Writers of Pro Football Prospectus 2008

Most Recent FO Features

HarrisChr11.jpg

» Film Room: Chris Harris

Is Harris one of the league's top cover corners, or a product of the system in which he plays? Cian Fahey says the answer lies somewhere in the middle.

21 Oct 2011

Varsity Numbers: Adj. Score Revisited

by Bill Connelly

Seven weeks represents a bit of a benchmark for the data we track at Football Outsiders. Brian Fremeau releases his first official OFEI and DFEI ratings, which means the first Off. F/+ and Def. F/+ rankings are released as well. Beyond that, however, we are halfway through the season, and we can begin to dig into the data in a few more interesting ways. Today I am going to revisit a concept I first discussed here last year and used all summer for my team profiles at SB Nation: Adjusted Points (and, therefore, Adjusted Wins).

I have always loved the relatability of a lot of baseball stats. WAR (Wins Above Replacement) boils a complicated measure down to, basically, how many wins a player is worth. Measures like EqA (Equivalent Average) give you something more telling and accurate than, say, batting average. But since people know what a good or bad batting average is, they scale it to where it resembles batting average. Something like DIPS (Defense Independent Pitching Stats) takes figures more reflective of pitching quality and equates them to an ERA-type measure.

Clearly, FO readers have begun to figure out what good or bad DVOA, F/+, S&P+, etc., ratings look like, but the casual reader still might be a little thrown by it. As I was looking into ways to improve our F/+ performances against the spread, I began to wonder what S&P+ might look like in a different format. What would it tell us if we looked at a single-game S&P+ performance in terms of a point figure? This would give us an opponent-adjusted, tempo-adjusted (since S&P+ is a per-play measure) way to judge offenses, in a more recognizable form.

I did not want to overthink this process -- I simply used regression equations to convert the "+" scores into something resembling real scores.

The purpose (what would teams' records look like if everybody played the same perfectly average opponent every week?) and concept (take a team's weekly "+" score and apply it to a format that is easier to understand) of this measure are at least relatively simple. Each team produces an Off. S&P+ number and a Def. S&P+ number in a given week based on the same type of opponent adjustments that make up the full-season rankings. But if you see that Team A had an Off. S&P+ of 109.7 for that week, that doesn't mean much to you out of context. If you see that their Adj. Score for that game was 31.1, however, then you get a much better grasp for how well they played. This gives us a slightly different look than full-season data gives us, simply because it factors in a consistency aspect, as we'll see below.

To see what Adj. Record can tell us about the 2011 season thus far, let's look at the top teams according to the AP poll. But first, a couple of notes:

  • Because this is an opponent-adjusted number, and because opponents continue to play games after you play them, a given team's Adj. Points and Adj. Wins for games already played will continue to shift as the season progresses. Such is life when you're looking at this type of data before the season actually ends.
  • One concept I didn't explain very well the first time around is this: for every game that is played, it is not guaranteed that one team will end up with an Adj. Win and one with an Adj. Loss. Both teams' performances are being compared to the opponent's season averages, so if both teams play well, or if both teams play particularly poorly, then there could be two Adj. Wins in a given game or two Adj. Losses. For instance, in Michigan State's 28-14 win over Michigan, both teams got an Adj. Win. Compared to State's averages, Michigan earned 33.1 Adj. Points on offense and allowed 25.2 Adj. Points on defense. Meanwhile, State earned 27.3 Adj. Points on offense and allowed 5.4 Adj. Points on defense. Both teams played well enough to beat an average team, but Michigan State played better and earned the win.
  • Last year, when I unveiled this concept, it was recommended that I use full-game data instead of just non-garbage time data. I thought about this for a while, and it made sense to me. A "score" comes from the entire game, and that is what is used below.

With that in mind, let's look at some data.

