Quarterback Statistics Commonly Misused

[Oldfan]

this thread is also known as "how to look at stats in a way that jason campbell wont be held accountable for being mediocre".

and by the OPs logic, it is impossible to compare campbell to any other QB other than QBs who have played behind the exact same line/WRs/RBs/coaches/ETC, which at this point is absolutely nobody. LOL

i guess we'll never know if barry sanders was better than trung candidate. after all, they didnt have the same blockers, runningbacks coach, head coach, scheme, quarterback, wide receivers, defense, special teams, or mailing address.

Get off that crap, BLC. Not every damn thread is about being for or against Jason Campbell. This thread is about the use of statistics. It should be debated on logic or not at all.

If you watched them both play and need statistics to determine that Barry Sanders was a better back than Trung Candidate, you need to find something to do with your spare time besides watch football.

I watched Jason Campbell play and thought he did alright last season and I expect him to do better this season, but I would trade him in a heartbeat along with two number one picks for Jay Cutler -- because I've seen him play too.

[Oldfan]

I didn't indulge in the writing, but is this the "Sabermetrics of Football" or to the QB position?

I'm a bit fascinated with baseball sabermetrics

This thread highlighted the QB position, but if you want to read some interesting sabermetric-type stuff on football, I'd recommend this site:

http://footballoutsiders.com/

[Brandon Lloyd Christmas]

Get off that crap, BLC. Not every damn thread is about being for or against Jason Campbell. This thread is about the use of statistics. It should be debated on logic or not at all.

If you watched them both play and need statistics to determine that Barry Sanders was a better back than Trung Candidate, you need to find something to do with your spare time besides watch football.

I watched Jason Campbell play and thought he did alright last season and I expect him to do better this season, but I would trade him in a heartbeat along with two number one picks for Jay Cutler -- because I've seen him play too.

according to you, even if i watched them both play, their statistics are not comparable because of all the parameters which you have set. therefore it is impossible to compare two players at the same position because of all the variables that you have brought into play, your A-H. so comparing cutler to campbell is bunk because they both have so many variables around them.

so for instance, comparing ryan leaf to peyton manning is impossible because of all the variables. or my original example, trung candidate to barry sanders.

and i agree with you that i think trading for cutler is a good idea, but you have to admit that the parameters you set are way to constrictive in terms of comparing players performances. to the degree which you have gone, theres technically no way to compare players unless they are on the same team, for instance brunell and campbell in 06 (and even then one could argue by your rules that because one had portis and one had betts their performances are not comparable).

[Oldfan]

according to you, even if i watched them both play, their statistics are not comparable because of all the parameters which you have set. therefore it is impossible to compare two players at the same position because of all the variables that you have brought into play, your A-H. so comparing cutler to campbell is bunk because they both have so many variables around them.

That's exactly right. At this point, the kind of statistics offered by NFL.com have neligible value as measurements of individual performance and the team stats aren't much better. They are statistics for people who don't understand statistics.

You will get better statistics at footballoutsiders.com but they aren't very reliable either. The stats guys on that site will admit it; their numbers are a work in progress.

I grade QBs mostly on what I can see, just as a scout would do, but I also depend on the opinions of experts to verify my own.

[thedevilhimself]

It is funny when you start getting stats confused with oppinion . As Oldfansays you cannot really soley based on statistics compare two players in two different systems because there simply are too many variables .

You can compare the play of two people on the field because you can make a dynamic assesment of the players based on their environment at the time . Any asesment in that case is subjective and depending on the observer may be fair or biased . In one team is playing against a piss poor defense while the WR and RB are running wild and he has all the time in the world to pass you would expect more from a Qb for example than if the QB is playing on a piss poor team with a leaky offensive line stone hand reciever and RB who eaks out inches per carry .

The problem is people use stats - and I know I do - to try and validate thier oppinions . That would be okay if the stats were controlled and comparable but not so okay if you are bending incomparable stats to make your point .. That is like trying to compare the land speed of different mammals to the density of concrete because hey they are both number right ?

[thedevilhimself]

I still say Cutler isn't worth a trade . His attitude and arrogance will see him as a free agent in 2 years anyway . He would not be half the QB he is right now if not for Branden Marshell . Now he is a player .

[Mania]

Actually I think the qb rating is so antiquated. If you notice on many occasions, the teams with the losing record have qb's with high ratings due to the fact that they have to pass a lot to catch up, therefor the qb has more yards.

This needs to be changed, now!

[Oldfan]

It is funny when you start getting stats confused with oppinion . As Oldfansays you cannot really soley based on statistics compare two players in two different systems because there simply are too many variables .

My thread was inspired by the common misuse of stats on both sides of the Cutler v. Campbell comparision. I love a good debate, but far too many posters are using cherry-picked, useless stats to unfairly trash one QB or the other. That's not real debate. It's just a stupid argument.

[Going Commando]

I'm a firm believer in the 33-33-33 ratio. All facets are equal.

It would be disastrous to spend as much money on special teams as you do on offense or defense. You have to go with what the market dictates. Maybe in the grand scheme of things, a linebacker could be as important to a team as a QB or LT (I don't believe this its hypothetical), but you certainly don't pay them more than a comparable player at those other positions, and you certainly don't draft them ahead of a QB or LT (ahem Aaron Curry). You would be getting bad market value.

[SirClintonPortis]

I agree on a general level with what Oldfan posited in the opening post. This thread is a breath of logical fresh air.

[KDawg]

It would be disastrous to spend as much money on special teams as you do on offense or defense. You have to go with what the market dictates. Maybe in the grand scheme of things, a linebacker could be as important to a team as a QB or LT (I don't believe this its hypothetical), but you certainly don't pay them more than a comparable player at those other positions, and you certainly don't draft them ahead of a QB or LT (ahem Aaron Curry). You would be getting bad market value.

I never said anything about payscale.

Nor did I mean anything about payscale. It's a ratio of how important each facet is to success. I would spend almost as much time looking at special teams as I would offense/defense. In a way, you evaluate special teams when doing offense and defense anyways, as alot of those guys play other positions.

I'd also be spending much more of my resources on kicker/punter and ensuring we have a QB who is a decent backup QB and a great holder.

We have a great snapper, thank god.

[Confucius]

Good thread regardless where you stand on the spectrum of this argument.

I hate to potentially muddy the waters (who am I kidding - no I don't), but I wanted to throw something else out there. Two caveats - one, please understand this post is not a refutation of anyone's argument, just food for thought/debate. Two, I am not a statistician or mathematician - so please feel free to point out errors in my thought process or language/terminology.

Seems to me, one of the oft overlooked element any statistical argument comes from the inclusion of the absolute and finite. In this particular discussion, this becomes apparent in the absolute comparisons of one team to another when using the percentage approach, the comparison of offense to defense as if they come from the same mathematical foundation, and the assumption of larger sample size being the defining criteria for the validity of analysis.

-= Example 1 - Positional % of teams =-

I know OF and others have agreed that these percentages change from one team to another, and are variable based on criteria such as scheme, coaching, etc.

When determining a position or units percentage contribution - does it matter if we're talking about a percentage of what? Yes and no. If we're only picking from a defined and relatively consistent set of criteria - e.g. the Gibbs coached 2004 to 2008 Redskins - then I think the valuation method is fine. Our quarterback generally equals xx% of the team works for me.

The "% of what" question really does matter though when we're talking about being able to assign this value across multiple criteria (e.g. all teams from 2001 to 2008). Take for example the Patriots (pick a year) vs. the Browns (again, pick a year ) - when we start saying that the QB is 10% of the team, it insinuates some level of equality to the position across teams (and I realize while OF never said this directly, it might be assumed by some), such that Derek Anderson is responsible for 10% and so is Tom Brady.

If by some system we start to assign values to the teams relative to each other though (and these are completely made up, but we could use something like DVOA I assume), e.g. if the avg team has a value of 100, then the Browns might have a team value of 50 and the Patriots a team value of 150. Derek Anderson is then worth 5 and Tom Brady has a value of 15. We now have a methodology for comparing one player to another using some sort of semi-statistical evaluation. If we flip the two are we to assume Derek Anderson would still be worth 10% of the Patriots offense and maintain an individual value of 15? I don't know...thoughts?

