. But if we're talking about the effectiveness of the 400 RB carries last season overall, the overwhelming bulk of those plays just weren't all that interesting statistically. What mattered a lot more was that we were committed to calling them.

 

[Nightman]

Removing outliers means you remove them from both ends of the spectrum and not just the inconvenient end. If you want to take away the long runs then you need to take away the TFL as well.

You also have to take away the 3rd and 1s, 4th and 1s and 1yd TDs.

If Murray had 7 40yd+ that is at least 280 yds on 7 carries. Of course it would bring his average down by a lot. That is 15% of his yards on 2% of his carries.

You would have to take away the top 7 runs of all the average backs as well.

 

[superonyx]

I usually discard the idea that teams try to tackle better RBs any harder than they try to tackle bad ones. They're defending formations and tendencies more than anything. For the really great players, they're probably keying on them more because it's a safer assumption he's going to get his no matter what, but play action for the Dallas Cowboys is going to be effective no matter who we've got in the lineup if we're committed to rushing the ball often.
I do agree that those tough conversions on the ground are demoralizing. Not sure that having the other guys demoralized more of the time necessarily correlates with winning more. When a better QB can then come in and erase the demoralizing points with big plays of his own, it doesn't matter all that much.

Short yardage and goal line, though, where you can extend drives or get points regardless of what the defense does, that matters. Which is why it also shows up in the victory correlation data.

I would think that looking at a running backs number of carries that are losses or very small gains are as important as any data. A 1 yard run on 2nd and 10 puts you in a terrible 3rd and 9 situation. A nice 5 yard run puts you in a very manageable 3rd and 4 where you have a much higher success rate on 3rd down.

Being consistent in positive yards also allows you to stick with the run.


I can't help but believe that the reason the Packers left Dez in single coverage on 4th and 2 in the fourth quarter is because they believed we would run the ball in that situation.

 

[FuzzyLumpkins]

Like they do in Olympic scoring, but hopefully less corrupt thinking would be involved...

It's what you do in all statistical analysis. If you want to bias out outliers then bias them all out and not just the ones that help further your argument. When you take away someone's best runs and they are still league average that tells me they are better than average. Context is also important. How do the league average guys fare when you take out their best runs? If all it does is shift the graph then it doesn't really posit anything.