QB Rating scaled to Strength of Schedule

[big_neil]

QB's naturally put up better numbers against weaker teams. Therefore, it seems to make sense to scale their rating by their opponents winning percentage. That is, if a QB has a 100 rating against undefeated teams, he'd have a 100 scaled rating.

Take Peyton Manning for example, his rating is about 105, but Indy's strength of schedule is only .370, the weakest of any team since 1998. So his scaled rating is:

105*.370= 38.9

Drew's rating is about 92, but Dallas strength of schedule is .505, so his rating is

92 * .505 = 46.6

Now when you look at his 75 rating for the last game against 9-2 Denver, his scaled rating is:

75 * .82 = 61.7

While on the rest of the season Drew's rating was 94.4 against 47-53 opponents or:

94.4 * .470 = 44.3

So while it may seem like his rating wasn't that good, it was one of his better performances when you consider the opponent.

 

[Eskimo]

Actually, this doesn't make much sense at all.

I would agree it would make sense to scale against strength of competition, but the scaling shouldn't be according to record. It'd be more reasonable to scale according to the performance of the pass defense of the opponents.

 

[Hostile]

QB rating is the single most useless individual stat that the NFL tracks.

JMO

 

[big_neil]

Here is Drew's scaled ratings for his games this season, sorted by highest, see if you think this shows his best performances in order:

SD 86
NY 65
DE 61
PH 48
PH 48
WA 45
SE 40
AR 36
OA 27
DE 23
SF 18

 

[big_neil]

It'd be more reasonable to scale according to the performance of the pass defense of the opponents.

Good idea. Can you provide those?

 

[Eskimo]

Good idea. Can you provide those?

I think this is an idea that is probably best left to statisticians. It is more along the lines of the analysis that Fox Sports does when they make up their power rankings with their DVOA type stuff. It is also the stuff that baseball sabermetricians love and have pioneered.

It may already exist out there. Maybe check Fox sports website and see if they have a ranking you could put to use.

 

[Alexander]

QB rating is the single most useless individual stat that the NFL tracks.

JMO

Most statistics are useless. They only serve speculative purposes for the most part.

In a game with as many team variables, it does not make much sense to place full weight on any of them.

This is not baseball.

 

[jterrell]

QB's naturally put up better numbers against weaker teams. Therefore, it seems to make sense to scale their rating by their opponents winning percentage. That is, if a QB has a 100 rating against undefeated teams, he'd have a 100 scaled rating.

Take Peyton Manning for example, his rating is about 105, but Indy's strength of schedule is only .370, the weakest of any team since 1998. So his scaled rating is:

105*.370= 38.9

Drew's rating is about 92, but Dallas strength of schedule is .505, so his rating is

92 * .505 = 46.6

Now when you look at his 75 rating for the last game against 9-2 Denver, his scaled rating is:

75 * .82 = 61.7

While on the rest of the season Drew's rating was 94.4 against 47-53 opponents or:

94.4 * .470 = 44.3

So while it may seem like his rating wasn't that good, it was one of his better performances when you consider the opponent.

Its an interesting concept but you apply it wayyy to liberally with scaling.

Its probably a nice consideration to factor into the ranking itself.

With many stats when applied to only one player they make sense but when looking league wide they lose their meaning. Manning is a good example. Yes his QB rating may be inflated by soft opponents but it isn't inflated to the tune of 60 points, lol.

What might work is to use the defensive pass ratings of each team and then figure the mean, i.e. teams 16 and 17. As you move away from those either subtract or add 1 point for each position.

But even that can be problematic. Denver for instance gives up a fairly high amount of yardage passing. This is because you can't run on them so teams pass alot. But they also get a ton of picks. In fact they are tied for 3rd in INTs.

 

[jterrell]

QB rating is the single most useless individual stat that the NFL tracks.

JMO

The current incarnation of QB rating does need overhauling but is a far better judge of success than Joe Blow's eyes.

What I'd like to see added is a factor taking into account first downs and passing performance on 3rd downs. And consideration given for catchable balls as well as half ending heaves.

Over the course of a full season generally the highest rated qbs are the most successful. So the system is not perfect but isn't fatally flawed or useless.

The top 5 currently are:
Carson Palmer, Peyton Manning, Drew Brees, Tom Brady, Marc Bulger.

I would not argue any of those as legit top 5 QBs this season. The 26-30 is also accurate with Brooks, Harrington, Frerotte, Losman, Orton.

I would much rather use QB rating as a gauge than say some GB fan telling me how great Favre is and its only because of his WRs and OL that he ranks 16th. The "reason" he is ranked 16th is his inordinately high INT number.

 

[big_neil]

Yes his QB rating may be inflated by soft opponents but it isn't inflated to the tune of 60 points, lol.

Yes, I didn't want to confuse everyone but I was going to apply a normalization by adding 50. That way the scores look more like QB Rating. Given that the best teams record vs. opponents is about .550, the highest Manning could have achieved is about 58.

Then we'd have:

P. Manning 88.9
D. Bledsoe 96.6

 

[big_neil]

The true normalization factor should be a function of the leagues best opponent average. 50 would assume .500 opponent record is the best. But if the best were .750 the normalization would be 25. So the formula would be 100-(League Best Pct). So it would be more like 47 as a normalization:

P. Manning 85.9
D. Bledsoe 93.6

But the difference is the same regardless of a flat norm. factor.

 

[Sarge]

Actually, this doesn't make much sense at all.

I would agree it would make sense to scale against strength of competition, but the scaling shouldn't be according to record. It'd be more reasonable to scale according to the performance of the pass defense of the opponents.

You can't go by strength of opponent or opponent defense. Peyton Manning is reponsible for running up big numbers against these teams.

Why should the next QB in line have the disadvantage of the damage done by Peyton Manning?

 

[big_neil]

You should at least factor out head to head, or beating a team lowers your opponent average and brings down your rating.

 

[Sarge]

You should at least factor out head to head, or beating a team lowers your opponent average and brings down your rating.

Factor out head to head how? The strength of schedule can't even be determined until the last game has been played.

One QB can go up against an 0-4 team, in 4 weeks, another QB can be going up against the same team who is now 4-4.

If you are going to use head to head - then you can't even have a QB rating until the end of the year.

As Hos stated, the rating system is a joke, but using head to head would be making a bad system worse, if that is even possible.

 

[big_neil]

Well if Indy is undefeated, they are 10-0. So their opponents are 37-53 overall, but 37-43 versus the rest of the league, so their record vs other teams is .460, so Manning's scaled, normalized rating would be

.46 * 105 + 47 = 96.3

Dallas opponents are 56-55, but factor out the 4-7 record opponents have and it's 52-48 (.520)

So Drew's would be:

.52*92+47= 94.9

 

[big_neil]

So Drew's scaled, normalized rating vs Denver with head to head removed is

.8*75.2+47=107.1

 

[Sarge]

Well if Indy is undefeated, they are 10-0. So their opponents are 37-53 overall, but 37-43 versus the rest of the league, so their record vs other teams is .460, so Manning's scaled, normalized rating would be

.46 * 105 + 47 = 96.3

Dallas opponents are 56-55, but factor out the 4-7 record opponents have and it's 52-48 (.520)

So Drew's would be:

.52*92+47= 94.9

This can get into a very complex scenario here. Suffice it to say, I am of the opinion that you can't use strength of schedule in the equation and if you do, it can't be added in until the end of the year which would be a waste of time.

 

[SmashFactorGolf]

Its like an stat----------the longer term it is run and or kept track of: the more validity it will tend to have.............kinda like day to day weather vrs. long term climate..........with the later being highly predicable and the former pretty variable

 

[THUMPER]

QB rating is the single most useless individual stat that the NFL tracks.

JMO

Actually, I believe that ranking defenses based on total yardage is the most useless stat.

 

[Hostile]

Actually, I believe that ranking defenses based on total yardage is the most useless stat.

Please note I used the word "individual" as a clarifier.