AP
Rk
Team Actual
Rec.
Adj.
W/L
Adj. Off.
PPG
Rk Adj. Def.
PPG
Rk Adj.
Margin
Rk
1 LSU 7-0 7-0 36.2 17 11.7 3 24.5 3
2 Alabama 7-0 7-0 44.4 1 13.1 5 31.4 1
3 Oklahoma 6-0 6-0 36.6 14 11.8 4 24.8 2
4 Wisconsin 6-0 6-0 40.7 6 23.6 33 17.1 10
5 Boise State 6-0 6-0 38.9 7 15.5 8 23.4 4
6 Oklahoma State 6-0 6-0 37.4 10 25.4 40 12.1 20
7 Stanford 6-0 6-0 36.3 16 18.2 11 18.2 8
8 Clemson 7-0 5-2 34.5 24 28.1 63 6.5 38
9 Oregon 5-1 6-0 42.9 2 24.0 35 19.0 7
10 Arkansas 5-1 5-1 33.2 32 25.3 39 7.9 29
11 West Virginia 5-1 4-2 36.4 15 25.9 42 10.5 22
12 Kansas State 6-0 4-2 26.8 72 22.3 25 4.5 47
13 Nebraska 5-1 5-1 33.2 33 26.3 47 6.9 35
14 South Carolina 6-1 6-1 30.5 47 15.5 7 15.0 15
15 Michigan State 5-1 5-1 28.4 60 7.6 1 20.9 6
16 Virginia Tech 6-1 6-1 31.5 42 17.5 9 13.9 18
17 Texas A&M 4-2 6-0 41.1 5 23.0 28 18.1 9
18 Michigan 6-1 6-1 37.3 12 23.2 30 14.0 17
19 Auburn 5-2 5-2 31.2 44 27.8 59 3.4 52
20 Georgia Tech 6-1 5-2 36.7 13 29.1 71 7.5 31
21 Houston 6-0 5-1 37.5 9 27.2 55 10.3 24
22 Washington 5-1 5-1 36.0 18 31.4 94 4.6 45
23 Illinois 6-1 6-1 31.1 45 20.8 20 10.3 23
24 Georgia 5-2 7-0 32.8 37 19.0 15 13.8 19
25 Arizona State 5-2 6-1 33.5 29 26.2 45 7.3 33

Through seven weeks, 16 teams have played well enough to be undefeated against an aforementioned "perfectly average" opponent every week. Ten of them are in the AP top 25 (LSU, Alabama, Oklahoma, Wisconsin, Boise State, Oklahoma State, Stanford, Oregon, Texas A&M, Georgia). The other six are as follows: Baylor (real record: 4-2), Florida State (3-3), North Carolina (5-2), Notre Dame (4-2), Penn State (6-1) and TCU (4-2).

Of the remaining undefeated teams (according to actual record), Kansas State and Clemson have trod the most questionable ground. Adj. Score suggests that Kansas State might have lost to an "average" opponent, given an "average" number of breaks, the weeks they played Kent State and Miami; Clemson, meanwhile, played quite poorly against Wofford in Week 2 and was slightly below average against Boston College.

Adj. Points and Adj. Wins are simply more tools for the overall arsenal, and I really enjoy them. I think they are wonderful for looking at trends within a season, and I plan to do just that in coming weeks.

This Week on SB Nation

F/+ Rankings

As mentioned, this is the first week where F/+ ratings included breakouts for Off. F/+ and Def. F/+. And powered by S&P+, our No. 1 team is a bit of a surprise. As always, full F/+ rankings can be found here.

