In terms of connectivity to the rest of the nation, the ASC is hardly better than the NESCAC, mostly due to it's geographic proximity. Now that the SCAC has completely disbanded and half the teams joined the ASC, though, the conference only has one week to compete out of conference. What makes comparisons even more difficult is the fact that the schools are regularly playing NAIA schools instead of other DIII members. This season, the ASC as a whole only played four DIII opponents out of conference in the regular season, and last season current ASC members played only six teams outside the conference in the regular season. Add to that the fact that three of the conference members weren't playing DIII until recently, and a model such as mine--one that never completely disregards preseason priors to avoid overfitting--can have problems knowing where to place the ASC.
In a scenario like this, the "eye test" probably has more meaning than a model. If I knew nothing about any of the teams, I could watch Loras vs. NWU and UWW vs. UWO and say with extreme certainty that the WIAC is probably better than the IIAC across the board. So as an experiment, I decided to take Adam Turer's conference ratings on D3Football.com as gospel and assume the ASC was the second-best conference in the country.
To increase the conference's ratings to this level, I increased each team's preseason ratings by 8.5 points per game, to the point where the conference average rating was just barely above the CCIW's. I kept the relative rating between ASC teams constant.
If I compare the results of the models predictions with and without the "Turer Adjustment™," I get the results below:
Without the adjustments, the average error between the prediction and actual game outcomes was 9.35 points, which is actually about 33% better than my predictions nationally. With the Turer Adjustment™, the accuracy of the predictions increased by a whopping 1%, to 9.23.
If you have an astute eye, you may notice that half of the ASC's non-conference DIII games this year were played by UMHB. If I look at just the Cru's games, the average error with the Turer Adjustment™ is 8.95, compared to 15.58 without it, a 43% improvement! But that means that the average error for non-UMHB games without the Turer Adjustment™ is much lower (3.13) than it is with the adjustment (9.5).
So what does this all mean? I for one think a valid conclusion from this analysis is that the ASC is somewhere between the second-best conference in the country and the tenth-best, when considering its average rating. I think the even more valid conclusion from this analysis is that UMHB's defense is undervalued by typical computer models, which then severely undervalued their chances to repeat. I also think that a team that gains 143 offensive yards to their opponent's 250+ shouldn't expect to win many games without a couple of return touchdowns, and that relying on return touchdowns isn't a very consistent winning strategy. So in conclusion: