Correlation of Passer Rating to Winning - game by game wk 6

[Mr. WEO]

I posted numbers. Make of it what you will.

 

More meaningless stats... 97 yards passing or 500 yards passing. If you win you look good. If you lose you look for excuses.

 

The numbers indicate nothing more than the Bills had balance with EJ.

 

the rational??? The team surrounding him.

 

You left out a lot of numbers.

[Rocky Landing]

First this: http://www.si.com/mo...ingdifferential . Passer rating differential (using the traditional stat) is the mother of all football stats. It's virtually impossible to argue with, I think.

 

And this: http://www.coldhardf...ats/2013/5/PRD/.

 

Also, here's an interesting takedown of ESPN's total QBR rating on Cold Hard Football Facts: http://www.coldhardf...r-tv-stat/7978/ .

 

 

It is. There's no other stat that comes close.

Honestly, I think the passer rating differential is a bit silly. The passer rating quantifies completions per attempt, yards per attempt, touchdowns, and turnovers. It's not that far from saying: "the team that scores the most points wins--100% of the time!" Of course the teams with the highest differential in these four categories will be the teams to make it to the Super Bowl. But that isn't predictive. It is a statistic that gains meaning as the season progresses. I would suspect (and I haven't looked at this) that a strong team in a weak division would have a higher differential than one coming out of a tough division. And, at this point, at the beginning of the post season, is the passer rating differential more useful than looking at those four categories (completions, yards, touchdowns, and turnovers) separately?

 

That being said, the passer rating differential does do one thing that the passer rating itself completely ignores: the performance of the opposing team (as well as the performance of the subject team, for that matter). But, that really means that the passer rating is more of a comparative stat between team performances, and less of an indicator of QB talent.

[Rocky Landing]

I posted numbers. Make of it what you will.

 

More meaningless stats... 97 yards passing or 500 yards passing. If you win you look good. If you lose you look for excuses.

 

The numbers indicate nothing more than the Bills had balance with EJ.

 

the rational??? The team surrounding him.

Balance? You can't use numbers that subjectively and call them "stats."

[Orton's Arm]

First this: http://www.si.com/mo...ingdifferential . Passer rating differential (using the traditional stat) is the mother of all football stats. It's virtually impossible to argue with, I think.

 

And this: http://www.coldhardf...ats/2013/5/PRD/.

 

Also, here's an interesting takedown of ESPN's total QBR rating on Cold Hard Football Facts: http://www.coldhardf...r-tv-stat/7978/ .

 

 

It is. There's no other stat that comes close.

 

I read each of the links. The first one was eloquently written, and makes about as good a case for passer rating as could be made. However . . . the New York Times performed a multiple linear regression analysis to determine which statistical variables were most correlated with winning. They came up with six: yards per pass attempt, yards per rush attempt, interception percentage, and the defensive equivalents thereof. Collectively, these six variables predicted 80% of the observed variation in teams' winning percentages. (The other 20% was presumably due to fumbles, penalties, sacks, special teams play, and other factors.) Yards per pass attempt was three times as important as either INT percentage or yards per rush attempt.

 

If passer rating looks good at predicting winning, it's because it relies on some of the same variables used in the New York Times' analysis: yards per attempt and interception percentage. But it also adds in an additional, unnecessary variable: completion percentage.

 

To show why this is a bad idea, consider the following:

Trent Edwards career quarterback rating: 75.5

Terry Bradshaw career quarterback rating: 70.9

 

Or--if you want to compare QBs from the same era--consider this:

John Elway career quarterback rating: 79.9

Kelly Holcomb career quarterback rating: 79.2

 

Suppose you were to use only the raw ingredients found in the New York Times multiple linear regression analysis, while discarding variables which didn't make it in. The single most important stat from their analysis was yards per pass attempt, so let's start there.

 

Trent Edwards career yards per pass attempt: 6.5

Terry Bradshaw career yards per pass attempt: 7.2

 

John Elway career yards per pass attempt: 7.1

Kelly Holcomb career yards per pass attempt: 6.6

 

If I was to create an overall rating for QB play--and have the measure be relatively simple--I'd use a combination of yards per attempt, air yards per attempt, and INT percentage. INT percentage would constitute 1/4 of the rating, with the other 3/4 evenly divided between yards per attempt and air yards per attempt. The rationale being that you don't know the extent to which YAC yards are the quarterback's doing, and the extent to which they're the result of the receiver's efforts. As a compromise, you give him full credit for air yards, and 50% credit for YAC yards.

Edited October 16, 2014 by Orton's Arm

[dave mcbride]

I read each of the links. The first one was eloquently written, and makes about as good a case for passer rating as could be made. However . . . the New York Times performed a multiple linear regression analysis to determine which statistical variables were most correlated with winning. They came up with six: yards per pass attempt, yards per rush attempt, interception percentage, and the defensive equivalents thereof. Collectively, these six variables predicted 80% of the observed variation in teams' winning percentages. (The other 20% was presumably due to fumbles, penalties, sacks, special teams play, and other factors.) Yards per pass attempt was three times as important as either INT percentage or yards per rush attempt.

 

If passer rating looks good at predicting winning, it's because it relies on some of the same variables used in the New York Times' analysis: yards per attempt and interception percentage. But it also adds in an additional, unnecessary variable: completion percentage.

 

To show why this is a bad idea, consider the following:

Trent Edwards career quarterback rating: 75.5

Terry Bradshaw career quarterback rating: 70.9

 

Or--if you want to compare QBs from the same era--consider this:

John Elway career quarterback rating: 79.9

Kelly Holcomb career quarterback rating: 79.2

 

Suppose you were to use only the raw ingredients found in the New York Times multiple linear regression analysis, while discarding variables which didn't make it in. The single most important stat from their analysis was yards per pass attempt, so let's start there.

 

Trent Edwards career yards per pass attempt: 6.5

Terry Bradshaw career yards per pass attempt: 7.2

 

John Elway career yards per pass attempt: 7.1

Kelly Holcomb career yards per pass attempt: 6.6

 

If I was to create an overall rating for QB play--and have the measure be relatively simple--I'd use a combination of yards per attempt, air yards per attempt, and INT percentage. INT percentage would constitute 1/4 of the rating, with the other 3/4 evenly divided between yards per attempt and air yards per attempt. The rationale being that you don't know the extent to which YAC yards are the quarterback's doing, and the extent to which they're the result of the receiver's efforts. As a compromise, you give him full credit for air yards, and 50% credit for YAC yards.

 

Bradshaw's career numbers are irrelevant - what were Pittsburgh's offensive and defensive passer ratings in their great years?

 

1974 - 48.9 to 44.3 (!!) -- Joe Gilliam played more than Bradshaw that year, and neither of them were good.

1975 - 86.7 to 42.8 - nuff said.

1978 - 81.5 to 51.8

1979 - 76.6 to 56.4

 

I should also mention that comparing rating stats from the 1960s-1970s to the 1980s and after is like comparing dead-ball era baseball to the steroid era. It shouldn't be done. Ever. For chrissake - Fran Tarkenton led the NFL in 1969 with an 87.2 rating. Bradshaw was the league MVP and Player of the Year in 1978 with an 84.7 rating (which was good; Staubach led the NFL with an 84.9 rating).

 

The key thing to remember about this stat: it is RELATIONAL and can't be evaluated outside of its chronological context.

Edited October 16, 2014 by dave mcbride

[Hanoverbills]

When Terry Bradshaw played, defensive backs could bump receivers the whole way down field. I would like to see how good Tom Brady would be against the Steelers of the 70s and with the same rules of 70s. You can"t compare qb ratings from the 70s and know.

Edited October 16, 2014 by Hanoverbills

[Orton's Arm]

Bradshaw's career numbers are irrelevant - what were Pittsburgh's offensive and defensive passer ratings in their great years?

 

1974 - 48.9 to 44.3 (!!) -- Joe Gilliam played more than Bradshaw that year, and neither of them were good.

1975 - 86.7 to 42.8 - nuff said.

1978 - 81.5 to 51.8

1979 - 76.6 to 56.4

 

I should also mention that comparing rating stats from the 1960s-1970s to the 1980s and after is like comparing dead-ball era baseball to the steroid era. It shouldn't be done. Ever. For chrissake - Fran Tarkenton led the NFL in 1969 with an 87.2 rating. Bradshaw was the league MVP and Player of the Year in 1978 with an 84.7 rating (which was good; Staubach led the NFL with an 84.9 rating).

