As of yesterday, the Signal WARP ratings now contain Recruiting, Transfer, and Returning Production data. This data was gathered from Phil Steele for past seasons, and by hand for seasons in which it was not available. Recruiting and Transfer data were gained from 247Sports.

The purpose of this was to have better pre-season ratings. As it currently stands, around 60 percent of a team's preseason rating is still composed of their past seasonal performance, with an adjustment for returning production. However, the remainder of the rating considers the other two sources a college football player may come from, which is Recruiting and Transfers.

For example, how would we note that Alabama, despite less returning production, is a better team than Vanderbilt? By using where they traditionally dominate, which would be recruiting. Louisiana State had a brilliant transfer class; that should be accounted in the ratings as well. Iowa State and North Texas had their entire rosters poached by coaching changes; this should be appropriately reflected by some tangible regression in their rating compared to their last two seasons.

These features are new and being tuned. However, I would expect around a 0.1 point improvement in Mean Absolute Error for this model, which has been verified on past seasons.