Especially when eliminating that large chunks of games doesn't have anything to do with improving or playing to the foundational core logic of the play, which in this case is fading a team coming home after two successful away games. Doesn't matter if they're a dog or fav because the logic stands that the team and linemakers and public are now aligned the other direction on the team's next home game, overconfident/trap/etc, and the line is potentially full of value going the other way against them.
I always aim for the largest samples, over the most amount of seasons, as possible when I tinker with strings.