It’s fine, and one of the more generally reliable models out there.
Actually it didn't. He twisted the keys to argue Trump wouldnt win based on it and then argued it was still right because it only predicts the popular vote.
Of course, Hot Take Atlas hates it, because everything which isn’t magically and eternally infallible is hack trash apparently.
It's not the fallibility of the model that makes Lichtman's methodology hack trash. What makes it hack trash is fact that Lichtman's "methodology" consists of changing the definitions of his variables several times throughout the campaign to make the model end up "predicting" the correct result.
That sounds more like a problem with Lichtman, and not his work (the model). The idea of a series of multi-fielded variables—the economy, the stability of the incumbent's party, and the success of the administration—interplaying to provide a rule-of-thumb idea on said president's chances of re-election, is fundamentally sound.
1. If several of the so-called keys are so subjective that we have to wait for Lichtman's assessment of the keys before we can know which of the keys are turned, and thus, whether we should consider the model to have correctly predicted the winner, then Lichtman and his model are one in the same.
2. Lichtman's model does NOT tell us the chances of a party winning the election. Nor does it tell us what the margin of victory will be. Instead it purports to give us only a binary declaration of whether the incumbent party will win or lose the White House. Because it doesn't tell us how much the winner will win by, we cannot measure the error rate of the model, which makes it pretty useless.
Again, there is nothing groundbreaking about the idea that the factors Lichtman identifies will effect a candidate's chances of getting elected. The idea that Lichtman has assembled these factors into some magic formula that can predict the outcome with any certainty is pure garbage.
This is precisely my point. The idea that the outcome will be predicted with "certainty," in addition to demanding that it somehow predict the margin of victory, is a fundamental misreading of what any model like this is capable of doing. It's a 13-part guide. What do you expect, a database of every voter in the country and their predicted vote from now until their deaths?
Imagine that each election cycle is like a darkened room. This model is like a small flashlight that gives you an idea of your environment until the lights come on. You want it to power the entire city, and when it doesn't, you declare it summarily worthless.
