A word about Adj. Scores

I've noticed something both encouraging and discouraging about a lot of the comments to my 2013 season previews, both in the threads themselves and at other sites.

Good: More people seem to be talking about and reacting to the Adj. Scores measure, one I like quite a bit.

Bad: Nobody really understands what Adj. Scores are.

That's my fault. Communication is key, and if this many people are misunderstanding something, then I haven't explained it well enough. So let's talk a bit more about it.

The short version

Adj. Score takes your performance in a game and adjusts it for pace, quality of opponent, and general luck. Instead of gauging your performance versus the opponent at hand, it looks at how you would have done -- how many points you would have both scored and allowed -- against a perfectly average opponent.

The long version

1. If you are reading this, you probably know by now that my go-to measure of choice is the S&P+ system of ratings I house at Football Outsiders. Teams are officially evaluated by their full-season S&P+ ratings, which encompass every non-garbage time play from every game. Each game and category also have single-game S&P+ scores. They are adjusted for the opponent at hand and calibrated to the same "+" scoring system as the full-season version, where 100.0 is perfectly average, anything lower is below average, and higher is above average.

2. Back in 2010, I decided to play with single-game S&P+ scores to perhaps make them a bit more accessible and interesting. I calibrated the single-game scores so that they would land on the same scale as regular, old points scored. The average number of points scored in an FBS college football game in 2012 was about 28, so a single-game offensive S&P+ score of 100.0 would be the same as about 28.1 Adj. Points.

3. Since your offense and defense end up with single-game S&P+ scores (and, therefore Adj. Points scores), and since 100.0 = perfectly average, then I decided it would be interesting to see which teams played good enough to beat a "perfectly average" team each week of the season. If the Adj. Points you "scored" were greater than the Adj. Points you "allowed," then you got an Adj. Win.

4. Single-game scores can skew averages, and these are in no way part of the overall ratings system, but I realized pretty quickly that you can get a pretty nice look at in-season trends by using Adj. Scores. For instance, look at today's Kansas State preview. For the season as a whole, KSU fielded both a top 20 offense and defense according to F/+ (the combination of my S&P+ and Brian Fremeau's FEI) and a top 40 offense and defense according to S&P+. Breaking S&P+ out into passing, rushing, down/situations, red zone, etc., tells us quite a bit. And with Adj. Points, we can learn a bit more.

Adj. Points Per Game (first 4 games): KSU 35.9, Opponent 29.0 (plus-6.9)
Adj. Points Per Game (next 5 games): KSU 39.9, Opponent 22.5 (plus-17.4)
Adj. Points Per Game (last 4 games): Opponent 24.6, KSU 24.2 (minus-0.4)

From this, we see that KSU's offense was about eight points better than average in September, 12 points better than average in October(ish), and four points worse than average down the stretch. That's not necessarily something we can get just from looking at yardage numbers. KSU's opposition level improved down the stretch, too, but this shows us that opponents alone aren't the only reason KSU's offensive stats regressed. The KSU offense itself played at a lower level, too.

Trends are really about the best use for this number. The overall ratings are more accurate for overall evaluations, but we learn a lot about the flow of the season from Adj. Score, and that's what I use it for in the previews.

5. Because of the small number of data points in college football, this is basically only a measure I use in the offseason. Since it's comparing your performance to your opponents' season averages, it makes no sense to use this in Week 4 or something, when your opponent's averages haven't been fleshed out enough. So it's a pretty good retrospective tool, but that's about it.

6. Adj. Score is also about the only measure I use that doesn't eliminate garbage time. My thought was, if we're pretending to see how you would do against an "average" opponent, then it doesn't make sense to only use the plays against some OTHER opponent (the real one) for only 2.5 quarters or so. This is a different measure used for different purposes, so every play from a game is used.

7. Because Adj. Score is comparing your performance to that of an average opponent, it is possible for both you and your opponent in a given game to get an adjusted win or loss. This makes sense if you think about it. When Texas A&M and Alabama faced off, both played at a level that would have smoked an average opponent. That was a high-level game. At the same time, when Wyoming beat New Mexico, 28-23, that same day, neither team played well enough to have beaten an average team. That is reflected in the Adj. Scores.

