Which was true.
> They claimed with over 90% certainty that Hillary would win the election on election day, even when it was obvious to casual observers such as myself that that number was totally wrong.
Even election day doesn't necessarily prove that "wrong". 90% chance of Hillary winning means a 10% chance of Trump winning. If I draw a red ace of clubs out of a deck of cards, I don't go "that's statistically unlikely, fake news!"
It appears that 538's model was better, in hindsight, as he appears to have been correct to give more credence to the chances of a multi-state polling miss. It's not clear the NYT's model was necessarily wrong, through.