Is QB Rating useless?

[Temo]

6 Daunte Culpepper 90.75
7 Chad Pennington 89.32
8 Tom Brady 88.36
9 Drew Brees 87.54
10 Trent Green 87.48
11 Jeff Garcia 86.43
12 Dan Marino* 86.38

http://www.profootballhof.com/history/story.jsp?story_id=2355

Culpepper had a good rating for such a lame athlete...

That "lame athlete" was the QB behind the highest scoring offense in the history of the NFL until Brady's Pats last year.

Also, let me repeat... you cannot compare QB Rating across era's. That's why I broke it down into 3 different eras in my post a little while ago. For the same reason that Brady Anderson isn't considered a better hitter than Joe DiMaggio.

 

[theogt]

I think it places too much emphasis on TD passes and completion percentage and not enough on ypa and ints which are much more important stats for a QB.

It places EXACTLY equal weight on YPA, INTs, and TDs.

This can be mathematically proven.

 

[AdamJT13]

It places EXACTLY equal weight on YPA, INTs, and TDs.

This can be mathematically proven.

I think what he meant is that TD passes and completion percentage should not be weighted equally with YPA and interceptions.

 

[theogt]

I think what he meant is that TD passes and completion percentage should not be weighted equally with YPA and interceptions.

I see. But still that's not a valid criticism. They're not "weighted equally" in the sense that one is more important than the other (or as equally important as the other). They're "weighted equally" in the sense that the final formula creates equal variance across each variable.

 

[Eskimo]

I see. But still that's not a valid criticism. They're not "weighted equally" in the sense that one is more important than the other (or as equally important as the other). They're "weighted equally" in the sense that the final formula creates equal variance across each variable.

It depends on what you mean by "valid".

If what I mean is say correlation of the combined QB statistic to points scored or wins, my guess is it could be shown that ypa and INTs are more strongly associated with these measured outcomes of performance than TDs and comp%.

Since that is what we are actually trying to encompass in the QB statistic, I could argue the current QB rating systems equal weighting is an invalid construct.

 

[theogt]

It depends on what you mean by "valid".

If what I mean is say correlation of the combined QB statistic to points scored or wins, my guess is it could be shown that ypa and INTs are more strongly associated with these measured outcomes of performance than TDs and comp%.

Since that is what we are actually trying to encompass in the QB statistic, I could argue the current QB rating systems equal weighting is an invalid construct.

True. But are we trying to get as close to wins or as close to ranking among peers? I'd clearly choose the latter as that's what it's named for -- i.e., a QB rating. If I were choosing a stat solely to correlate with winning, I'd certainly change quite a few things about it. My point about correlation with winning % earlier in the thread was only that it's not totally useless as the OP suggested.

 

[burmafrd]

Well to start with you need an official scorer like baseball has.

Manning had at least 2 picks dropped. Now on one of them TOOMER stopped when Manning expected him to continue. If that had been a pick then manning gets blamed but it was Toomer's mistake. So that pick should not be charged to Manning. It should be charged to Toomer.

People try and say that that evens out- but how do you know?

KC Joyner uses a bad decision measure which is a good idea but like all too many of his stats is too limited.

YPA is a very valid stat- but I would also have YPC as well. THen break that up to 2 further stats: distance pass traveled to receiver and yards after catch. The Yards after catch should go to the WR not the QB.
IT should be weighted so that the farther the ball travels in the air the more credit goes to the QB.

Any tipped ball automatically does not get charged to the QB. Any ball that hits a WR and is subsequently intercepted should not be charged on the QB- unless the QB bounced the ball off of his helmet or hit him in the butt, etc.
Any ball that hits the receivers hands from that point on is his no matter whether he catches it, drops it, lets it bounce off his hands and gets picked, etc. The QB has done his part and gotten the ball to where it needed to be.

Thats just a few things that would help make the rating a better stat.

 

[sonnyboy]

I think what he meant is that TD passes and completion percentage should not be weighted equally with YPA and interceptions.

"YPA is a very valid stat- but I would also have YPC as well. THen break that up to 2 further stats: distance pass traveled to receiver and yards after catch. The Yards after catch should go to the WR not the QB.
IT should be weighted so that the farther the ball travels in the air the more credit goes to the QB."


YPA is the stat. It is by far the most telling. What's odd about the passer rating system is that it even considers completion percentage.

Two QB's achieve an 8.00 YPA. QB A does it while completing 65% of his passes and QB B completes 60%.

QB A ends up with the better rating and is rewarded for dinking and dunking.

TD passes are significant, but I believe total points scored by an offense including FG's should be factored into the passer rating. You need to come up with some kind of equation were you double count TD pass points and add them to all other points.