 

The key thing to remember about this stat: it is RELATIONAL and can't be evaluated outside of its chronological context.

 

> I should also mention that comparing rating stats from the 1960s-1970s to the 1980s and after is like comparing dead-ball era baseball to the steroid era. It shouldn't be done. Ever.

 

That limitation is true of quarterback rating. It is not true of yards per attempt. You can meaningfully compare the yards per attempt stats from quarterbacks of different eras. Below is a list of QBs, followed by their career yards per attempt, and the years in which they were active:

 

Sammy Baugh: 7.3 (1937 - 1952)

Y. A. Tittle: 7.4 (1948 - 1964)

Earl Morrall: 7.7 (1956 - 1976)

Bart Starr: 7.8 (1956 - 1971)

Johnny Unitas: 7.8 (1956 - 1973)

Fran Tarkenton: 7.3 (1961 - 1978)

Roger Staubach: 7.7 (1969 - 1979)

Joe Montana: 7.5 (1979 - 1994)

Steve Young: 8.0 (1985 - 1999)

Jim Kelly: 7.4 (1986 - 1996)

Tom Brady: 7.4 (2000 - present)

Peyton Manning: 7.6 (1998 - present)

Drew Brees: 7.5 (2001 - present)

 

Now for some comparison QBs:

Ryan Fitzpatrick: 6.5

Trent Edwards: 6.5

J.P. Losman: 6.6

Kelly Holcomb: 6.6

Joe Ferguson: 6.6

Todd Collins: 6.5

 

Yards per attempt is robust across different eras. Guys like Sammy Baugh and Y.A. Tittle have comparable yards per attempt stats to Drew Brees or Tom Brady. Passer rating is not robust across eras, because it takes completion percentage into account. Any stat that takes completion percentage into account can't let you meaningfully compare a QB who focuses on short passes with one that prefers to throw to intermediate or deep targets. The same flaw which prevents you from using passer rating to meaningfully compare QBs from different eras also prevents you from using that stat to meaningfully compare QBs from the same era, if those two QBs have different playing styles. This is why Kelly Holcomb (preference for short passes) has a nearly identical career passer rating to John Elway (preference for longer passes).

 

The argument for passer rating is that all quarterbacks today gravitate toward short passes to approximately equal degrees. That argument is false. A guy like Trent Edwards gravitates toward dump-off passes far more than a standard-issue NFL starter. The question is whether we should statistically reward him and other QBs like him for bumping up their completion percentages with all those dump-offs. Taking completion percentage into account is an absolutely terrible idea, and will create large statistical distortions unless you know in advance that the two QBs you're comparing are each about equally likely to dump the ball off short.

Edited October 16, 2014 by Orton's Arm

[Greg F]

Why comparing QB's based on quarterback ratings in different era's is futile.

 

[BarleyNY]

ESPN's QBR rating system tries to factor out some of the flaws in the standard passer rating. I'd be interested to see if there is a better correlation.

[PlayoffsPlease]

People have gone astray from the original point. The original point is that in any given game, the QB that has the better passer rating that day, usually leads his team to victory 13/15 times in the most recent week. It is not attempting to correlate a number to winning over a career or even a season.

[dave mcbride]

 

 

> I should also mention that comparing rating stats from the 1960s-1970s to the 1980s and after is like comparing dead-ball era baseball to the steroid era. It shouldn't be done. Ever.

 

That limitation is true of quarterback rating. It is not true of yards per attempt. You can meaningfully compare the yards per attempt stats from quarterbacks of different eras. Below is a list of QBs, followed by their career yards per attempt, and the years in which they were active:

 

Sammy Baugh: 7.3 (1937 - 1952)

Y. A. Tittle: 7.4 (1948 - 1964)

Earl Morrall: 7.7 (1956 - 1976)

Bart Starr: 7.8 (1956 - 1971)

Johnny Unitas: 7.8 (1956 - 1973)

Fran Tarkenton: 7.3 (1961 - 1978)

Roger Staubach: 7.7 (1969 - 1979)

Joe Montana: 7.5 (1979 - 1994)

Steve Young: 8.0 (1985 - 1999)

