New QB rating system set to debut? (1 Viewer)

[CalBear]

Here are the correlation coefficients of the three numbers for this week's action:

Passer vs. QBV: .76

QBV vs. QBR: .725

Passer vs. QBR: .75

In the end, they're all measuring the same stuff. My assertion is that any attempt to solve a problem where one of the systems is producing unintuitive results, will result in different unintuitive results. To wit: Of the six comments Pasquino makes about QBR producing unintuitive results in week 2, his QBV system produces results just as poor on two of them (Brees and Newton); passer rating does a better job at matching our intuition those cases than either QBV or QBR.

Although there seems to be some problems with your passer rating data. Passer rating cannot be below zero, and you're showing two negative numbers. There look to be other issues as well.

 

[Jeff Pasquino]

FYI I'm tracking NFL Passer Rating, ESPN Total QBR and my QBV stats in Hot Reads.

In case you have no idea what QBV is, read this.

Looks like your columns are mislabeled. Dalton had a Passer rating of 107.0 - not QBV.-QG

Correct, columns 3 and 4 are QBV; 5 and 6 are passer rating. Working on it getting fixed. Thanks

 

[Jeff Pasquino]

Here are the correlation coefficients of the three numbers for this week's actionasser vs. QBV: .76QBV vs. QBR: .725Passer vs. QBR: .75In the end, they're all measuring the same stuff. My assertion is that any attempt to solve a problem where one of the systems is producing unintuitive results, will result in different unintuitive results. To wit: Of the six comments Pasquino makes about QBR producing unintuitive results in week 2, his QBV system produces results just as poor on two of them (Brees and Newton); passer rating does a better job at matching our intuition those cases than either QBV or QBR. Although there seems to be some problems with your passer rating data. Passer rating cannot be below zero, and you're showing two negative numbers. There look to be other issues as well.

One of the goals of QBV was a K.I.S.S. approach that didn't take a team of ESPN editors to watch every snap and "grade" each play. QBV is just math based on the box score. Perfect? No. Pretty good? So far, I'd say yes.I understand looking at correlations, but I think the general "does it make sense" is the way to look at it. Does it matter if QB Jones is 15th or 18th? Probably not, but the Top 10 and the bottom 5-10 should make sense from a "I watch football and that makes sense" point of view.Matt Ryan and Aaron Rodgers being several spots behind Colt McCoy in Total QBR makes no sense. Cam Newton 29th for the week makes zero sense. Mark Sanchez was not a Top 10 QB last week, yet QBR says it was.

 

[Jeff Pasquino]

What a travishamockery this week with Total QBR. Unreal.

I cannot believe that Chad Henne is QB #6 for the week. Brutal.

 

[Chase Stuart]

http://espn.go.com/nfl/story/_/id/7050006/nfl-week-4-total-qbr-season-leaders

http://espn.go.com/nfl/story/_/id/7050004/nfl-week-4-total-qbr-leaders

QBR hates Sanchez, and I can't say I disagree.

 

[CalBear]

Here are the correlation coefficients of the three numbers for this week's actionasser vs. QBV: .76QBV vs. QBR: .725Passer vs. QBR: .75In the end, they're all measuring the same stuff. My assertion is that any attempt to solve a problem where one of the systems is producing unintuitive results, will result in different unintuitive results. To wit: Of the six comments Pasquino makes about QBR producing unintuitive results in week 2, his QBV system produces results just as poor on two of them (Brees and Newton); passer rating does a better job at matching our intuition those cases than either QBV or QBR. Although there seems to be some problems with your passer rating data. Passer rating cannot be below zero, and you're showing two negative numbers. There look to be other issues as well.

One of the goals of QBV was a K.I.S.S. approach that didn't take a team of ESPN editors to watch every snap and "grade" each play. QBV is just math based on the box score. Perfect? No. Pretty good? So far, I'd say yes.I understand looking at correlations, but I think the general "does it make sense" is the way to look at it. Does it matter if QB Jones is 15th or 18th? Probably not, but the Top 10 and the bottom 5-10 should make sense from a "I watch football and that makes sense" point of view.Matt Ryan and Aaron Rodgers being several spots behind Colt McCoy in Total QBR makes no sense. Cam Newton 29th for the week makes zero sense. Mark Sanchez was not a Top 10 QB last week, yet QBR says it was.

