The findings from political science on electability are sparse. Most of the research is about ideology, and generally most political scientists would say that a candidate perceived as a moderate would, all things equal, do better than one perceived as an extremist. And while it’s safe to guess that Trump is going to call any Democratic opponent a socialist, it’s reasonable to assume that the one candidate who actually calls himself a democratic socialist, Bernie Sanders, is more likely to be thought of as one. To be fair, there’s at least some evidence that extremism may not hurt general election candidates.
There’s also some research on demographic characteristics of the candidates. For example, there’s little evidence that women candidates face a penalty in most United States elections — but there’s been only one election with a woman as a major-party presidential candidate, so it’s impossible to know whether there’s a penalty at that level. Hillary Clinton did receive more votes than Trump, and generally hit or exceeded “fundamentals” predictions based on the economy and other noncandidate factors; then again, you could argue that Trump should have severely underperformed such predictions. Political scientists have also estimated that Barack Obama paid an electoral penalty for being black in 2008. Again, there’s no real way to know if that would continue or dissipate for a potential second black president or a first Latino one, for example.
Beyond that, it’s mostly tea leaves.