Jim Kelly: 7.4 (1986 - 1996)

Tom Brady: 7.4 (2000 - present)

Peyton Manning: 7.6 (1998 - present)

Drew Brees: 7.5 (2001 - present)

 

Now for some comparison QBs:

Ryan Fitzpatrick: 6.5

Trent Edwards: 6.5

J.P. Losman: 6.6

Kelly Holcomb: 6.6

Joe Ferguson: 6.6

Todd Collins: 6.5

 

Yards per attempt is robust across different eras. Guys like Sammy Baugh and Y.A. Tittle have comparable yards per attempt stats to Drew Brees or Tom Brady. Passer rating is not robust across eras, because it takes completion percentage into account. Any stat that takes completion percentage into account can't let you meaningfully compare a QB who focuses on short passes with one that prefers to throw to intermediate or deep targets. The same flaw which prevents you from using passer rating to meaningfully compare QBs from different eras also prevents you from using that stat to meaningfully compare QBs from the same era, if those two QBs have different playing styles. This is why Kelly Holcomb (preference for short passes) has a nearly identical career passer rating to John Elway (preference for longer passes).

 

The argument for passer rating is that all quarterbacks today gravitate toward short passes to approximately equal degrees. That argument is false. A guy like Trent Edwards gravitates toward dump-off passes far more than a standard-issue NFL starter. The question is whether we should statistically reward him and other QBs like him for bumping up their completion percentages with all those dump-offs. Taking completion percentage into account is an absolutely terrible idea, and will create large statistical distortions unless you know in advance that the two QBs you're comparing are each about equally likely to dump the ball off short.

What's your point? Seriously. Guys with an 80 rating in 1972 were generally near the top of the league, and they had more ypa. It's a core feature of passer rating. Anyway, they were all throwing deep back then. Also, ae you honestly saying that int rate shouldn't count? The problem with your earlier argument - which you've changed - was that you were comparing ratings across eras. Please explain to me why passer rating differential isn't an indicative stat. Switching the topic to ypa is a red herring.

Edited October 17, 2014 by dave mcbride

[Orton's Arm]

What's your point? Seriously. Guys with an 80 rating in 1972 were generally near the top of the league, and they had more ypa. It's a core feature of passer rating. Anyway, they were all throwing deep back then. Also, ae you honestly saying that int rate shouldn't count? The problem with your earlier argument - which you've changed - was that you were comparing ratings across eras. Please explain to me why passer rating differential isn't an indicative stat. Switching the topic to ypa is a red herring.

 

> Also, ae you honestly saying that int rate shouldn't count?

 

INT rate should count. It was one of the key variables found in the NY Times linear regression.

 

> The problem with your earlier argument - which you've changed - was that you were comparing ratings across eras.

 

My earlier argument has not changed. The ease or difficulty of accumulating passing yardage does not seem to have changed much across eras, as demonstrated by the fact that the yards per attempt stats from the best QBs of the '40s, '50s, or '60s are comparable to the best modern era quarterbacks. If quarterback rating can't be meaningfully used across eras--and I agree it cannot be--it's because it takes completion percentage into account. For this reason, it cannot be used to meaningfully compare any two quarterbacks with differing propensities for dumping the ball off short, regardless of era.

 

> Please explain to me why passer rating differential isn't an indicative stat.

 

Passer rating is a meal with some good ingredients (yards per attempt and INT percentage) and some junk food ingredients (completion percentage). In the aggregate, the effect of the nutritious ingredients will tend to outweigh the effect of the junk. But once you start talking about specific quarterbacks, the junk food aspect of passer rating can quickly lead to problems. If you're looking at an individual quarterback, you shouldn't use a hodgepodge stat like passer rating. Instead, you should break passer rating into its component pieces. Those pieces are air yards, YAC yards, INT percentage, completion percentage, and TD percentage. By examining each piece individually, you'll develop a far more meaningful perspective of the QB's play than would have been the case had you relied solely on passer rating.

 

> Switching the topic to ypa is a red herring.

 

Yards per attempt is not a red herring. To the extent that passer rating is useful, it's because it relies on yards per attempt and interception percentage.

Edited October 17, 2014 by Orton's Arm

[dave mcbride]

> Also, ae you honestly saying that int rate shouldn't count?