-= Example 2 - The valuation of Offense and Defense =-

There is a defined, finite limit on how good a defense can be. I realize this may be overly simplistic, but a defense starts at zero, and aims to stay as close to zero as possible, allowing zero yards and zero points.

The offense on the other hand, starts at zero and aims to reach infinity. There is no defined limit to the success of an offense. So when Addicted stars doing his analysis, saying a good offense can overcome a bad defense, I do believe he actually has a theoretical leg on which to stand.

The more I think through this one though, I'm honestly not sure how turnovers, defensive touchdowns, offensive turnovers leading to field goals, etc. will ruin my thoughts on this - may completely blow me out of the water.

-= Example 3 - Larger sample sizes =-

At various points in this thread and others where statistics are discussed, the issue of what represents a valid sample size is usually brought up. In this thread, I believe it was OFs reference to DVOA being sampled across 20 years, hence making it more reliable.

More reliable in adjusting to the mean of the 20 years? Yes. More reliable in predicting the current situation? Not necessarily. For the answer to the 2nd question to be "yes" - we'd first have to prove that the current year is close to the mean of the past 20 years - something which I think is debatable. For large sample sizes to always be more accurate - an unchanging absolute value of whatever you're measuring (in this case - offense vs. defense importance ratios) must not change significantly over the 20 year sample size.

This is case of the sample size actually being too large to be accurate. Human life expectancy is a great example of this. You can skew stats however you like - taking the largest sample size possible (e.g. as long as we've been keeping census data) you'd arrive at a much lower # than just taking the past 10 years of census data - but most would agree that the current decade of census data, despite being a much smaller sample size, is a more accurate representation of the current situation. Again...thoughts?

[Oldfan]

I agree on a general level with what Oldfan posited in the opening post. This thread is a breath of logical fresh air.

Thank you, SCR.

[addicted]

Addicted, I think we've taken our debate on this as far as it can go. I think your position is based on nothing more than an anomaly in the simplistic yardage stat which can be explained by the "bend but don't break" defensive teams in the league who give up lots of yards but not a lot of points. You don't agree, but that's okay. It was an interesting exchange.

Your over simplifying and ignoring current thrends and statistics because you simply don't want to look in that direction and blowing off facts with the lame excuse of "bend don't break" which doesn't answer any real question I brought up and I see that as you just brushing this off and now ending the discussion.

While I agree good exchange but I honestly have no choice but to think a bit less of you now for not taking me on. You posted in this thread before I had a chance to answer you that you were certain my opinion was wrong and now you simply can't disprove it so your walking away. The next time you decide to hammer someone you might want to slow down and let that person respond, rushing like you did made you look foolish.

You've posted thank you and your welcomes tons of times on this thread but dismiss my opinion because you don't want to take it on. I do not feel as if you ever proved me wrong and you need to know that you've haven't won this debate with me. I am certain that the only reason you believe me to be wrong about this has to do with the study you read. Had you done your own research I can't see how you would come to that conclusion. Since you want to walk away I will accept that and move on.

[addicted]

First this was a fantastic post, thank you for sharing it with us.

-= Example 1 - Positional % of teams =-

If by some system we start to assign values to the teams relative to each other though (and these are completely made up, but we could use something like DVOA I assume), e.g. if the avg team has a value of 100, then the Browns might have a team value of 50 and the Patriots a team value of 150. Derek Anderson is then worth 5 and Tom Brady has a value of 15. We now have a methodology for comparing one player to another using some sort of semi-statistical evaluation. If we flip the two are we to assume Derek Anderson would still be worth 10% of the Patriots offense and maintain an individual value of 15? I don't know...thoughts?

I understand where you are going with this. My thought is that each teams positional value is depandant on who that team has on its roster. For example it is foolish to say that the Patriots success relys on the RB position but it wouldn't be foolish to say the Redskins success relys on the RB position. You simply can not say that every team values its positions the same. The Oakland Raiders, NY Giants, Washington Redskins, Tennessee Titans to name a few teams rely on the RB position much heavier then other teams like the Patriots, the Colts, the Saints, the Cardnials.

With that end in mind I believe that if you are going to do a situational team by team analysis you will find each percentage is different as you go team by team. So that to me asks the question, is it possible to assign the value of position league wide as Oldfan suggested.

