Weird ESPN QBR ratings

[Rod Johnson]

6 hours ago, NVRamsFan said:

Kizer and his 1 TD 1 INT 56.8% comp 232 yard day was higher QBR rating than Goffs 3 TD 0 INT 67.6% comp 355 yard day. Kizer 73.6 Goff 44.1 all because Kizer ran for 1 TD. QBR is complete garbage.

Kizer is just better

[RandyMossIsBoss]

I think some are way too harsh on QBR just because it's ESPN and its' first use by their anchors was to hype Tebow.

 

It's not all that hard to understand. It's basically a poor man's DVOA applied to the QB. Obviously it takes into account QB rushing, but more importantly it looks at context. For example, a 3 yard completion on 3rd & 2 is much more valuable than a 12 yard completion on 3rd & 20, but basic stats will tell you the latter play was essentially 4x more effective/valuable. QBR revolves around the expected points added (EPA) stat, so again, converting a 3rd & 10 with a 15 yard pass adds a lot more expected points than completing that same pass, from the same spot, on 1st & 10. Another example: a hail mary INT with 1 second left is less costly than an INT in the red zone. 

Any conversation of QBR should be pitting it against passer rating, which it was made to replace (much like PER was made to, and successfully did, replace efficiency rating in the NBA). For every ridiculous example of QBR not reflecting what actually happened, you will find just as many if not more for passer rating. Now if you think both are ridiculous and useless attempts to measure QB play with a single number, that's another story, but there's no reason to clown QBR and then treat passer rating as the holy grail. In all honesty QBR is probably a better indicator of QB performance than traditional passer rating (at the very least it is more highly correlated with wins, which is natural since that's what it ultimately measures: how much are you helping/hurting your team's chances of winning?)

 

The real flaw in QBR, and why I tend to stay away from it out of principle is that it is NOT purely statistical. It's like an odd mix of guys viewing the tape a la PFF, and applying metrics a la FO. I either want something based off of the box score (passer rating, DVOA, EPA) or based on observance (PFF grades), QBR tries to mix both, assigning blame to players. Is that necessarily a bad thing? Maybe not, but I prefer the two to be clearly separated. Also there is a clutch factor added on top of everything else (some of which already takes into account things considered as "clutch"), which weighs close games in the 4th quarter heavier than other situations. So basically they're giving double credit, or dinging you twice, for playing good/bad in these situations. 

 

 

At the end of the day QBR is just another tool in the toolshed, it shouldn't be the be all and end all, nor should it be treated like a number Trent Dilfer picks out of a hat.

[Soko]

2 hours ago, Forge said:

Also, I really hate that ESPN shoves that stat down your throat as if it really means anything at. It makes listening to any analyst or writer nearly unbearable. Like, I really like Bill Barnwell, have since the Grantland podcast, but because he has to reference that damn stat every 35 seconds when talking about any given quarterback, it's just brutal to listen to. 

ESPN has been largely unwatchable for a while now (although not to the extent of NFLN).

[Forge]

5 minutes ago, Yin-Yang said:

ESPN has been largely unwatchable for a while now (although not to the extent of NFLN).

Oh for sure. I can't remember the last time I actually watched something on that platform besides a football game. That being said, I have a lot of time where I'm listening to the radio / podcasts. There's only a few things that I can listen to though because it becomes insufferable. 

[MWil23]

ESPN has become a caricature of what's wrong with sports reporting/television today. Their allegedly massive post Thanksgiving layoffs will just continue to add more proof to this. 

[Forge]

3 minutes ago, MWil23 said:

ESPN has become a caricature of what's wrong with sports reporting/television today. Their allegedly massive post Thanksgiving layoffs will just continue to add more proof to this. 

Yeah, I've been kind of curious where they are going with that. Last time, they got rid of some quality reporters (including pretty much their entire NHL staff...it's like they don't even acknowledge it as a sport any more). Supposedly this one will be bigger with more on air talent...which quite frankly, doesn't really bug me since I don't watch. The thing that kills me is that they layoff what talent they have, but then pay Gruden 10 million a year, Mike Greenburg a crazy amount (I understand this one a bit, at least), not to mention Stephen A and numerous others. They hitch their wagon to the wrong people, and then they make flat out awful programming decisions. 

[frenchie]

QBR was invented for the sole purpose of making mobile quarterbacks appear to be better than what they are...  ESPN basically said as much when they began using it.

[Malik]

Far too much subjective data to even call it a statistic. It punishes QBs who play 4 quarters very well versus those who come back. A QB evaluating stat could definitely be added to football it would just have to rely on actual measurables instead of vaguely defined ideas of "pressure to the QB" and not asinine things like comeback victories.

[ViperVisor]

A big problem is that it makes no sense how every other week a guy will be within 5 points of being a Perfect QB.

Keenum last week was 98.9

21/29    304    4-2  

Now imagine a QB going cheese Madden and he is 20/20 500 yards and 10 TDs. That would be only 1 better at a 99.9

Or 10/10 for 100 yards passing but also ran the ball 20 times for 200 yards and 5 TDs.

If this was a MLB pitcher grade you would only give out the top top grades if a guy struck out 26 and 1 guy grounded out.

The guy ESPN hired from here is smart http://www.advancedfootballanalytics.com/

The end results he got were interesting and a good gauge of how good a team was per drop back more-so than the QB although it was sorted by QB.