 

INT rate should count. It was one of the key variables found in the NY Times linear regression.

 

> The problem with your earlier argument - which you've changed - was that you were comparing ratings across eras.

 

My earlier argument has not changed. The ease or difficulty of accumulating passing yardage does not seem to have changed much across eras, as demonstrated by the fact that the yards per attempt stats from the best QBs of the '40s, '50s, or '60s are comparable to the best modern era quarterbacks. If quarterback rating can't be meaningfully used across eras--and I agree it cannot be--it's because it takes completion percentage into account. For this reason, it cannot be used to meaningfully compare any two quarterbacks with differing propensities for dumping the ball off short, regardless of era.

 

> Please explain to me why passer rating differential isn't an indicative stat.

 

Passer rating is a meal with some good ingredients (yards per attempt and INT percentage) and some junk food ingredients (completion percentage). In the aggregate, the effect of the nutritious ingredients will tend to outweigh the effect of the junk. But once you start talking about specific quarterbacks, the junk food aspect of passer rating can quickly lead to problems. If you're looking at an individual quarterback, you shouldn't use a hodgepodge stat like passer rating. Instead, you should break passer rating into its component pieces. Those pieces are air yards, YAC yards, INT percentage, completion percentage, and TD percentage. By examining each piece individually, you'll develop a far more meaningful perspective of the QB's play than would have been the case had you relied solely on passer rating.

 

> Switching the topic to ypa is a red herring.

 

Yards per attempt is not a red herring. To the extent that passer rating is useful, it's because it relies on yards per attempt and interception percentage.

I think you're missing my point about red herrings. Your emphasis on ypa is a red herring because no one is saying it's unimportant and because it's a core feature of the stat you're deriding. Completion percentage isn't unimportant, btw; there's usually a pretty strong correlation between completion pct and ypa. I think your view is skewed because you focus on Bills QBs too much. Look at Brees, Young, Manning, etc. Also, even while Montana's ypa numbers were usually not stratospheric, what made that offense so good is that their passing plays usually resulted in positive yardage (i.e., he completed a high percentage). That, combined with a high td rate and a low int rate (plus a solid running game), led to an unstoppable offense. It all matters, and that's why the overall rating for both offense and defense is so important. I do agree that ypa is central though.

 

For what it's worth, after 6 games opponents have a 94.3 rating against the Bills D, and our offense has an 83.9 rating. Last year under Pettine, it was 74.9 (the Bills offense had a 75.0 rating).

 

A 94.3 rating for our D would be the worst in team history, I think, although it's worth noting that QB rating is up across the board this year (probably because of the new rules that restrict DBs from playing like they have done throughout NFL history).

Edited October 17, 2014 by dave mcbride

[BarleyNY]

Why comparing QB's based on quarterback ratings in different era's is futile.

 

Great illustration that shows how misleading it is to compare stats of QBs from different eras.

[Orton's Arm]

I think you're missing my point about red herrings. Your emphasis on ypa is a red herring because no one is saying it's unimportant and because it's a core feature of the stat you're deriding. Completion percentage isn't unimportant, btw; there's usually a pretty strong correlation between completion pct and ypa. I think your view is skewed because you focus on Bills QBs too much. Look at Brees, Young, Manning, etc. Also, even while Montana's ypa numbers were usually not stratospheric, what made that offense so good is that their passing plays usually resulted in positive yardage (i.e., he completed a high percentage). That, combined with a high td rate and a low int rate (plus a solid running game), led to an unstoppable offense. It all matters, and that's why the overall rating for both offense and defense is so important. I do agree that ypa is central though.

 

For what it's worth, after 6 games opponents have a 94.3 rating against the Bills D, and our offense has an 83.9 rating. Last year under Pettine, it was 74.9 (the Bills offense had a 75.0 rating).

 

A 94.3 rating for our D would be the worst in team history, I think, although it's worth noting that QB rating is up across the board this year (probably because of the new rules that restrict DBs from playing like they have done throughout NFL history).

 

> Your emphasis on ypa is a red herring because no one is saying it's unimportant and because it's a core feature of the stat you're deriding.

 

But why use quarterback rating at all when YPA is better? (Especially when you mix three parts YPA with one part interception percentage.)

