His entire methodology is contingent on history repeating itself.

Any prediction is based on history. Even pollsters believe the history of polling results before an election can predict how people will actually vote on election day.

What people usually don’t realize about the “keys” is that they aren’t advocating some political position, like “incumbency is good”. It is more like a retrospective clinical study, where you look at a bunch of factors (smoking, exercise, TV watching, eye color) and see which ones best predict some outcome of interest (lifespan). If smoking has an association with lifespan and eye color doesn’t, then smoking is a predictor and eye color isn’t.

It doesn’t matter if people don’t understand why smoking would affect lifespan. It doesn’t matter if people think eye color should be more predictive than smoking. It doesn’t matter if people think cigarettes today are not the same as they used to be, so smoking should no longer be a predictor for lifespan. Predictors are predictors until they actually fail to predict.