RK TEAM RECORD F/+ LAST
WK
CHANGE S&P+ RK FEI RK OFF.
F/+
RK DEF.
F/+
RK
1 Boise State 6-0 +36.6% 4 +3 306.5 1 0.250 5 +15.4% 10 +21.1% 1
2 Alabama 7-0 +35.3% 1 -1 297.0 2 0.271 2 +16.4% 6 +18.9% 3
3 LSU 7-0 +32.5% 3 +0 275.0 4 0.320 1 +16.3% 7 +16.2% 6
4 Oklahoma 6-0 +30.4% 2 -2 279.1 3 0.260 3 +13.4% 13 +17.1% 4
5 Wisconsin 6-0 +28.9% 5 +0 274.1 5 0.254 4 +20.3% 1 +8.7% 17
6 Stanford 6-0 +25.4% 7 +1 262.2 8 0.240 9 +12.5% 16 +12.9% 7
7 Notre Dame 4-2 +25.0% 6 -1 269.0 6 0.200 13 +15.6% 9 +9.4% 13
8 Oklahoma State 6-0 +24.7% 9 +1 260.1 9 0.237 10 +8.4% 25 +16.3% 5
9 Oregon 5-1 +23.7% 8 -1 265.6 7 0.191 14 +14.8% 11 +8.9% 16
10 Texas A&M 4-2 +22.6% 12 +2 249.1 10 0.248 6 +18.4% 2 +4.2% 36
11 Michigan State 5-1 +20.4% 18 +7 247.0 12 0.213 11 +8.8% 23 +11.5% 10
12 Michigan 6-1 +17.8% 11 -1 242.2 13 0.185 16 +11.4% 18 +6.4% 30
13 Georgia Tech 6-1 +15.7% 10 -3 237.5 17 0.167 18 +12.7% 15 +3.1% 42
14 Arkansas 5-1 +15.7% 13 -1 230.1 26 0.201 12 +8.8% 24 +6.8% 27
15 West Virginia 5-1 +15.5% 16 +1 221.0 35 0.241 8 +15.9% 8 -0.4% 59
RK TEAM RECORD F/+ LAST
WK
CHANGE S&P+ RK FEI RK OFF.
F/+
RK DEF.
F/+
RK
16 Penn State 6-1 +15.3% 15 -1 248.6 11 0.105 37 -4.0% 79 +19.3% 2
17 Georgia 5-2 +15.0% 17 +0 239.8 15 0.142 27 +13.0% 14 +2.0% 46
18 Clemson 7-0 +14.8% 21 +3 218.1 40 0.241 7 +17.5% 3 -2.8% 73
19 Virginia Tech 6-1 +14.7% 26 +7 236.9 18 0.148 25 +7.6% 29 +7.1% 23
20 Arizona State 5-2 +14.3% 23 +3 227.6 28 0.186 15 +11.0% 20 +3.3% 40
21 Toledo 4-3 +14.1% 28 +7 238.2 16 0.130 30 +7.1% 30 +7.0% 25
22 Temple 5-2 +13.9% 27 +5 236.5 19 0.134 28 +5.5% 40 +8.4% 19
23 North Carolina 5-2 +13.7% 19 -4 232.1 22 0.152 23 +11.5% 17 +2.2% 45
24 Florida State 3-3 +13.7% 24 +0 240.4 14 0.112 33 +5.9% 38 +7.8% 21
25 Ohio State 4-3 +12.3% 40 +15 226.0 31 0.154 22 +0.2% 57 +12.1% 9

"What The...?" Team of the Week

Toledo. They fought No. 25 Ohio State to a statistical draw on the road, they stayed close to No. 1 Boise State longer than most and they romped over No. 22 Temple (a "What The..." team in their own right). Honestly, their worst result came against a BCS team: a disputed overtime loss to No. 67 Syracuse. They have, for the most part, dominated who they are supposed to dominate, and they have popped into the F/+ top 25 because of it. If they dominate Miami (Ohio) and look good in beating both Western Michigan and Northern Illinois at home in the coming weeks, they could potentially hang in the top 30 or so.

Big Movers

Notable Rises

Virginia (23 spots, from 68th to 45th). Georgia Tech was already showing signs of fraying, and if they continue to slide, then that will damage the impact of Virginia's 24-21 win over the Yellow Jackets in Charlottesville. Still, the Hoos are 4-2 on the season and, with home wins over N.C. State tomorrow and Duke in a few weeks, could reach bowl eligibility for the first time since 2007.

SMU (17 spots, from 50th to 33rd). After getting somewhat manhandled (defensively, at least) by Central Florida last year, the Mustangs returned the favor on the Knights, romping to a 38-17 win. They laid an egg against Texas A&M in Week 1, but have looked great ever since.

Ohio State (15 spots, from 40th to 25th). Thanks mostly to the fact that Michigan State has continued to wreck shop defensively, Ohio State can lay claim to playing like at least an average offense for four straight weeks now. They have work to do, but as the previously suspended batch of players work their ways back into the rotation, the Buckeyes should continue to, at the very least, play at a top-25 level.