 

> Completion percentage isn't unimportant, btw; there's usually a pretty strong correlation between completion pct and ypa.

 

Why do I need something with a correlation to YPA when I can have YPA straight up? Including completion percentage as one of their variables did not meaningfully increase the predictive power of the New York Times' linear regression model.

 

> I think your view is skewed because you focus on Bills QBs too much.

 

I admit to being more familiar with Bills QBs than with non-Bills QBs. However, let's say there's some hypothetical way a statistical measure could overstate a QB's performance. If someone asks, "How do I know any real QB will ever have his performance overstated in the way you're describing?" I can almost always point to some post-Kelly Bills' QB to give him his answer.

 

> even while Montana's ypa numbers were usually not stratospheric, what made that offense so good is that their passing plays usually resulted in positive yardage

 

I disagree with the way you've phrased this. A three yard dump-off on third-and-long results in positive yardage and an increased completion percentage, but does not result in a good offense. What made Montana's offense so good is that he kept the chains moving. It's better to kill a defense with a thousand small cuts, than it is with a few long bombs. I'll grant that Montana produced better offensive results (and a higher completion percentage) than most other quarterbacks with similar YPA stats. But Montana may be a statistical outlier, not part of some larger pattern. For every guy like him, there are dozens like Trent Edwards or E.J. Manuel. Guys who increase their completion percentage not by throwing short passes to targets moving horizontally, but by dumping the ball off to stationary targets. Hitting a horizontally moving target in perfect stride (as Montana did) is very difficult. Dumping the ball off to someone stationary, less than 10 yards away, is easy. As a general rule--and not just in football--you see easy things attempted far more often than difficult things.

 

> For what it's worth, after 6 games opponents have a 94.3 rating against the Bills D, and our offense has an 83.9 rating. Last year under Pettine, it was 74.9 (the Bills offense had a 75.0 rating).

 

Statistics like the above are reasonably useful. But I'd be more interested to see the yards per pass attempt our defense allowed last year, and the YPA it's allowing this year.

[dave mcbride]

> Your emphasis on ypa is a red herring because no one is saying it's unimportant and because it's a core feature of the stat you're deriding.

 

But why use quarterback rating at all when YPA is better? (Especially when you mix three parts YPA with one part interception percentage.)

 

> Completion percentage isn't unimportant, btw; there's usually a pretty strong correlation between completion pct and ypa.

 

Why do I need something with a correlation to YPA when I can have YPA straight up? Including completion percentage as one of their variables did not meaningfully increase the predictive power of the New York Times' linear regression model.

 

> I think your view is skewed because you focus on Bills QBs too much.

 

I admit to being more familiar with Bills QBs than with non-Bills QBs. However, let's say there's some hypothetical way a statistical measure could overstate a QB's performance. If someone asks, "How do I know any real QB will ever have his performance overstated in the way you're describing?" I can almost always point to some post-Kelly Bills' QB to give him his answer.

 

> even while Montana's ypa numbers were usually not stratospheric, what made that offense so good is that their passing plays usually resulted in positive yardage

 

I disagree with the way you've phrased this. A three yard dump-off on third-and-long results in positive yardage and an increased completion percentage, but does not result in a good offense. What made Montana's offense so good is that he kept the chains moving. It's better to kill a defense with a thousand small cuts, than it is with a few long bombs. I'll grant that Montana produced better offensive results (and a higher completion percentage) than most other quarterbacks with similar YPA stats. But Montana may be a statistical outlier, not part of some larger pattern. For every guy like him, there are dozens like Trent Edwards or E.J. Manuel. Guys who increase their completion percentage not by throwing short passes to targets moving horizontally, but by dumping the ball off to stationary targets. Hitting a horizontally moving target in perfect stride (as Montana did) is very difficult. Dumping the ball off to someone stationary, less than 10 yards away, is easy. As a general rule--and not just in football--you see easy things attempted far more often than difficult things.

 

> For what it's worth, after 6 games opponents have a 94.3 rating against the Bills D, and our offense has an 83.9 rating. Last year under Pettine, it was 74.9 (the Bills offense had a 75.0 rating).

 

Statistics like the above are reasonably useful. But I'd be more interested to see the yards per pass attempt our defense allowed last year, and the YPA it's allowing this year.

