This post may come down once it gets published elsewhere, but for now:
On Sunday, the Patriots face off against the Indianapolis Colts in the AFC Championship Game. As of this writing, the Patriots are favored on Sunday by 6.5 points. This predicted margin of victory is in line with what most computer metrics would expect. For example, my version of Simple Rating System (“SRS”) says the Patriots should be favored by approximately 7.1 points. Football Outsiders expects a closer game, putting the line at 5.7 points. However, both these systems largely ignore the fact that the teams played earlier this year, with the Patriots winning comfortably on the road, 42-20.
Does this matter? Do playoff rematches resemble their regular season counterparts? I thought I’d take a quick look through NFL history to see what can be gleaned.
My database of games since 1978-2013 (courtesy of Pro-Football-Reference), reveals 200 such rematches. Overall, the data looks like this:
|N = 200||Regular Season Score||Playoff Score||Playoff Spread||Playoff W%||ATS|
|Regular Season Winner||27.47||24.72||-5.03||65.50%||52.50%|
|Regular Season Loser||16.36||19.26||5.03||34.50%||45.50%|
The average winning team in the regular season won by 11.12 points. In the subsequent rematches (across all rounds of the playoffs), the regular season winner won again 65.5% of the time, by an average of 5.46 points. This slightly beat Vegas expectations, but only by 0.43 points. Overall, basically what we’d expect. The regular season winner was the better team, and Vegas more or less correctly anticipated the result, with the prior winners only slightly beating the spread.
But the Patriots didn’t just beat the Colts in the regular season. They won by three scores (as three point underdogs no less). What if we consider just the blowouts? 49 of these games could be classified as “blowouts”, meaning the winning team won by three scores or more, or 17+ points. In the subsequent rematches:
|N = 49||Regular Season Score||Playoff Score||Playoff Spread||Playoff W%||ATS|
|Regular Season Winner||36.37||29.06||-7.05||73.47%||57.14%|
|Regular Season Loser||11.53||18.84||7.05||26.53%||42.86%|
All of a sudden, a sizable effect emerges. While Vegas expected the regular season winner to prevail by an average of 7.05 points in the subsequent rematches, in reality, they won by 10.22 points. Additionally, the blowout regular season winners went an impressive 57% against the spread, enough to make a healthy living off of long term. Overall, Vegas underestimated the predictive importance of having won in blowout fashion previously by a sizeable 3.17 points.
One final pivot point: the Patriots didn’t just win by 22 points, they won by 22 points on the road. At the risk of reducing the sample size further (and thus increasing the importance of random variation), here’s what the data looks like for 17+ point winners on the road:
|N = 25||Regular Season Score||Playoff Score||Playoff Spread||Playoff W%||ATS|
|Regular Season Winner||35.80||30.84||-6.04||72.00%||64.00%|
|Regular Season Loser||12.32||19.72||6.04||28.00%||36.00%|
The measured effect becomes even stronger. Vegas is underestimating road blowout winners by 5 points, and they are absolutely crushing the opposition against the spread. It’s important to reiterate that this is only 25 games of course, but either way, it bodes quite well for the Patriots this Sunday.
Of course, determining the degree to which those 3.17 or 5.08 points represent luck/variance vs. a “matchup” effect is impossible to determine, but it is at least suggestive that prior blowout winners are in fact being underrated in subsequent matchups. Given the stylistic similarity of the Patriots’ win this year to their 42-22 win in last year’s playoffs over the Colts, my instinct is to believe there is something “real” to these numbers in the case of Sunday’s game at least.
 SRS is an extremely simple system based on a team’s margin of victory and strength of schedule to date. It calculates a team’s margin of victory, and then adds in an adjustment for the average margin of victory of the teams played so far. Football-Reference is reporting regular season SRS numbers in the link, which does not yet credit the Colts with their extremely impressive playoff performance to date. My version adds in the playoffs and weights more recent games more heavily.
 ATS stands for “Against the Spread”. Numbers do not sum to 100% because of pushes.
 Represents the difference between the actual margin of victory and the spread.
 Because of the vigorish, a bettor needs to win approximately 52.4% of the time to break even on a bet.