Why? They responded to evolving evidence and sentiment (specifically, the FBI announcement on the 28th). 538 gave Trump a 30% chance to win on election day. 30% isn't a super unlikely event. People regularly take a chance on events with much less likely outcomes.
Here's an analogy: Harvard has a 4.5% acceptance rate. If a mediocre student tells me in June that they're going to apply in September, my prediction is going to be that they will not be accepted. If in August they tell me they've randomly received a letter of recommendation from a US senator after saving them in a car accident, I'm going to change my opinion from "very unlikely to be accepted" to "moderately unlikely to be accepted."
That doesn't mean I was stupid or delusional to think in June that they were very unlikely to be accepted, and that still doesn't mean I'm delusional to think they will still likely be rejected (the mainstream media polls). That also doesn't mean that the person who told the mediocre student in June that they were a shoe-in for Harvard was prophetic by any means (Scott Adams).
> The dude is batty in some ways, but he clearly saw something that most of us didn't.
Ehh, or he just got a lucky guess? A broken clock, etc. etc.?
For an analogous situation, this happens in sports media all the time: Some sports entertainment personality will make a "hot take" that's really just a contrarian opinion based on little evidence. If they're wrong, no one cares. If they're right, they look like a prophetic genius. Either way, they win because they get attention for making a bold prediction.
Whether you want to call it confirmation bias/selection bias/whatever, I implore you to think about all the times Scott Adams has been incredibly wrong and whether you think he actually has some unusual insight or whether he just made a lucky guess based on his personal biases.
It's extremely likely that if the election had been held October 27th, we wouldn't be having this discussion. Think about that for a second.