To seriously try and grade QBs on a 100 scale you would have to look at every play and grade how much was the protection above or below the avg. How did the pass options run the routes. How good was the coverage before the throw. Was it the best target to throw to? How good was the defense after the ball left the QBs hand? How good was the WR with going for the catch.

A nearly impossible task.

 

I know what Passer Rating is. It is a simple and somewhat flawed equation based on a handful of things. From there we can argue why a QBs X is low which is tanking his rating or Y is high which is boosting it. It's hard to argue QBR when we are only vaguely shown what it is.

[LETSGOBROWNIES]

On November 16, 2017 at 6:55 AM, NVRamsFan said:

Kizer and his 1 TD 1 INT 56.8% comp 232 yard day was higher QBR rating than Goffs 3 TD 0 INT 67.6% comp 355 yard day. Kizer 73.6 Goff 44.1 all because Kizer ran for 1 TD. QBR is complete garbage.

Can't even let us have this one small victory huh?

[VegasOdds101]

Pats vs Cowboys 2015...... final score: Pats 30 Cowboys 6.

Brady: 20/27  275yds  2tds  0 ints. 130.9 rating,  2 rushes 3 yds  1 td

Weeden: 26/39  187yds  0 tds 1 int. 66.9 rating,  3 rushes 18 yds  0 tds

ESPN QBR:

Brady 24.1

Weeden 27.5

At some point later, after this discrepancy was reported everywhere and how dumb the stat was they changed the ratings to 34.4 for Brady and 21.5 for Weeden.  And remember 50 is supposed to be average. It's a ridiculous system that obviously puts feelings into account.

[Danger]

This stat is garbage.

 

He also had 2 rushing attempts for -2 yards. Was sacked twice for -15 yards.

[Cakeshoppe]

On 11/16/2017 at 11:19 AM, RandyMossIsBoss said:

I think some are way too harsh on QBR just because it's ESPN and its' first use by their anchors was to hype Tebow.

 

It's not all that hard to understand. It's basically a poor man's DVOA applied to the QB. Obviously it takes into account QB rushing, but more importantly it looks at context. For example, a 3 yard completion on 3rd & 2 is much more valuable than a 12 yard completion on 3rd & 20, but basic stats will tell you the latter play was essentially 4x more effective/valuable. QBR revolves around the expected points added (EPA) stat, so again, converting a 3rd & 10 with a 15 yard pass adds a lot more expected points than completing that same pass, from the same spot, on 1st & 10. Another example: a hail mary INT with 1 second left is less costly than an INT in the red zone. 

Any conversation of QBR should be pitting it against passer rating, which it was made to replace (much like PER was made to, and successfully did, replace efficiency rating in the NBA). For every ridiculous example of QBR not reflecting what actually happened, you will find just as many if not more for passer rating. Now if you think both are ridiculous and useless attempts to measure QB play with a single number, that's another story, but there's no reason to clown QBR and then treat passer rating as the holy grail. In all honesty QBR is probably a better indicator of QB performance than traditional passer rating (at the very least it is more highly correlated with wins, which is natural since that's what it ultimately measures: how much are you helping/hurting your team's chances of winning?)

 

The real flaw in QBR, and why I tend to stay away from it out of principle is that it is NOT purely statistical. It's like an odd mix of guys viewing the tape a la PFF, and applying metrics a la FO. I either want something based off of the box score (passer rating, DVOA, EPA) or based on observance (PFF grades), QBR tries to mix both, assigning blame to players. Is that necessarily a bad thing? Maybe not, but I prefer the two to be clearly separated. Also there is a clutch factor added on top of everything else (some of which already takes into account things considered as "clutch"), which weighs close games in the 4th quarter heavier than other situations. So basically they're giving double credit, or dinging you twice, for playing good/bad in these situations. 

 

 

At the end of the day QBR is just another tool in the toolshed, it shouldn't be the be all and end all, nor should it be treated like a number Trent Dilfer picks out of a hat.

This is a very good post. If you have a stat you agree with all the time, you run the risk of having a stat that isn't telling you anything. If we want advanced stats to have value, they should tell us something we don't know, and hence it should sometimes produce unintuitive results. Otherwise, it's just a way to make arguments look more convincing. So sometimes useful stats will be unintuitive and that's a good thing.

 

My big problem with QBR is that it's black-boxed. We don't get to see what their weightings are for any of the inputs and so then we just have to take their word that the outputs are telling us something useful when they don't accord with our intuitions. With something like Passer rating, which is very transparent, or DVOA, which is gray-boxed (we can't compute it ourselves but at least we know their methodology for computing it), at least we can look at the assumptions that go into it and decide if we agree with them. With QBR, when it spits out something really odd, we have no idea if it's telling us something valuable but unintuitive or if we're just seeing  a product of the problems we already know are built into it. 

[BroncosFan2010]

QBR is clearly created using a random number generator (Or perhaps manatees take numbered balls and place them into a series of tubes at ESPN HQ) and its sole use is to generate a discussion that includes the words 'ESPN' at some point.

[SWATcha]

On 11/16/2017 at 10:32 AM, EliteTexan80 said:

This. Absolutely zero value. You'd be better served to compare QBs based on what kind of car they drive over this wanna be analytic review.

Let's make this happen ET! Inquiring minds want to know this type of info!

Rogers drives a lifted Ford F-150 that's susceptible to fires, golf clubs, remote control aircrafts, and Clay Mathews in general.