To me the answer is yes you can but it wouldn't be easy to do. The way I would try and tackle this would be like this:

1. First you would need to know and the collect the following information:

a. Yardage average for each team per game.

b. The QB stat yardage per game

c. The RB stat yardage per game

Once you know that you would be able to determine the percentage of yardage per game each position was responsible for. For example and to keep this simple say the Redskins had an average of 220 yards per game offensively. From your research you find out that Jason Campbell averages 100 yards per game and Clinton Portis averages 120 yards per game. That would mean that on a 100 point scale Clinton Portis averages 60% of the teams yards and Jason Campbell averages 40% of the teams yards. You then would know for the Washington Redskins that the RB is relied on more heavily then the QB position is.

2. You'd do this for all of the 32 teams.

3. Next you'd need to do that for a set peroid of years. I'd do it for 10

4. Then you'd review the numbers and come to some conclusions. I believe what you'd find is that offensively the QB position normally produces at least 60 to 70 percent of teams offensive yardage on average. The WR's numbers are contained within the QB's numbers so they wouldn't be needed. If you find that the QB numbers are as I believe them to be, you will know how to grade the RB and the QB percentage wise.

The thing I wouldn't know how to guage to see what percentage of worth to a teams success would be the teams offensive line. Even if a QB is the highest sacked QB in the league in a season its still well within the realm of possibility that this QB could easily be producing 80% of his teams yards. So how would you judge that?

-= Example 2 - The valuation of Offense and Defense =-

There is a defined, finite limit on how good a defense can be. I realize this may be overly simplistic, but a defense starts at zero, and aims to stay as close to zero as possible, allowing zero yards and zero points.

The offense on the other hand, starts at zero and aims to reach infinity. There is no defined limit to the success of an offense. So when Addicted stars doing his analysis, saying a good offense can overcome a bad defense, I do believe he actually has a theoretical leg on which to stand.

The more I think through this one though, I'm honestly not sure how turnovers, defensive touchdowns, offensive turnovers leading to field goals, etc. will ruin my thoughts on this - may completely blow me out of the water.

I agree. Thanks for saying I have a leg to stand on, heh I know I do. I know that my analysis is correct. A potent offense can overcome a bad defense each and every year in the NFL where as a Defense/Special teams can not do so as easily. I don't think its possible to review Defenses and Special teams like that.

-= Example 3 - Larger sample sizes =-

At various points in this thread and others where statistics are discussed, the issue of what represents a valid sample size is usually brought up. In this thread, I believe it was OFs reference to DVOA being sampled across 20 years, hence making it more reliable.

More reliable in adjusting to the mean of the 20 years? Yes. More reliable in predicting the current situation? Not necessarily. For the answer to the 2nd question to be "yes" - we'd first have to prove that the current year is close to the mean of the past 20 years - something which I think is debatable. For large sample sizes to always be more accurate - an unchanging absolute value of whatever you're measuring (in this case - offense vs. defense importance ratios) must not change significantly over the 20 year sample size.

This is case of the sample size actually being too large to be accurate. Human life expectancy is a great example of this. You can skew stats however you like - taking the largest sample size possible (e.g. as long as we've been keeping census data) you'd arrive at a much lower # than just taking the past 10 years of census data - but most would agree that the current decade of census data, despite being a much smaller sample size, is a more accurate representation of the current situation. Again...thoughts?

That was the point I tried to make earlier. Looking back at data 20 years ago doesn't mean anything since the leagues changed so much since then. You should limit any research to today's NFL I believe to a maximum of 10 years but I prefered the last 6 seasons. Looking at data that old makes little sense to me.

[Oldfan]

Confucious,

Let's suppose that a car manufacturer commissions an extensive survey of new car buyers to determine the four most important factors that determine the car buying decision. The percentage of buyers making the factors their top criterion might look like this:

Factor A = 40%

Factor B = 30%

Factor C = 20%

Factor D = 10%

Total 100%

When we want to know the IMPORTANCE of the factors involved in anything, the 100% scale is useful.

From my OP: All fans agree that the QB position is the most important, but many fans seem to overrate the value. So, how important is the QB to winning? Let's use 100% to represent the whole and try to break down the factors.

So, we are talking about the importance of each factor I listed to winning football games.

There is a defined, finite limit on how good a defense can be.