 

This is the crux of it -- you really don't watch enough other good QBs, and the Bills have screwed up your perspective. Did you watch the Niners at all from 1981 to 2001 (or so)? They lived off of that sort of play.

 

Let's look at the 2001 Niners, who went 12-4 and were one of the best teams in the league (they lost their playoff game in really tough weather conditions in Green Bay). They finished 4th in yards, 3rd in points, and had a QB rating differential of 25 points (95.0 to 70.3). Jeff Garcia didn't drive the ball deep and had a YPA of 7.0. Yet he had a great TD rate-to-INT rate ratio (6.3 percent to 2.4 percent) and completed a very respectable 63 percent of his passes (bear in mind that this was 2001).

 

The Niners were 10th in the league that year in YPA but 3rd in passer rating, 2nd highest in TD-per-throw percentage, 3rd lowest in INT-per-throw percentage, and 4th in completion percentage. Not coincidentally, they were also 3rd in points scored and 4th in yardage.

 

The bottom line: for the Niners that season (and it was a typical Niners season) the passer rating correlated very strongly -- indeed, virtually exactly -- to the team's level of success on offense.

 

Watch more good teams and notice how they emphasize completing passes, scoring passing TDs, and not throwing picks (as well as putting up a high ypa, of course). The Bills are a bad team to base your observations on. No Bills QB has thrown 25 TD passes since Jim Kelly in 1991, believe it or not.

Edited October 17, 2014 by dave mcbride

[Orton's Arm]

This is the crux of it -- you really don't watch enough other good QBs, and the Bills have screwed up your perspective. Did you watch the Niners at all from 1981 to 2001 (or so)? They lived off of that sort of play.

 

Let's look at the 2001 Niners, who went 12-4 and were one of the best teams in the league (they lost their playoff game in really tough weather conditions in Green Bay). They finished 4th in yards, 3rd in points, and had a QB rating differential of 25 points (95.0 to 70.3). Jeff Garcia didn't drive the ball deep and had a YPA of 7.0. Yet he had a great TD rate-to-INT rate ratio (6.3 percent to 2.4 percent) and completed a very respectable 63 percent of his passes (bear in mind that this was 2001).

 

The Niners were 10th in the league that year in YPA but 3rd in passer rating, 2nd highest in TD-per-throw percentage, 3rd lowest in INT-per-throw percentage, and 4th in completion percentage. Not coincidentally, they were also 3rd in points scored and 4th in yardage.

 

The bottom line: for the Niners that season (and it was a typical Niners season) the passer rating correlated very strongly -- indeed, virtually exactly -- to the team's level of success on offense.

 

Watch more good teams and notice how they emphasize completing passes, scoring passing TDs, and not throwing picks (as well as putting up a high ypa, of course). The Bills are a bad team to base your observations on. No Bills QB has thrown 25 TD passes since Jim Kelly in 1991, believe it or not.

 

> Did you watch the Niners at all from 1981 to 2001 (or so)? They lived off of that sort of play.

 

They lived off of 3 yard dump-offs to stationary targets on third and long?

 

> The Niners were 10th in the league that year in YPA but 3rd in passer rating . . .

 

I'll grant the statistics you provided are strong evidence that Garcia's 49ers had a top-5 offense. Not the 10th best offense indicated by YPA. I also acknowledge that a West Coast offense produces benefits not fully measured by YPA; and that these benefits are measured by passer rating. If Montana-style (or even Garcia-style) play was typical of NFL quarterbacks, passer rating might indeed be a better statistical tool to use than yards per attempt.

 

Below is a useful definition of the West Coast offense.

 

> the West Coast Offense can be described as quick, horizontal timing routes to stretch a defense horizontally, and passing to set up the run.

 

From my own personal experience throwing footballs, I know that it's more difficult to hit a target moving horizontally than one that's moving vertically. The easiest target to hit, of course, is one that's stationary. The question then becomes: how many teams trust their quarterbacks enough to make passes to horizontally moving targets the basis of their offenses?

