The Game Ball Thread: Week 5 vs Colts

Super Nomario

Member
SoSH Member
Nov 5, 2000
14,012
Mansfield MA
Did anyone else get the feeling that in the second half the Pats wanted to work on their passing game? They seemed to have no trouble getting 3-7 yards from Michel per play, yet they threw the ball anyway.
They came out pass-heavy and stayed pretty pass-heavy except for a drive towards the end of the second quarter when it looked like they were killing time before the half. I don't agree with your characterization of the run game - Michel gained two yards or fewer on 9 of his 19 runs (one wiped out by holding), but he had three runs of 10+. It was hit or miss. The pass game was the consistent chain-mover.
 

DennyDoyle'sBoil

Found no thrill on Blueberry Hill
SoSH Member
Sep 9, 2008
42,271
AZ
They came out pass-heavy and stayed pretty pass-heavy except for a drive towards the end of the second quarter when it looked like they were killing time before the half. I don't agree with your characterization of the run game - Michel gained two yards or fewer on 9 of his 19 runs (one wiped out by holding), but he had three runs of 10+. It was hit or miss. The pass game was the consistent chain-mover.
Isn’t that sort of how run games go? I mean, 18 for 98 with a TD is what it is. It’s 5 yards a carry and a TD, which is great.
 

Bergs

funky and cold
SoSH Member
Jul 22, 2005
21,612
This immediately made me think, "why don't we use median yards per carry instead of average" - apparently I'm not the first person to make this observation.
Because there would be next to no differentiation. All you really need is an indexed standard deviation to get a sense of how a given runner's mean is being driven by outliers. And pretty much any run over 15 yards is an outlier.
 

SeoulSoxFan

I Want to Hit the World with Rocket Punch
Moderator
SoSH Member
Jun 27, 2006
22,089
A Scud Away from Hell
well, not LeVeon Bell-ing your fresh new star-in-the-making RB is probably a factor too.
Speaking of, I thought Michel was great at being patient and reading the block. It seems clear that he does not possess Lewis-esque twitchy moves (gone with the knee injuries?).

However, a few of his runs were mini-Bell-ish in the way he did not overrun his blockers (also unlike Lewis, ala last year's Jags fumble run).
 

DennyDoyle'sBoil

Found no thrill on Blueberry Hill
SoSH Member
Sep 9, 2008
42,271
AZ
This immediately made me think, "why don't we use median yards per carry instead of average" - apparently I'm not the first person to make this observation.
Interesting. I’ve always just assumed it evens out over the season or even after 100 carries or so but maybe not.

I don't disagree with you, but it runs counter to the idea that Michel was gaining 3-7 yards consistently, which is what I was responding too.
Gotcha. Yeah, I agree.
 

williams_482

Member
SoSH Member
Jul 1, 2011
391
Isn’t that sort of how run games go? I mean, 18 for 98 with a TD is what it is. It’s 5 yards a carry and a TD, which is great.
It's worth noting how big the difference between a high variance and low variance 5.0 YPC can be.

To take a fairly extreme example, let's give two RBs four runs on 1st and 10 from the 50.
RB1 runs for -1, 0, 1, and 20 yards, worth -0.65, -0.58, -0.46, and +1.28 Expected points, a total of -0.41
RB2 runs for 3, 4, 6, and 7 yards, worth -0.15, +0.01, +0.29, and +0.44 Expected Points, a total of +0.59

Thats a swing value of an entire expected point, purely from how those 20 yards are distributed.
 

cgori

Member
SoSH Member
Oct 2, 2004
3,999
SF, CA
Because there would be next to no differentiation. All you really need is an indexed standard deviation to get a sense of how a given runner's mean is being driven by outliers. And pretty much any run over 15 yards is an outlier.
It's worth noting how big the difference between a high variance and low variance 5.0 YPC can be.

To take a fairly extreme example, let's give two RBs four runs on 1st and 10 from the 50.
RB1 runs for -1, 0, 1, and 20 yards, worth -0.65, -0.58, -0.46, and +1.28 Expected points, a total of -0.41
RB2 runs for 3, 4, 6, and 7 yards, worth -0.15, +0.01, +0.29, and +0.44 Expected Points, a total of +0.59

Thats a swing value of an entire expected point, purely from how those 20 yards are distributed.
This pretty much gets at what I was thinking. I see that a std deviation stat would also give this, but doesn't using the median shift things toward de-emphasizing the outliers naturally?

These things come into my mind quite often when they are talking about Bryce Love (Stanford running back) who will have runs of 1,3,1,55,3,3,4,38 and so forth. Even though he is a great runner, a few stuffed/crappy rush attempts really stall a drive.
 

BaseballJones

ivanvamp
SoSH Member
Oct 1, 2015
24,375
This pretty much gets at what I was thinking. I see that a std deviation stat would also give this, but doesn't using the median shift things toward de-emphasizing the outliers naturally?

These things come into my mind quite often when they are talking about Bryce Love (Stanford running back) who will have runs of 1,3,1,55,3,3,4,38 and so forth. Even though he is a great runner, a few stuffed/crappy rush attempts really stall a drive.
Barry Sanders was kind of like this. Lots of runs where he lost yardage or got nothing, but then a handful of absolutely explosive runs in a game. Very different than a guy like Emmitt Smith, who obviously had a lot of long runs over his career but was more of a consistent 4-6 yard guy.
 

williams_482

Member
SoSH Member
Jul 1, 2011
391
This pretty much gets at what I was thinking. I see that a std deviation stat would also give this, but doesn't using the median shift things toward de-emphasizing the outliers naturally?

These things come into my mind quite often when they are talking about Bryce Love (Stanford running back) who will have runs of 1,3,1,55,3,3,4,38 and so forth. Even though he is a great runner, a few stuffed/crappy rush attempts really stall a drive.
Success Rate is probably the best available simple statistic for evaluating RB performance. It implicitly accounts for at least some value form variance and situational performance, and correlates better with winning than YPC does.

The problem with medians is that almost every single RB in the NFL has a median cary of 3 yards. You could look at YPC for "middle third" runs or something along those lines, but that loses enough intuitiveness that you are probably better off looking at something a little more inclusive.
 

Bergs

funky and cold
SoSH Member
Jul 22, 2005
21,612
This pretty much gets at what I was thinking. I see that a std deviation stat would also give this, but doesn't using the median shift things toward de-emphasizing the outliers naturally
No. At least not in the context of this conversation.