Your point is not relevant because you are confusing the importance factor with the grading element. In my estimate, the importance of the defense to a winning effort remains at 34% whether the defense grades out as an A, B, C or D -- just as Factor A remains at 40% importance to the car buyer whether the manufacturer's car grades out as an A, B, C or D in satisfying Factor A.

And, just incidentally, you are mistaken on the grading. There is no logical reason to expect that a defense cannot be built to stop any offense no matter how good it is. The closer the offense gets to the infinite, the zero points defense potential rises to match it (the position of zero isn't fixed).

[sample size] More reliable in adjusting to the mean of the 20 years? Yes. More reliable in predicting the current situation? Not necessarily.

It is just as important to stop the run as it is to have an effective running game.

It is just as important to stop the pass as it is to have an effective passing game.

Can you counter the logic of either of those statements?

Can you think of any reason that either of those statements won't remain true over time?

We have 20 years of data supporting what we could logically expect -- in winning football games, defense and offense are equal in importance.

[Confucius]

OK - that helps to explain it. Good analogy.

Seems like maybe you're talking a priori knowledge - e.g. logic would dictate if my offense is important your defense would be equally important - where I've moved into the realm of a postiori by bringing in the weighting of empirical trending (e.g. offenses are more productive in this era).

So even though my thought process was centered around a numerical output - it doesn't change the relative values (importance) of the input. In other words - Even though offenses now put up much higher numbers (and by correlation, make defenses seem less effective) - the relative % importance in the outcome of a game hasn't changed.

Think I gotcha.

[Oldfan]

Think I gotcha.

Good man.

[Confucius]

lol - thanks.

So now we know that logically speaking, in terms of importance in the game of football, because they are diametrically opposed, defense must equal offense. The actual % value can vary and is up for debate - but they must be equal.

That being said - when the game is played and there is a winner and loser - curious as to what would be a good logical/statistical/mathematical (or car analogical ) way to explain it? I realize I'm potentially going off topic here - but this is one of the better threads of this type I've seen in a while.

Where I'm trying to go - is despite the equal importance - there was some inequality in efficiency and effectiveness for there to be a winner and loser. If we can find a way to put solid number theory on that, can we then derive the root cause of this imbalance, and maybe start to find a way to actually compare individual players. Or are we forever stuck in the loop of "no we can't b/c nothing is ever objective and exists in an independent vacuum"? (And honestly, being a fan of subjectivism I tend to say we're stuck in the loop )

With that potential derailer thrown out there I'm out for a while. Looking forward to some responses.

[superozman]

Oldfan – I’d like to hop in this discussion, however, I am confused at what you are trying to convey. I have gathered the following:

1. You stated the QB stats commonly misused due to you’re A-H factors:

a. Sacks

b. Completion %

c. Interceptions

d. Touchdowns

e. Points per game

f. QB Rating

MY TOUGHTS – First I’d start with this quote – “So, stats comparing Jason Campbell to, say Jay Cutler or Matt Cassel, are pretty much useless” Really, honestly? Comparing stats between QB’s gives you your first gauge to delve deeper into the issue. Here’s my question to you. If you lived on the moon for the past 10 years. And I gave you QB STATS of QB Rating, Touchdowns, Interceptions, Completion percentage, Yards, and games played…Who would you pick as your top 3 quarterbacks looking at stats alone. Now tell me the 3 QB’s you chose would differ greatly if you looked deeper into everything. YOU CAN COMPARE.

I do agree with you, people do use too many arbitrary statistics for arguments. I hate that.

“if Jim Zorn…useful QB stats”. Whole-heartedly agree. However, it would be ridiculous to do…unless…you would propose to the NFL this job and you’ll do it under a volunteer basis haha. 

2. Overrated Turnover Stat

a. overrating conservative offenses with game manager QBs

I, IMHO, can definitely disagree with this – how can you UNDERrate the turnover stat??? Seriously. A good quarterback will PROTECT the football. If I look at stats, and see a QB, over time, consistently throws 20 tds, and has 17 picks and 3 fumbles a year…I don’t care if that means +60 in your terms set above, I don’t want a Quarterback coughing up the ball that much.

