They pretty much have to give him what he wants.

Does this type of stuff piss off your teammates? Or do they all just say, "I would do that too if I could get more money."

The numbers say Bulger is a top tier quarterback. And by my estimates he's worth an extra 1.8 wins per season to the Rams. For an explanation, and a list of all 2006 QBs' "wins added" see the link behind my name.

The Rams have a top 6 QB. They need to pay him as such.

Re 2:

I remember reading in Tim Green's book (see link on my name) that holdouts were mostly treated with humor by the team. If I remember correctly (it's been some time since I read) he basically says that anyone with a "legitimate" way out of camp would do well to take it.

Brian,

It looks like your rating system correlates pretty well with DVOA. It's more practical though because it can be easily calculated by anyone and does not require play-by-play data.

You might want to do some historical studies using those numbers, although you might have to tinker with the constants based on era (e.g. completion % has gone up steadily for the last 20 years or so).

they updated the story to say he's getting something like 6 years, 60 mil, 20mil guaranteed.

Re 2:

I remember reading in Tim Green's book (see link on my name) that holdouts were mostly treated with humor by the team. If I remember correctly (it's been some time since I read) he basically says that anyone with a "legitimate" way out of camp would do well to take it.

I love every stat that confirms my idea that Tom Brady is overrated.

Elvis Grbac, Brad Johnson, Kerry Collins, Jon Kitna, Tom Brady.

I'm surprised Bulger's contract wasn't the largest in league history. If Freeney can become the highest paid defensive player in league history in this climate, I'd expect Bulger could have a Manning-like contract.

3: Brian, I like the work you've been doing. Here's a handful of ideas on your QB rating system.

1. "Wins added" sounds like a cumulative stat, but your formula comes from rates: completion rate, sack rate, and interception rate.

Consider two QBs, both of whom put up equally spectacular numbers. Both end up around 2.5 wins added. But one of them has 20 attempts per game (eg Ben Roethlisberger 2005) and one of them has 30 attempts per game (Peyton Manning). The latter is presumably adding more wins.

In the case of Marc Bulger, he was way over the NFL average for attempts, so your wins added estimate is probably low.

2. I know you didn't want to do it, but you're going to need to put in some sort of yards/completion or yards/attempt figure if you want better results (example: Brad Johnson should be way, way lower.) The QB does not completely control completion percentage, interception rate, or sack rate either.

3. Your system doesn't require any play-by-play, so you could go way back to the early days of the league if you know how to manipulate the pro football reference database easily.

4. If you figure out a good way to make this a completely cumulative stat (figuring in attempts in some way) and get some historical data, you could make a list of greatest QBs of all time in terms of wins added.

Keep up the good work.

Bugler?

Elvis Grbac, Brad Johnson, Kerry Collins, Jon Kitna, Tom Brady.

Peyton Manning, Jeff Garcia, David Carr, Mark Brunell?

Yaguar- I agree with your points, but let me clarify a couple things.

1. Yes, wins added does sound cumulative. But what I mean is "given a totally average running game, receivers, and defense, how many wins would this QB's abilities add to his team over 16 games?" Also, I intentionally want to filter out attempts. For example, Kitna had so many attempts because his team was playing from behind. If we didn't account for attempts, he would appear to be highly skilled, when he is only average. I could do a straight-up analysis without considering attempts. It just depends on what you're looking for.

2. I'd like to add "air yards," i.e. pass yards minus YAC. But I can't find good YAC stats by QB. I know FO has them but I don't think they're publicly available. Until I find them, I'm using completion percentage as a proxy (similar to DIPS in baseball).

3. Doug Drinen is the master of historical analysis. I might leave that to him, but it would be very easy to rerun my analysis with your suggestion.

4. I think that was the original intent for the original NFL passer rating. It was invented by an executive of the NFL HOF.

I hate Bugler, always playing Reveille every morning.

