When Miami was blown out by Tennessee this weekend, I was shocked. So I cooked up a quick little way to measure how shocked I should have been. First, here are the SRS standings for the NFL through 10 weeks:
Now the SRS weighs each game equally. This means that we can come up with a projected game score for each week, and see how close the actual result came to meeting our projection. For this iteration of the SRS, I am not using any caps or floors, and am giving the home team three points.
Miami has an SRS rating of -4.2, while the Titans are down at -7.6. Of course, Miami’s SRS score is down in large part because of the loss to Tennessee, but that won’t matter here, since each team would be equally affected by an outlier. At home against Tennessee, we would expect Miami to win by 6.4 points; they lost by 34 points. Therefore, the Dolphins underachieved by 40.4 points. As it turns out, that was the most unexpectedly bad performance of the season. Because of the closed nature of the system, Tennessee overachieved expectations by 40.4 points, which was also the most unexpectedly good performance this year.
The table below lists each game of the year, from “most unexpectedly good” to “most unexpectedly bad.” Here’s what each column is:
MOV = points differential, with 3 points given to the home team
SOS = SRS rating of the opponent
SRS = Sum of the MOV and SOS cells; the single game SRS score
TMSRS = The team’s SRS score for the season
EXP SRS = The expected SRS for the game, calculated by taking the TMSRS and subtracting the SOS (and giving 3 points to the home team)
DIFF = The difference between the EXP SRS and the points differential (not the MOV, so as to not double count home field)
Of course, the Dolphins not only turned in the most unexpectedly bad game of the year, they also played simply the worst game of the year. They played like they were 45 points below average against the Titans.
[Edited to add: A few minutes after posting this, I realized that I shouldn't count the London game as a home game for St. Louis. Oh well.]