SITE NEWS: We are moving all of our site and company news into a single blog for Sports-Reference.com. We'll tag all PFR content, so you can quickly and easily find the content you want.
Also, our existing PFR blog rss feed will be redirected to the new site's feed.
Pro-Football-Reference.com » Sports Reference
For more from Chase and Jason, check out their work at Football Perspective and The Big Lead.
ShareThisThe Simple Rating System: 2007 Offenses and Defenses
Yesterday, we looked at the simple rating system for each team. One of the nice things about the SRS is that we can also break it down into components, such as offense and defense. To be sure, these are far from perfect -- points scored is just a proxy for offense, and points allowed is just a proxy for defense. But for the most part, they're pretty good.
We can include the offensive strength of schedule ratings, too. The Pittsburgh Steelers scored 393 points last year, and the average team scored 343 points, meaning Pittsburgh scored 2.9 PPG more than average. But the Steelers played the Rams, the Cardinals, the Dolphins, Cincinnati twice, Denver, and Cleveland twice. That's half their schedule against really bad defenses, all of whom allowed more than 3 PPG above average (which you can see on the second chart). The Steelers offensive SOS (i.e., the quality of the defenses the Steelers' offense faced) was just under 2.0 PPG easier than average, which means the Steelers were "morally" about 0.9 PPG better at scoring points than average. The table below shows every team's grade.
Off Rat Off SOS nwe 15.86 0.73 dal 7.81 1.06 ind 6.61 0.18 gnb 5.67 0.17 jax 4.81 0.81 sdg 4.33 0.27 nyg 2.82 1.20 hou 2.54 0.54 cle 2.21 -1.22 nor 1.93 -0.07 ari 1.91 -1.65 cin 1.58 -0.48 det 1.41 1.48 min 1.39 0.27 pit 0.91 -1.97 sea 0.83 -2.05 was 0.19 1.01 phi 0.14 0.82 den -0.18 1.51 chi -0.21 0.60 tam -2.34 -1.53 ten -2.85 0.02 oak -3.48 0.52 nyj -3.99 0.95 mia -4.14 0.86 bal -4.98 -0.48 buf -5.59 0.35 car -5.72 -0.72 atl -5.76 -0.26 ram -6.47 -1.22 kan -7.35 0.21 sfo -9.91 -1.91
As you can see here, the 2007 Patriots topped the '98 Vikings for the best offense in NFL history. The '49ers offense was close to historically bad, too. The Steelers and Seahawks had the easiest schedules for offense, while the Lions and Broncos had the toughest schedules for scoring points.
Here are the defensive ratings:
Def Rat Def SOS ind 5.39 0.08 phi 5.11 2.18 sdg 4.45 0.51 pit 4.30 -0.58 was 4.28 1.97 nwe 4.20 -0.36 tam 3.56 -1.25 ten 3.56 0.43 gnb 3.33 -0.17 min 2.40 0.15 jax 2.02 -0.66 kan 1.89 1.14 dal 1.66 0.28 buf 1.49 1.92 chi 1.44 1.50 sea 0.93 -2.57 nyg 0.44 0.69 nyj 0.30 0.80 car -0.09 -0.09 bal -1.76 0.55 sfo -2.03 -0.96 oak -2.50 0.69 hou -2.52 -0.21 cle -3.28 -1.09 den -3.77 0.11 cin -3.96 -1.59 mia -4.23 1.39 nor -4.45 -1.89 atl -4.85 -0.66 det -4.96 1.10 ari -5.85 -2.60 ram -6.50 -0.81
Of note, the Eagles defense played a pretty difficult schedule, while the Seahawks and Cardinals faced some really weak defenses. This sort of analysis is useful when looking at a team like Seattle. On the surface, it's easy to say the the Seahawks tied for 6th in the league in fewest points allowed. But when you play the schedule Seattle played, that's not so impressive. After adjustment, the Seahawks D falls to 16th best. I know these past couple of days seem like Seahawks bashing, but when you play a cupcake schedule, your raw stats aren't indicative of your actual ability.
This entry was posted on Tuesday, June 17th, 2008 at 4:25 am and is filed under History, Statgeekery. You can follow any responses to this entry through the RSS 2.0 feed. Both comments and pings are currently closed.
Indy--#1 DEFENSE--who'd o' thunk it?
Chase--re: your reply to my comment in yesterday's post: I agree that most people will pick NE for 12-13 wins. But how much predictive value can the GREATEST OFFENSE ever have? We know that their record, points scored, etc. will regress--they just can't be the best ever 2 years in a row--and it's mathmatically impossible that they win ALL their games again. Let's face it, 15-0 was difficult enough--they could have lost game 16 if they weren't the GREATEST OFFENSE EVER. Actually, I think the 11.3 wins SRS predict them for is about right--they could have easily lost vs. Ravens, Eagles, Colts, and Giants in the regular season.