Um. Yea. I think your analysis is pretty irrelevant without some yards per attempt figure. While yards per attempt is somewhat correlated with completion percentage, you imply (incorrectly) that it is very dependent on completion percentage. For example (and yes, I'm the asshole who picks one example to debunk an entire study. but you have david carr ranked ahead of Tom Brady for chrissakes!): David Carr, 68.3% completion, 6.3 yards per attempt. Tom Brady, 61.8% completion, 6.8 yards per attempt. While I understand what you're saying when you want to do 'yards in the air' there's also the argument that b/c Brady is a better passer, he attempts deeper throws and normally puts them in spots where his receivers can get more YAC (and vice versa: Carr threw a bunch of dink and dunk passes, which were generally unsuccessful plays and hence should not be valued the same as brady's). Basically it looks like your stats depend too much on completion percentage - even more so than the traditional QB rating.

Until you can get "air yards," wouldn't yards/attempt or yards/completion be better than having no reference at all in your system to the length of the passes? Seeing Brunell and Johnson and Carr so high on the list is jarring.

Don't pass plays designed for high YAC (slants etc.) require a high degree of accuracy and timing from the quaterback? A correctly delivered ball is crucial for the success of those kinds of plays. Why insist on penalizing a quarterback for possessing that kind of skill? That seems arbitrary.

I hate Bugler, always playing Reveille every morning

Perhaps, but he makes some tasty snacks.

Brian,

Good work. To those who think that Carr is overrated in his system, look at it this way: Carr's "dink and dunk" passing last year minimized INT's and sacks. Given the play of the rest of the team (below average in just about everything) I would say he's close to being right. Look at his reply in #10. He computes how much each QB is worth to his team if they had an AVERAGE running game, defense, off. line, etc. Did the Texans have an AVERAGE running game (NOT!!). These factors prob. explain why Brady is lower. He has above average teammates. Just my 2 cents.

Well, this site USED to be safe for work. Good thing I now have "#1 Site for College Girls!" ads on my screen. Super.

16.

I agree with most of the others, its way too dependent on Completion percentage, when YPA is a much better stat. It contains completion percentage AND yardage.

"Air Yards" is just going to make things less accurate. Hitting a receiver in stride is much more responsible for YAC than some skill the receiver has.

re 16:

Fair enough. But I still don't understand how all completions can be treated the same. There just seems to be a logical break - completing 68% of your passes when you don't ever throw it further than 5 yards cannot be more valuable than completing 61% of your passes when you don't ever throw it shorter than 20 yards (an extreme example). And, considering that Y/A were not included in the regression (nor yards in the air), I am led to believe that the findings, although a good step towards a new performance rating, simply do not 'hold water'.

Re 18: I think we should be able to get a good estimate of how much of YAC is due to the QB and how much is due to the receivers. You could see how much variation there is between receivers on the same team. If there is little variation, than it is mostly due to the qb. If there is a lot of variation, then it is mostly due to the receivers.

Some more clarification: I completely agree YPA is THE MOST IMPORTANT STAT in football. But I also think it has a lot more to do with receiver ability than most people realize (see New Orleans '06).

What my system does is uses YPA as the metric for passing effectiveness, and I use the regression coefficients from YPA to help estimate wins.

Then, to isolate QB contribution to YPA, I assign all QBs the league average yds per catch. Next I calculated the effect of a 1-pt increase in comp% on YPA. That yields a nice little equation to see the average contribution of comp% on wins, holding yds per catch constant.

This method discards some data, i.e. "air yards." But, like baseball's DIPS, my theory is that sometimes the model benefits from excluding data with lots of random noise or with lots of external contributions.

Part of my post is an appeal for someone to point me towards YAC (by QB), so we can finally have a valid replacement for QB ratings we've been waiting for.

Thanks for all the great suggestions. It's a work in progress.

Yo rageon, firefox plus adblock equals safe for work.

On a totally different note, so after all these years, Whitey ended up hiding in St. Louis?

"But I also think it has a lot more to do with receiver ability than most people realize (see New Orleans ‘06)."

New Orleans went from Aaron Brooks to Drew Brees in 2006. I think YAC can be just as easily passed off as the QB improvement.

20. All that would tell you is which reciever had more YAC, not why. Theres no way to determine if its because of the receiver, or the routes theyre running, or even coverage schemes.

The patriots are famous for being really bad against 2nd WRs because of coverage schemes. The 2 WR isnt proof that its the WR is the difference, its because the 2 is undercovered.

Unless you take WRs and compare them running exactly the same routes against the same coverage, theres no way to tell, so I think its only fair to say that its both the QB and the WR responsible for it.