3. Importance of Quarterback (and other spots)

21% = Coaching

34% = Defense

11% = Special Teams

12% = O line

12% = RB and receivers

10% = QB

This is just your opinion. My opinion is it varies TOO MUCH per team to even attempt this. In a reply I know you stated as an average, but what the heck good is an average when the +- for a team to those percentages varies so greatly. And it varies per game. There is no point to your percentage rating, and IMHO proves nothing. And this seems to be the most discussed.

[jrockster21]

So what you did, was make a bunch of arbitrary numbers up to prove that stats don't really prove anything?

LOL.

Oldfan, you just spent a whole lotta time on nothing.

[Oldfan]

lol - thanks.

So now we know that logically speaking, in terms of importance in the game of football, because they are diametrically opposed, defense must equal offense. The actual % value can vary and is up for debate - but they must be equal.

That being said - when the game is played and there is a winner and loser - curious as to what would be a good logical/statistical/mathematical (or car analogical ) way to explain it? I realize I'm potentially going off topic here - but this is one of the better threads of this type I've seen in a while.

Where I'm trying to go - is despite the equal importance - there was some inequality in efficiency and effectiveness for there to be a winner and loser. If we can find a way to put solid number theory on that, can we then derive the root cause of this imbalance, and maybe start to find a way to actually compare individual players. Or are we forever stuck in the loop of "no we can't b/c nothing is ever objective and exists in an independent vacuum"? (And honestly, being a fan of subjectivism I tend to say we're stuck in the loop )

With that potential derailer thrown out there I'm out for a while. Looking forward to some responses.

I have no problem going off-topic at this point, but I won't have much to offer except my pessimism.

Football is the consummate team sport. The defense interacts with and influences the efficiency of the offense and special teams and vice versa. So, how do we isolate the efficiency of one unit from the others?

I think that an accurate ranking of the team units is a huge undertaking and only after we have those stats as a basis could we think about isolating individual stats which is far more difficult.

[Oldfan]

superozman: . If you lived on the moon for the past 10 years. And I gave you QB STATS of QB Rating, Touchdowns, Interceptions, Completion percentage, Yards, and games played…Who would you pick as your top 3 quarterbacks looking at stats alone. Now tell me the 3 QB’s you chose would differ greatly if you looked deeper into everything. YOU CAN COMPARE.

On the moon, I couldn't answer your question. You have given me worthless measurements. You might just as well have said that the volume of a room on earth is 80,000 cubic feet then asked: How high is the room?

“if Jim Zorn…useful QB stats”. Whole-heartedly agree. However, it would be ridiculous to do…unless…you would propose to the NFL this job and you’ll do it under a volunteer basis haha.

I offered that hypothetical to explain what would be needed to come up with useful QB stats.

I, IMHO, can definitely disagree with this – how can you UNDERrate the turnover stat?

Very easily. Turnovers are worth are four points, touchdowns are worth seven. In the TD/INT ratio, if you treat turnovers as equal in value to touchdowns, you have overrated them.

In a reply I know you stated as an average, but what the heck good is an average when the +- for a team to those percentages varies so greatly.

My purpose has been explained. Your question indicates you haven't grasped it. We are not concerned here with the question of how important Tom Brady is to the Patriots, we are concerned with how important are QBs to the team effort so what we want is an average.

[superozman]

On the moon, I couldn't answer your question. You have given me worthless measurements. You might just as well have said that the volume of a room on earth is 80,000 cubic feet then asked: How high is the room?

Completely missed my point haha. Lets rephrase. If I come back from the moon. And I am given just those stats for all the QB’s that played in that time and picked the top 3 QB’s, based on STATS alone. I would pick the same ones as a person with the most in depth analysis.

I offered that hypothetical to explain what would be needed to come up with useful QB stats.

Yea – those would be the best stats, I agree.

Very easily. Turnovers are worth are four points, touchdowns are worth seven. In the TD/INT ratio, if you treat turnovers as equal in value to touchdowns, you have overrated them.