We have a simple passing rating system that does a pretty good job of predicting success. If you (or anyone) proposes a system to replace it, the replacement system should be meaningfully differentiated from the existing system, and the differences should align with our intuitive understanding of good quarterback play. With correlation around .75, I think both QBV and Total QBR fail on the first point; they are not sufficiently differentiated from passer rating to provide meaningful distinctions. For the players who are rated significantly differently by QBV or Total QBR, you can point to some who agree with our intuition and some who don't. I just don't see what we're getting from that.

 

[Joe Summer]

Through 5 weeks, the player with the highest QBR ranking is........ Alex Smith

 

[SlaX]

Through 5 weeks, the player with the highest QBR ranking is........ Alex Smith

Or, it could be Aaron Rodgers and Alex Smith just led week 5's numbers.

 

[Jeff Pasquino]

Here are the correlation coefficients of the three numbers for this week's actionasser vs. QBV: .76QBV vs. QBR: .725Passer vs. QBR: .75In the end, they're all measuring the same stuff. My assertion is that any attempt to solve a problem where one of the systems is producing unintuitive results, will result in different unintuitive results. To wit: Of the six comments Pasquino makes about QBR producing unintuitive results in week 2, his QBV system produces results just as poor on two of them (Brees and Newton); passer rating does a better job at matching our intuition those cases than either QBV or QBR. Although there seems to be some problems with your passer rating data. Passer rating cannot be below zero, and you're showing two negative numbers. There look to be other issues as well.

One of the goals of QBV was a K.I.S.S. approach that didn't take a team of ESPN editors to watch every snap and "grade" each play. QBV is just math based on the box score. Perfect? No. Pretty good? So far, I'd say yes.I understand looking at correlations, but I think the general "does it make sense" is the way to look at it. Does it matter if QB Jones is 15th or 18th? Probably not, but the Top 10 and the bottom 5-10 should make sense from a "I watch football and that makes sense" point of view.Matt Ryan and Aaron Rodgers being several spots behind Colt McCoy in Total QBR makes no sense. Cam Newton 29th for the week makes zero sense. Mark Sanchez was not a Top 10 QB last week, yet QBR says it was.

We have a simple passing rating system that does a pretty good job of predicting success. If you (or anyone) proposes a system to replace it, the replacement system should be meaningfully differentiated from the existing system, and the differences should align with our intuitive understanding of good quarterback play. With correlation around .75, I think both QBV and Total QBR fail on the first point; they are not sufficiently differentiated from passer rating to provide meaningful distinctions. For the players who are rated significantly differently by QBV or Total QBR, you can point to some who agree with our intuition and some who don't. I just don't see what we're getting from that.

and what simple system does this? If you mean NFL Passer Rating, that fails to fairly cover QBs who have a more complete game with running skills (Vick, Tebow, Rodgers). ESPN Total QBR started this debate - again - that NFL Passer Rating is flawed, and I contend that ESPN's method is not good. Including subjectivity on a rating is akin to having judges scoring a sport, which we all know is a problem. Not only that, but they have to watch the game - every play - to grade out a QB. Well, so much for wanting to go back in time and grade out a Joe Montana or Dan Marino.To judge a scoring system by its correlation isn't fair either in total, as there are only 24-32 data points each week. If you look at the Top 10 and the Bottom 5 and they pass the sniff test, then you might be on to something. That's what I believe QBV does better than QBR or Passer Rating.

 

[Reepicheep]

Through 5 weeks, the player with the highest QBR ranking is........ Alex Smith

Or, it could be Aaron Rodgers and Alex Smith just led week 5's numbers.

Interestingly, Alex Smith is very quietly #3 in the traditional passer rating at 104.1

 

[CalBear]

and what simple system does this? If you mean NFL Passer Rating, that fails to fairly cover QBs who have a more complete game with running skills (Vick, Tebow, Rodgers).

