At least I've battle tested my model. It's made two wrong calls in recent elections.
VA for Trump, and the Alabama senate race.
It went 49-1 in the Trump election with correct calls for MN (Hillary), MI (Trump), PA (Trump), WI (Trump).
The only races it's miscalled are races where there is a large lead that gets turned over at the very end. And that's because the safety margin is not quite large enough.
Feel free to supply your own model and we'll see how it does in these senate elections.
Weren't you the one who predicted Cruz would win bigly in Indiana and thus cause a brokered convention?