The 2023-24 league year

kerouac9

Klowned by Keim
Joined
Feb 14, 2003
Posts
38,389
Reaction score
29,775
Location
Gilbert, AZ
xc_hide_links_from_guests_guests_error_hide_media
Wouldn't this graph have been more interesting if the vertical axis was average yards allowed on third down? What does EPA add to this equation?

EPA is like taking the integral of an integral; it flattens everything. There's a 1/2-point difference from the top of the graph to the bottom.
 

daves

Keepin' it real!
Supporting Member
Joined
Nov 2, 2003
Posts
3,526
Reaction score
7,207
Location
Orange County, CA
Wouldn't this graph have been more interesting if the vertical axis was average yards allowed on third down? What does EPA add to this equation?
Maybe not; if the average distance to go on 3rd down is 7.5 yards, and you allow 7 yards, that's good! if average distance to go on 3rd down is 6.5 yards, and you allow 7 yards, that's bad! EPA captures that distinction better than raw yards.
 

kerouac9

Klowned by Keim
Joined
Feb 14, 2003
Posts
38,389
Reaction score
29,775
Location
Gilbert, AZ
Maybe not; if the average distance to go on 3rd down is 7.5 yards, and you allow 7 yards, that's good! if average distance to go on 3rd down is 6.5 yards, and you allow 7 yards, that's bad! EPA captures that distinction better than raw yards.
Does it? How is EPA determined?
 

football karma

Michael snuggles the cap space
Joined
Jul 22, 2002
Posts
15,245
Reaction score
14,299
That would suffice.
Before the Steeler game I was this close to posting a thread devoted solely to Murray’s outfits, then decided nah, that’s stupid. One of my all time fails.
its never too late for the Murray Fits thread
 

kerouac9

Klowned by Keim
Joined
Feb 14, 2003
Posts
38,389
Reaction score
29,775
Location
Gilbert, AZ
Pretty good explanation here.

In the context of 3rd down defense, giving up yardage for a first down allows the offense to continue its drive, resulting in a bigger increase in expected points for the offense.
It’s so abstracted tho. I’d much rather see it measured against success rate or something that has more range. The human mind can’t really grok the difference between .05 EPA and .07 EPA but when graphed like this and without interrogating the scale lines it looks like differences are more/less dramatic than they are.

I’m also extremely wary of small sample sizes and I wish they’d append the N for each version.
 

Latest posts

Forum statistics

Threads
553,553
Posts
5,407,941
Members
6,317
Latest member
Denmark
Top