 

Let's look at the Bills' opponents this season. Tannehill doesn't operate that kind of offense with the Dolphins. Nor Fitzpatrick with the Texans, nor Stafford with the Lions, or Cutler with the Bears. Brady could operate an offense like that if he wanted to. But against the Bills, the Patriots' yardage often came in very large chunks. It's not obvious to me that any of the Bills' opponents did things which wouldn't have been detected by yards per attempt, but would have been detected by passer rating.

 

Let's say a QB dumps the ball off to his RB a lot. Typically, you're not going to be able to sustain 49ers-style many play drives by doing that. You can sustain such drives by throwing quick slants to your #1 WR, and by other throws to horizontally moving targets. But not by always going to your dump-off option. But if a QB keeps dumping the ball off, sometimes he'll get lucky. Sometimes the RB or other dump-off option will generate a really big YAC. That YAC will inflate the QB's quarterback rating and yards per attempt stat. All those completions to RBs will also inflate his completion percentage. I'm not claiming that every non-Brady/Rivers QB we've faced is Trent Edwards. Normally, our opponent QBs provided a mix of short, intermediate, and deep passes. But when they did go short, it was far more likely to be a dump-off to a stationary target, than a pass to a target moving horizontally. The median gain per play will be much lower on a dump-off to a RB than on a quick pass to a WR moving horizontally. Due to the low per-play median gain associated with dump-offs, you're not going to be able to use them to inflict very many "death by a thousand small cuts" drives on opposing defenses. Because dump-offs have a relatively low median gain, and a high standard deviation (due to a few big YAC plays), their efficacy should be measured by yards per attempt--not by any stat which takes completion percentage into account.

[Orton's Arm]

Just to add to my earlier post: suppose you had a RB who obtains exactly 4 yards each carry. If you gave him the ball every time, your team would score a touchdown every drive. (Barring penalties.) This RB's average yards per carry isn't anything out of the ordinary. But his standard deviation is zero; and that--in combination with his respectable yards-per-carry stat--is what makes him special.

 

The 49ers offense of the '80s and '90s had a good, solid average yards per pass attempt. But in addition to that, the standard deviation of their average gain per pass play was relatively low. That low standard deviation, in combination with the relatively high average gain per pass play, better explains their offensive success than a standalone stat like yards per attempt. In their particular case, a low standard deviation per pass play was correlated with a high completion percentage.

 

But that correlation isn't always going to be present. If a QB dumps the ball off to his RB a lot, his completion percentage will be high. But if most of those dump-offs result in 0 - 3 yard gains, and a few result in 30 - 40 yard gains, the standard deviation will be high. The use of completion percentage as a proxy for standard deviation would result in a gross misinterpretation of on-field events--at least for this category of situation.

Edited October 17, 2014 by Orton's Arm

[K-9]

This is the crux of it -- you really don't watch enough other good QBs, and the Bills have screwed up your perspective. Did you watch the Niners at all from 1981 to 2001 (or so)? They lived off of that sort of play. ...

 

They sure did. Just look at the sheer number of receptions Roger Craig had during those years. Hell, he led the league in catches one year with 92. And I'd wager the vast majority of those receptions were caught within 5 yards of the LOS. They just killed on passes to the flats, most of which were caught behind the LOS. That's why I think the "Air Yards Per Attempt" stat is one of the most useless stats out there. Doesn't tell us anything pertinent about a QB or the system he operates; completely disregards it, actually.

 

GO BILLS!!!

[San-O]

Passer rating is often derided as a "misleading statistic". And this information is a very small one week sample size. But 13 of the 15 games were won by the team with the QB with the higher passer rating in week 6. Kyle Orton was pretty good, with a 94 rating, 14'th best out of 30 QBs this weekend. Problem was that Mr. Brady had a 139 rating, second best of the week.

 

At the other extreme Nick Foles was pretty weak with a 79 rating, but his counter part Eli Manning had a 76 rating.

 

The two outliers are wierd, some Ryan Fitzpatrick ended up ahead of Andrew Luck. 109, to 97. That was a weird game over all. And Charlie Whitehurst (87) finished ahead of Blake Bortles (88) m and that was a close rating in a 2 point game.

 

A one week sample is very small, I get that. But 87% is a pretty decent correlation (and I do know the difference between correlation and causation)

 

Brady didn't throw a pick, fumble and get sacked... or DID he?

Edited October 17, 2014 by San-O