Completely ignoring YPA is ridiculous.

re 23: I would agree that ignoring YPA would be ridiculous, but it is not being ignored.

YPA = comp% x (yards/catch)

So half of YPA is being considered, the part most controlled by the QB.

I used to think that hitting a receiver in stride is what made big YAC numbers because I suffered through Kyle Boller for too long.

But I think that's more of a function of man-to-man vs. zone coverage. When QBs/WRs read zone, they tend to do comeback/buttonhook type seam routes that usually result in no YAC.

I think YAC comes from WRs who are able to get separation (we used to say get open) in man-to-man coverage.

re 20: Great idea Dennis.

Re 23: I'm talking about a large-scale study using every team and several years worth of a data. I'm not sure exactly how to do it (I'm sure better minds than me could work out the details), but the gist of it is you take the variance or standard deviation in YAC for the receivers on each team. Then you combine the data for five seasons or however many.

That should allow a lot of the other factors to average out and should give a decent estimate of how much of YAC is due to the QB and how much is due to the receiver. We all agree they both factor into it, but it would be great to have an idea if it's about 50-50 or 80-20 one way or the other or whatever. That would help people like Brian know how much weight they should give to YAC in rating QBs and things like that.

I hate "26 million reasons why." One dollar is not a reason why! Perhaps he has 26 reasons why, as one million dollars seems like a pretty good reason.

Aren't they just all making fun of Owens's former PR woman's comments? I'd never heard that phrase before she said it.

If $26m is 26m reasons, $1 has to be at least a small reason.

Trying to compare Michael Vick’s passing skills to those of David Carr, when 43% of Vick’s passes were thrown at least 11 yards downfield vs. 18% of Carr’s, using INT% and CMP% straight up defies any sort of logic. I don’t think Vick is a very good quarterback, but that just isn’t a reasonable comparison.

It is entirely unsurprising to note that as downfield pass % increases, CMP% decreases and INT% increases. This is an issue that is actually pretty easy to deal with.

Did this in excel, didn’t take long.
Compared INT% vs. % of passes at least 11 yards downfield. Fit a line to it (this line in theory represents what a league-average QB’s INT% would be for any given pass distribution).

Repeat for CMP% vs. % of passes at least 11 yards downfield.

So now we can compare each QB’s INT% and CMP% to something relevant, simply by subtracting LGINT% from his INT% and LGCMP% from his CMP%. Results are unsurprising:

Vick
-0.14256
-3.02532

Young
0.27272
-5.90216

Ben
1.58928
2.16116

Romo
0.6224
7.4878

Hasselbeck
0.73344
-1.30332

Peyton
-1.57272
6.32216

Cutler
0.54384
0.28548

Grossman
1.05488
-4.30564

Palmer
-0.62304
3.21212

Kitna
0.57696
3.31212

Bulger
-1.66784
3.35652

McNabb
-1.16784
-2.54348

Rivers
-1.06232
-2.78904

Leinart
0.24808
1.19976

Favre
-0.00224
-4.91028

Eli
0.51984
-3.39252

Losman
0.48056
0.90632

Delhomme
-0.21944
-0.59368

Pennington
0.53576
2.45072

Brady
-0.4256
-0.5682

Smith
0.88544
-4.35932

McNair
-0.09248
0.35844

Brees
-0.65384
1.33952

Carr
0.54296
1.23912

general notes:
Peyton Manning is awesome, Carr is pick prone while not only slightly above expected average accuracy.

McNair drops down to being an average QB, Eli is terrible, yet still better than Alex Smith.

Hasselbeck, Vick, Roethlisberger, and Young threw the most downfield, and as a result got the biggest boost in relative CMP%. However, only Ben finishes above average after adjustment.

The next three adjustments I'd make would be:
A) add in fumbles. use the FO fumbles/2 and add them to pick total. I don't get why everyone flat out ignores fumbles; must be a huge Culpepper following in the stat community.

B) drop 4th Q INTs from the numbers. Much like running the ball in the 4th quarter doesn't cause winning, throwing picks in the 4th quarter doesn't cause losing (it's caused by losing).

C) Use completions + drops % instead of flat completion %. I don't have these numbers so I won't. But it's an obvious adjustment to make.