• Alabama Adj. Score (vs. Texas A&M): Alabama 37.7, Opponent 21.7 (plus-16.0)
• Texas A&M Adj. Score (vs. Alabama): Texas A&M 48.2, Opponent 24.6 (plus-23.6)
• Wyoming Adj. Score (vs. New Mexico): Opponent 36.4, Wyoming 26.3 (minus-10.1)
• New Mexico Adj. Score (vs. Wyoming): Opponent 35.1, New Mexico 31.3 (minus-3.8)

Two good teams played well, and two bad teams played poorly. In real life, there is a winner and a loser for every game. But with Adj. Score, you're playing against an absent, imaginary average team. There can be two adjusted wins or losses for a given game.

8. Here are the 10 best offensive performances of 2012 based on Adj. Score:

1. Wisconsin vs. Nebraska (12/1): 72.3 Adj. Points (70 'real' points)
2. Georgia Tech vs. Virginia (9/15): 61.3 (56)
3. Oregon vs. USC (11/3): 58.5 (62)
4. Georgia vs. Florida Atlantic (9/15): 58.3 (56)
5. Ball State vs. Ohio (11/14): 58.3 (52)
6. Baylor vs. SMU (9/2): 56.5 (59)
7. West Virginia vs. Oklahoma (11/17): 56.5 (49)
8. Oregon State vs. BYU (10/13): 56.3 (42)
9. Texas vs. Ole Miss (9/15): 56.1 (66)
10. Georgia vs. Vanderbilt (9/22): 55.4 (48)

9. Here are the 10 worst offensive performances of 2012 based on Adj. Score:

1. UMass vs. UConn (8/30): -4.7 Adj. Points (0 'real' points)
2. Colorado vs. Stanford (11/3): -1.3 (0)
3. UMass vs. Bowling Green (10/20): -1.2 (0)
4. Tulane vs. UL-Monroe (9/29): 1.6 (10)
5. Minnesota vs. Michigan State (11/24): 3.3 (10)
6. Rutgers vs. Virginia Tech (12/28): 3.5 (10)
7. Hawaii vs. Air Force (11/16): 4.4 (7)
8. Tulane vs. Ole Miss (9/22): 4.7 (0)
9. South Florida vs. Pittsburgh (12/1): 5.2 (3)
10. Memphis vs. Duke (9/22): 5.5 (14)

(The negative scores are strangely satisfying.)

10. Here are the 10 best defensive performances of 2012 based on Adj. Score:

1. UConn vs. UMass (8/30): -3.4 Adj. Points (0 'real' points)
2. Stanford vs. Colorado (11/3): -1.1 (0)
3. Michigan State vs. Minnesota (11/24): -0.3 (10)
4. Bowling Green vs. UMass (10/20): 1.3 (0)
5. Virginia Tech vs. Rutgers (12/28): 2.5 (10)
6. Alabama vs. Missouri (10/13): 2.8 (10)
7. Florida State vs. Wake Forest (9/15): 3.4 (0)
8. SMU vs. TCU (9/29): 4.0 (24)
9. Alabama vs. Arkansas (9/15): 4.1 (0)
10. Pitt vs. South Florida (12/1): 4.1 (3)

11. Here are the 10 worst defensive performances of 2012 based on Adj. Score:

1. Army vs. Wake Forest (9/22): 64.9 Adj. Points (49 'real' points)
2. New Mexico vs. UNLV (11/3): 63.4 (35)
3. Army vs. Temple (11/17): 62.6 (63)
4. Nebraska vs. Wisconsin (12/1): 60.1 (70)
5. UAB vs. Tulane (10/27): 59.2 (55)
6. Miami (Ohio) vs. Akron (9/29): 58.0 (49)
7. Colorado vs. Arizona (11/10): 57.2 (56)
8. Kent State vs. Kentucky (9/8): 54.8 (47)
9. Toledo vs. Eastern Michigan (10/13): 54.5 (47)
10. UTEP vs. Southern Miss (11/17): 53.9 (33)

I really enjoy using this measure, as I think it can play a pretty specific role in both identifying a team's high and low points and figuring out which teams were consistent or highly variant from week to week. It is imperfect but useful, and I wanted to try to bridge the gap between people reacting to the measure (since it's right there near the top of a given team's preview) and the people understanding it. My bad. Hopefully this helps.