No, because there are other VARIABLES besides points that your not including. If you want to look so in depth at stats, but then treat a value as set in stone, your disproving your original analysis. It’s not as simple as you say. What does your +4 take into consideration? Is +4 the average of all touchdowns/fg’s/no points that the turnovers have created. Does it include the fact of the TD’s/FG’s that COULD have been scored. Does it take into consideration the fact of turnover will change the pace of a game, mentally wear down a team? So your saying a 20/20 TD-Turnover ratio is +60? Terrible. So your saying 20 TD’s (140 points) doesn’t get Nullified until 35 turnovers are created? (140 points). Terrible.

My purpose has been explained. Your question indicates you haven't grasped it. We are not concerned here with the question of how important Tom Brady is to the Patriots, we are concerned with how important are QBs to the team effort so what we want is an average.

That’s why I posted, haha I haven’t grasped it at all. BECAUSE YOU CAN”T POSSIBLY PUT A NUMBER TO IT. If in a game you are playing the Patriots, your DEFENSE AND DEFENSIVE COACHING MUST stop the Patriots to win, so you would put a higher importance on that, and lower importance on the QB. If your playing the Lions, you can put a higher importance on the QB to win the game because your defense can slack against ORVLKERJFISKY. In emphasizing the QB you may throw 35 passes and put up 30 total points. However you can lower your importance on the QB, put a higher importance on the Defense, throw a total of 5 passes and win the game 10-0 because you dominated the defensive side of the ball.

[HBnotBlades]

This is a fantastic thread and I'm sorry that I've missed out on it for so long. There's some great discussion going on that usually has no place in your average extreme skins thread. Anyway...

superozman: . If you lived on the moon for the past 10 years. And I gave you QB STATS of QB Rating, Touchdowns, Interceptions, Completion percentage, Yards, and games played…Who would you pick as your top 3 quarterbacks looking at stats alone. Now tell me the 3 QB’s you chose would differ greatly if you looked deeper into everything. YOU CAN COMPARE.

On the moon, I couldn't answer your question. You have given me worthless measurements. You might just as well have said that the volume of a room on earth is 80,000 cubic feet then asked: How high is the room?

Oldfan, you're right in thinking that you can never compare QB's on an absolute, definitive scale, but you can't deny that the QB who has a higher rating, more TD's, completion %, etc. wouldn't have a higher probability of future success, even if placed with a different supporting cast. These are just measurements taken in the presence of noise, and in the case of the NFL, there is a very high noise level due to all the things that go into QB success. It's still possible to get meaningful data and the Football Outsiders DVOA ratings that were mentioned by Confucius are an excellent example of how to do this.

OK, on to one of the other topics...

So now we know that logically speaking, in terms of importance in the game of football, because they are diametrically opposed, defense must equal offense. The actual % value can vary and is up for debate - but they must be equal.

That being said - when the game is played and there is a winner and loser - curious as to what would be a good logical/statistical/mathematical (or car analogical ) way to explain it? I realize I'm potentially going off topic here - but this is one of the better threads of this type I've seen in a while.

If we look at this problem on a game by game basis with both teams assumed equal at the start of the game, then yes, offense and defense (leaving out special teams now for simplicity) are equally weighted. The team with the most points wins, so if the offense scores more points than the defense allows, said team would obviously win. Team A's offense is by definition diametrically opposed to Team B's defense.

I also think you can look at this from a prediction perspective and compare the anticipated performance of offenses and defenses. Let's assign percentages to the performance of an offense or defense with the 0th percentile being that particular offense or defense's typical performance. Thus, the Lions defense 50% performance would be far worse than the Steelers defense 50% performance. I was inclined to use 0% as the mean (then using +/-), but everyone's talking about percentages adding to 100% so I'll stick with that theme. Although here, a 100% total would mean a perfect performance for both the offense and the defense and a bigger blowout than we've ever seen.

So if our offense performs at if average 50% level for a game against the lions, the defense may only have to perform at a 25% level for us to win. However if the offense performs at a 50% level against the steelers, we would likely need the defense to have an 80% exceptionally high level performance to win.

Really, I'm a big fan of the analysis using the DVOA statistics (which do break down by position). Using those, I'm sure its possible to break down the likelihood of winning a game when each position plays plays some percentage above or below its typical DVOA performance for a game. These would be different for every team as mentioned early on, because every team has a different score for each position. Really I think you could just normalized the DVOA scores to a scale of 100 and then you have your pie chart of importance by position for a given team.

However there's still the factor of coaching...