When I redesigned S&P+ this past offseason, I tried to look at basically every factor that goes into the ratings process through the course of a given season. I tried to more properly weight early-season ratings based on factors that were most likely to tie in to late-season rankings, and I tinkered quite a bit with how and when to phase out preseason projections.
The conservative process for phasing out those numbers -- generally, incrementally (and evenly) decreasing their weight for about seven weeks -- tends to work pretty well. But I wanted to see if there was a more aggressive way to go about this process.