I guess my question in all of this is: if the Pats record last year was 12-4 with a SB loss, would their 2008 predicted wins be any different? (I know their points differential would have to be tweaked slightly to reflect the score difference. But would the predicted wins only drop, say, to 11.1?)
We know that their record, points scored, etc. will regress–they just can’t be the best ever 2 years in a row–and it’s mathmatically impossible that they win ALL their games again.
Not exactly. It's extremely unlikely that they win all their games again, but the probability of that happening is definitely nonzero. After all, if they can do it once, there's clearly nothing that makes it impossible for them to do it. And it's not like probabilities have a memory, the two seasons are independent.
Re 2:
the two seasons are independent.
Not exactly, defensive coordinators have 19 games of film to breakdown and try to stop the offense. On the other hand, the players on the Patriots have a full year of practice and experience with one another. There are probably other dependent variables, but those were off the top of my head.
I think SRS is great. But I'd like to point out these are not really offensive or defensive ratings, but "points scored" and "points allowed" ratings. There's a distinction.
Take the Ravens for example. Their offense was as awful as usual last year. But they also gave up an astonishing 26 fumbles, many deep in their own territory. That's 26 fumbles *lost*, not just fumbles. Think how that would affect points allowed.
Similarly, NE's defense probably appears over-rated by SRS. The offense scored so often, or at least gained favorable field position, that their defense had a much easier time keeping points off the board.
"the two seasons are independent."
Not exactly, defensive coordinators have 19 games of film to breakdown and try to stop the offense.
Ok, but what I meant was that in terms of probability, even though they obviously got very lucky in going 16-0, that alone doesn't make them any less likely to get lucky in the future. Luck doesn't just "run out" like that. Yeah, it's really unlikely that they will go 16-0, but going 16-0 once doesn't exhaust their supply of lucky bounces. It could happen again. I really, really hope it doesn't, but it could.
I think SRS is great. But I’d like to point out these are not really offensive or defensive ratings, but “points scored” and “points allowed” ratings. There’s a distinction.
Yes, there is a distinction, and Chase pointed it out in the very first paragraph of the post:
To be sure, these are far from perfect — points scored is just a proxy for offense, and points allowed is just a proxy for defense.
I agree with Alex.
I really, really hope it doesn't happen again.
Sorry, guess I should have said that it is virtually impossible. However since the probability is so low to begin with, what would be the probability that it would happen again--RIGHT after it happened? (the math would be 50% to the 16th power for the prob. to go undefeated, right?=.00152% chance [ie, 15 chances in 1 million]. To do it again= 2.3 chances in TEN BILLION!!!) Can we agree that technically it is a mathmatical impossibility???
If my math is incorrect, somebody please post the correct #'s. I assumed 50% probability of winning every individual game--I know that it is not exactly that, as the Pats would have greater probability to beat, say, the Dolphins.
Joseph,
While 2.3 chances in ten billion is practically impossible, New England has already gone undefeated once. If you look at the ten billion possibilities, a team won't go undefeated in season one in most of them. So since NE already did that, you don't want to count that against them. That's what Alex means when he says the seasons are independent.
For example, when you play roulette, you may see a 00 on a spin. What are the chances of getting two 00s in a row? It's really, really low -- 1 in 38, squared, or 1 in 1,444. But after the first spin of 00, the odds of a 00 on the next spin is "just" 1 in 38, just like NE's chances of going 16-0 this year are independent of what they did last season.
Also, I think New England has about a 75% chance of winning a given game, since they've averaged a little more than 12 wins a season over the last 7 years. And given that winning percentage, they'd have about a 1 in 100 chance of going 16-0. And as Chase explained above, their chances of going 16-0 this season aren't any lower just because they went 16-0 last season.
Let's say I've got a coin flipping league, and someone goes 16-0 in season 1. In season 2, their chances of going 16-0 are exactly the same as everyone else's: about 1 in 65,000. Doesn't matter what happened the season before, probability has no memory.
Now, if you want to find New England's chance of going 16-0 next season and the season after next, then you'd square the probability of them going 16-0 in a given season. It'd be about 1 in 10,000. So, really unlikely, but stranger things have happened.
Again, I really, really hope this doesn't happen, but it could. We must be prepared for the most dire of possibilities, and not console ourselves with faulty probability calculations!
Most points scored != best offense.
This is not a good enough criteria to judge the best offense in a given season, let alone all time.
A team with a terrible defense but a great offense will probably score LESS since they will have to drive a longer distance down the field.
But a team with a decent defense and great offense will have shorter fields to drive down, which statistically would result in more points.
Again, this criteria is not good enough to determine statistically which team was the best offense of all time.