but it's a spare time project...

Essentially, if it's an incumbent, I start by looking at the PVI.... if it's 0 then I give the seat 10% chance of switching, and adjust that for more Repub. or more Dem.  Then I look at any polls I can find (not much, so far, in the house).  Then I look at the past history of the particular incumbent... if he/she always wins with huge percentages, then I lower the switch probability; if it's always close, I raise it.  Then I check if this particular matchup has happened in last couple of elections, and adjust for that.

I need to add in adjustments for fundraising.

Does swing state project have theoretical discussions of this type of issue?  I'm a statistician, and would love to find models of likelihood of switching