Vick finished with a 100.2 passer rating last year, (#4 in the league), and Aaron Rodgers finished at 101.2 (#3) which means they both seem to be being accounted for reasonably well. Tebow has a 101.7 rating this year. I agree that the ESPN system is problematic and not an improvement. I just don't really see the argument for improvement being necessary, and I don't see any system which correlates strongly with the existing rating system as being worth spending time on.

 

[Jeff Pasquino]

and what simple system does this? If you mean NFL Passer Rating, that fails to fairly cover QBs who have a more complete game with running skills (Vick, Tebow, Rodgers).

Vick finished with a 100.2 passer rating last year, (#4 in the league), and Aaron Rodgers finished at 101.2 (#3) which means they both seem to be being accounted for reasonably well. Tebow has a 101.7 rating this year. I agree that the ESPN system is problematic and not an improvement. I just don't really see the argument for improvement being necessary, and I don't see any system which correlates strongly with the existing rating system as being worth spending time on.

OK - so first you say that "We have a simple passing rating system that does a pretty good job of predicting success."That's false on all accounts. None of these (Passer rating, QBR or QBV) PREDICT success - they just assess passt performance. I don't believe any of it is predictory.Then you agree with me that ESPN is problematic and not an improvement. We have common ground here. My entire point of QBV was that someone had to point out the fallacy of the ESPN Total QBR being a drastic improvement. I took the time to write up all the issues (at least most of them) and explain how you don't need a team of video analysts to figure out a better way. In a matter of days, by myself, I showed a better system than what ESPN did with a full staff working for months.None of them are perfect, but I contend that QBV > QBR > Passer Rating, if only because QBV and QBR don't just limit performance analysis of a QB to passing.

 

[Chase Stuart]

Donovan McNabb: Fifth in passer rating in week 6, fifth from bottom in ESPN QBR.

 

[Native]

Stafford is tied with Matt Cassel for 15th in QBR after six weeks, despite leading his team to a 5-1 record with 15 TD's, 4 INTs, 1729 yards (on pace for 4600+) and 62% completion percentage. Notable players ahead of him - Chad Henne, Jason Campbell and Philip Rivers (who has 6 TDs and 7 Int's).

I read through the small print of QBR and liked what I saw but some of these rankings still feel way off.

 

[Chase Stuart]

Stafford is tied with Matt Cassel for 15th in QBR after six weeks, despite leading his team to a 5-1 record with 15 TD's, 4 INTs, 1729 yards (on pace for 4600+) and 62% completion percentage. Notable players ahead of him - Chad Henne, Jason Campbell and Philip Rivers (who has 6 TDs and 7 Int's). I read through the small print of QBR and liked what I saw but some of these rankings still feel way off.

Stafford has the sexy numbers -- 5-1, 15-4 -- but he's not playing like a top five QB in the league.For example, the completion percentage you cite ranks 15th in the league.In net yards per attempt he ranks 16th (in yards per attempt he's 18th).Now Stafford does rank in the 6-8 range in ANY/A, which includes the TD/INT numbers. And from the perspective of ESPN QBR, that should correlate more closely than NY/A, anyway. I suspect that Calvin Johnson is getting a lot of the credit on his touchdowns/catches (deservedly so) which takes away from some of the credit Stafford might get. That and the garbage time numbers Stafford compiled against KC don't count for much.I think Stafford should probably be in that 8-12 range.

 

[Ghost Rider]

How does Philip Rivers, who threw interceptions on the Chargers last two drives with them down 31-20, finish with 91.9 and the 3rd highest QBR for the week?

 

[FavreCo]

How does Philip Rivers, who threw interceptions on the Chargers last two drives with them down 31-20, finish with 91.9 and the 3rd highest QBR for the week?

I guess it needs tweaked....A LITTLE.

 

[Synthesizer]

How does Philip Rivers, who threw interceptions on the Chargers last two drives with them down 31-20, finish with 91.9 and the 3rd highest QBR for the week?

Does the new rating system weigh some parts of the game more heavily than others? To be fair, Rivers was 21-27 for 280 yards and 2 TDs/0 INTs the rest of the game. Excellent numbers.