Washington (15 spots, from 52nd to 37th). The Huskies and quarterback Keith Price have improved steadily since their near-disaster in Week 1 versus Eastern Washington. They took down both Utah and Colorado like a good team should.

Others: Northern Illinois (78th to 58th), Nebraska (41st to 28th), Purdue (91st to 78th), UL-Monroe (103rd to 90th), Miami (45th to 34th).

Notable Tumbles

Florida (28 spots, from 14th to 42nd). Each week, the safety net of preseason projections has drifted away and damaged Florida's overall rankings. This young, banged up and (at the moment) rudderless squad needs to find some direction, and that is difficult to do with a rotation of two different true freshman quarterbacks.

Duke (16 spots, from 59th to 75th). The Blue Devils got pushed around in every way a team can possibly be pushed around in their 41-16 loss to Florida State, which completely negated whatever gains they had managed with good performances versus Tulane and Florida International.

Cincinnati (16 spots, from 33rd to 49th). Like Maryland and Miami, Cincinnati has been all over the map in 2011: looking rather poor versus Tennessee, then spectacular against N.C. State. More than anything, they likely fell this week because of the phasing out of preseason projections that took their great 2008 and 2009 seasons into account.

Tennessee (12 spots, from 20th to 32nd). Poor Derek Dooley. His team fell victim to some crazy endings (LSU, North Carolina) last year, and they have fallen victim to the injury bug this year, losing their two best offensive players in receiver Justin Hunter and quarterback Tyler Bray. Dooley is under heat to start winning games by now (I saw him compared to Alabama's Mike Shula earlier today), but a little luck wouldn't hurt right about now.

Others: Western Michigan (30th to 55th), Central Florida (29th to 51st), Tulsa (44th to 64th), Troy (83rd to 97th), Pittsburgh (35th to 47th), Texas (25th to 36th).

Upset Watch

Rutgers over Louisville (Spread: Louisville -2 | F/+ Projection: Rutgers by 8.7). Rutgers is undefeated in the Big East and, with freshman Gary Nova at quarterback, has played at a stable level recently. Their defense is potentially outstanding, and the offense just needs to stay out of the way. Meanwhile, Louisville is undergoing some serious growing pains with freshman quarterback Teddy Bridgewater getting acclimated to his first-string role. Louisville has not played even an average offensive game since their win over Kentucky on September 17, and Rutgers' is probably the best defense they've faced in that stretch.

Duke over Wake Forest (Spread: Wake Forest -3 | F/+ Projection: Duke by 0.3). Wake is 4-2 but untrustworthy from week-to-week; Duke is, in many ways, the same. F/+ gives this one to the home team, but I'm going to disagree.

Georgia Tech over Miami (Spread: Miami -2.5 | F/+ Projection: Georgia Tech by 2.3). F/+ is still taking into account Tech's spectacular September. But let's just say this game is a good use for Adj. Points. Here is Tech's Adj. Scoring Margin for each game they've played this year, going in chronological order: plus-13.3, plus-10.7, plus-20.5, plus-13.5 ... plus-2.8, minus-3.0, minus-5.0.

Utah over California (Spread: Cal -3 | F/+ Projection: Utah by 4.8). If you can figure out what to expect from either of these two teams from week-to-week, you are smarter than I am.

Favorite Moment of Last Weekend

It pretty much has to be this run, doesn't it? If Trent Richardson goes on to win the Heisman, I think I know the first clip in his montage.

Posted by: Bill Connelly on 21 Oct 2011

2 comments, Last at 21 Oct 2011, 6:33pm by Bill Connelly

Comments

1
by navin :: Fri, 10/21/2011 - 5:35pm

Great stuff again Bill.

Any reason why the two systems differ so much on Clemson? If I had to guess it would be because of big plays. Clemson seems to hit a lot of big plays, which makes their drives seem impressive. But on a play-by-play basis, they have a lot of unsuccessful plays. In one big win earlier this year, Boyd was something like 13-30 but with a ton of yards.