PFF seems low on the vols offense

[sjt18]

For those not aware or choose to read past it, this is how the grading scale works.

Each game players are graded on a scale from 1-100. 90-100 represents an elite grade, 85-89 represents an all-conference player, 70-84 is starter quality, 60-69 is backup quality and everything under 60 is replaceable.

PFF Grades: Tennessee vs. Virginia | Rocky Top Insider

Check out the individual PFF grades for Tennessee football's season opening win over Virginia Saturday afternoon.

www.rockytopinsider.com

So apparently UT only has 3 starter quality players. Yeah... that's objective.

 

[Carp]

So Dylan Sampson has 4 TD’s. 1 receiving and 3 rushing. But is rated at a 56.7? This score puts him in the replaceable column. Those 4TD’s don’t play into his rating?

Sampson averaged about 4.2 yards per touch, which is good but not great. I realize some of that is because he was in short yard situations, but overall I didn't think he had an amazing day or anything. PFF also grades the plays he didn't touch the ball.

The ratings are fun to look at but you can't read too much into it. PFF has no way of knowing specific assignments on any given play. Some plays are obvious (like if the defense is playing man). Others, not so much.

 

[VAVol85]

Interesting to see how Pili and Herring compare now that we know Pili will be out for a while. Similar number of plays but Herring actually graded higher.

 

[DiderotsGhost]

I'm not convinced that PFF is anything other garbage analytics. I work in data science and a lot of their 'grades' just feel like they are coming from some crap model that has no track record of evaluating things properly.

Their marketing is much better than their analytical work.

 

[Franklin_Vol09]

I'm not convinced that PFF is anything other garbage analytics. I work in data science and a lot of their 'grades' just feel like they are coming from some crap model that has no track record of evaluating things properly.

Their marketing is much better than their analytical work.

It’s easier to run the grades in the NFL. They have common opponents. PFF for me, is used more for offensive lineman, than the other positions.

 

[DiderotsGhost]

It’s easier to run the grades in the NFL. They have common opponents. PFF for me, is used more for offensive lineman, than the other positions.

It's really more of a 'quality of data' question than 'how to evaluate opponents' (though I agree there can be issues with that as well). They don't provide any explanation for their model. What data is being used? For a lot of positions like OL, it's very difficult to evaluate with automated processes. Whether they are using numeric data or using video footage, it seems very likely that the data is questionable, leading to poor results. Garbage in, garbage out.

They've marketed themselves as 'cutting edge', but a lot of their grades just seem like random nonsense. My guess is that underneath the hood is a very crappy model using flawed data.

 

[Franklin_Vol09]

It's really more of a 'quality of data' question than 'how to evaluate opponents'. They don't provide any explanation for their model. What data is being used? For a lot of positions like OL, it's very difficult to evaluate with automated processes. Whether they are using numeric data or using video footage, it seems very likely that the data is questionable, leading to poor results. Garbage in, garbage out.

They've marketed themselves as 'cutting edge', but a lot of their grades just seem like random nonsense. My guess is that underneath the hood is a very crappy model using flawed data.

Yep. It’s the garbage in garbage out scenario.

 

[pimo1]

PFF has value only if one understands how it works. They grade based on all plays. So lets say you have 2 linebackers 1 plays 20 snaps the other plays 40 snaps but their statline is the same. The guy with 20 snaps will have a significantly higher score. Thats a simplistic view but the gist. Also it grades every snap. So if a WR has a snap where he was covered, failed to get a release, dropped a ball etc he fets a low grade that snap. If an olineman gets blown up, gives up a sack/tfl same.

Because of all this PFF scores game to game dont really tell much but their performance in the context of that game. PFF has value over the course of time after most of a season etc. Also note a score over 60 is good as scores over 80 are hard to pull off especially if one takes a lot of snaps.

If you look at the scores in context of the game against UVA they make a lot more sense. Of course our defensive players did well because the defense was great and the offense was kinda okish especially first half. Before people start bashing stats understand then.

 

[sechamp2012]

I'm not convinced that PFF is anything other garbage analytics. I work in data science and a lot of their 'grades' just feel like they are coming from some crap model that has no track record of evaluating things properly.

Their marketing is much better than their analytical work.

And somebody is making a lot of money selling this bottom line. Very subjective you can’t have the same graders across the board even though they have standards it’s humans and a lot of data for anyone to check & double check and back grade the analyzers. it’s just one of many subjective measuring sticks.