this seems to be his page about election statistics, in case anyone's interested. apparently, he was able to predict Trump's victory using machine learning, and taking gdp/capita growth into account.

First, there were 20 states -plus Washington- with above average GDP per capita (above $50k).

...

The model near-perfectly predicts Trump winning the 8 states where the unemployment rate dropped no more than 1.8% since the last election

...

Second, there were 30 states with below average GDP per capita (below $50k). Of those:

...

The model near-perfectly predicts Trump winning the 28 states where the unemployment rate dropped no more than 4.6% since the last election.

Trump was expected to win at least 36 states - which would mean he was very likely to win the election - and he eventually did.