I would like to see how a passer rating system works using these factors:

A POINT%
B YPA
D INT%

 

[Eskimo]

True. But are we trying to get as close to wins or as close to ranking among peers? I'd clearly choose the latter as that's what it's named for -- i.e., a QB rating. If I were choosing a stat solely to correlate with winning, I'd certainly change quite a few things about it. My point about correlation with winning % earlier in the thread was only that it's not totally useless as the OP suggested.

I would agree it is not useless.

It is like most statistics - useful when interpreted appropriately. There is no one statistic that is going to be the "be all and end all" when it comes to football performance. There are far too many variables that are not easily measured.

If you want a game in which statistics can tell much/most of the story (offensively anyway) then you are a baseball fan.

 

[Yakuza Rich]

I wouldn't say it's completely useless. It's an okay barometer of QB play. The problems with QB rating is that it has two components that can be more easily manipulated in completion percentage and passing TD's. West Coast offenses throw shorter passes and tend to throw a lot more inside the red zone than more traditional offenses, thus WCO's tend to have higher QB ratings. I'd say anytime you get a WCO QB, you should probably take off about 5-10 points on their QB rating to see how well they are doing in comparison to QB's from different systems.

That being said, QBRD (QB Rating Differential)...a formula I "concocted" has an extremely high statistical correlation to winning in the NFL. All QBRD does is take a team's QB rating and subtracts it's QB rating allowed on defense from it. The higher the QBRD, the more games a team is going to win (in general). So it's tough to say that QB rating is useless when QBRD has historically had such a strong correlation to winning in the NFL.

One thing about QB rating as well, if you want to get a good idea of how well a QB played, I suggest to also look at ypa. Campbell's YPA was at 4.9 ypa IIRC, which is horrendous.




YAKUZA

 

[Yakuza Rich]

"YPA is a very valid stat- but I would also have YPC as well. THen break that up to 2 further stats:

YPC isn't really that important compared to YPA. The big thing that YPC does cover is dropped passes. But you still have to have quite a few dropped passes to see a major spike in ranking in YPC.





YAKUZA

 

[Eskimo]

I wouldn't say it's completely useless. It's an okay barometer of QB play. The problems with QB rating is that it has two components that can be more easily manipulated in completion percentage and passing TD's. West Coast offenses throw shorter passes and tend to throw a lot more inside the red zone than more traditional offenses, thus WCO's tend to have higher QB ratings. I'd say anytime you get a WCO QB, you should probably take off about 5-10 points on their QB rating to see how well they are doing in comparison to QB's from different systems.

That being said, QBRD (QB Rating Differential)...a formula I "concocted" has an extremely high statistical correlation to winning in the NFL. All QBRD does is take a team's QB rating and subtracts it's QB rating allowed on defense from it. The higher the QBRD, the more games a team is going to win (in general). So it's tough to say that QB rating is useless when QBRD has historically had such a strong correlation to winning in the NFL.

One thing about QB rating as well, if you want to get a good idea of how well a QB played, I suggest to also look at ypa. Campbell's YPA was at 4.9 ypa IIRC, which is horrendous.




YAKUZA

I agree with ypa as a very good QB statistic. The only thing I would like to see is how good a measure something like ypa minus some coefficient multiplied by INTs is as a predictor of things like winning and points scored.

 

[joseephuss]

YPC isn't really that important compared to YPA. The big thing that YPC does cover is dropped passes. But you still have to have quite a few dropped passes to see a major spike in ranking in YPC.





YAKUZA

Perhaps. YPC has gone done considerably during the evolution of football while YPA has stayed about the same. YPC has dropped from about 14.0 YPC to just under 11.5 YPC.

I like the idea of adding YPC because it is easty to do. You don't need an official scorer to determine how far the pass traveled or how much RAC the receiver provided. Or if the INT was the QBs fault or the receivers fault, which would be difficult to do as an outsider since you don't know the play called or the responsibilities of the QB and receiver against different defenses. YPC is just a straight forward stat that can be easily calculated for any QB past and present.

The QB rating is not useless. It just comes down to how it is used. Like any stat it can't be used in a vacuum. There is a lot more you have to consider when performing your evaluations. The QB rating or any stat you choose is just an aid. It is not meant to be the be all end all. Any scout, coach or GM using it as the only basis for their evaluation will quickly find themselves without a job.

 

[Sarge]

Qb rating is a good idea but it needs tweaking from it's current format.

 

[Temo]

If anyone is interested, Pete Palmer wrote a very interesting article on QB rating over on FO. He's not an FO employee, but he is somewhat of an icon in the area of sports statistics (in baseball and football).

Link:
http://www.footballoutsiders.com/2008/09/02/ramblings/stat-analysis/6505/