EverydayImRusselin
Well-known member
- Joined
- Apr 15, 2013
- Messages
- 2,015
- Reaction score
- 661
Popeyejones":1nb4t5td said:EverydayImRusselin":1nb4t5td said:Not that I disagree much with the premise of the OP, I just wanted to point out that PFF sucks for grading. I won't rehash it but suffice to say there are hundreds of articles available via google.
The critiques about PFF basically amount to two things:
1) In order to create an advanced statistic you have to use formulas to assign values to things. This is a catch-all critique of any type of algorithmic deduction. If someone took that the time to show that PFF's formulas are objectively poorly weighting I'd be all for that, but just complaining about it is toothless as it's a complaint about the existence algorithmic deduction, not PFF.
2) There is subjective classification (i.e. "noise") in PFF's "objective" system.. Well, yeah, duh, this is also true practically the entire rest of the statistical world, be it in survey design, coding schema, response error, etc. People seem to believe that this is a death sentence for PFF rather than being a general observation about enumerating real world events. Also worth noting that the vast, vast majority of critiques of PFF comes from fans who don't like how their own players are scored (read: if "objectivity" is a problem with PFF we've really amassed the least objective accusers in the world to lobby the complaint).
All in all, PFF obviously isn't perfect and obviously could be better (that's true of all things), but I find the critiques of it to be really toothless.
For the critique not to be toothless it would have to successfully address the following two questions, IMO: 1) Is there something better we should be using, and if so, what is it?; 2) Do its flaws make it entirely useless or even damaging? I think that's an incredibly hard argument to make with a straight face (e.g. anyone want to stand up and argue PFF is wrong and the Seahawks Oline is actually really good?).
ASIDE #1: When critiquing PFF nobody really seems to mention that every team in the NFL uses a similar per-play +/-X grading scheme as part of their postmortem for each game. This is what coaches are talking about when they talk about guys "grading out well" during in-week press conferences. It's not like PFF is doing some zany thing off in the wild; they're doing essentially what every team in the NFL does also.
ASIDE #2: No surprise that the critiques of PFF are basically the same as the critiques of ESPN's QBR algorithm, as they're really just critiques of algorithms rather than specific critiques of those algorithms. QBR obviously has its problems too, but nobody ever mentions what it's good at: it's better at predicting winning than any quarterback stat that came before it. Unless people do the leg work to model winning as based on QB play in a better way, they're just yelling at clouds, IMO.
Without going into it too much, the fact that they have grades on both teams within an hour or so of the game being finished tells me they aren't being that thorough. That's not to mention the fact they have no clue what each players assignment is on almost every single play.
Feel free to give them as much weight as you want, I give them very little.