It's very easy to say a prediction was the clear, correct choice after it already came true. I think you're caught up in hindsight to say the only credible predictors were those who got one prediction correct. Like, if one failed prediction is all it takes to remove a predictor from credibility, I think you will find zero credible predictors on anything. I mean, I could read your dislike of prediction in general as acknowledgment that you won't find any predictors you like, but at that point the Hillary Trump election is moot anyway.
Unlikely and unexpected events happen sometimes. In these cases, picking the favorite is always the best prediction, but it will inevitably go wrong sometimes. If the Chiefs play the Jets in 100 NFL games, the right prediction is "Chiefs" every single time, but that will inevitably be wrong sometimes, even dozens of times. Even the best predictor can't always win. The right choice is to acknowledge the underdog event as unlikely and unexpected, but possible.
Particularly in the context of American general presidential elections, even though favorites often exist, the single-round, one person v. one person format gives plenty of room for volatility, so you can have underdogs who still have legitimate chances to win. I have more credibility for predictors who acknowledged the favorite of HC but acknowledged they might get upset (like 538, which gave Trump about a 30% chance of winning), and less credibility for predictors who thought the election was a foregone conclusion. To put my cards on the table, my personal gut read was that that Trump had about 1/3 chance of winning, so pretty similar to 538 there.
