Democratic Primaries

...someone's ability to beat candidate Y in the other party.

To the other point, since not every election circumstance comes to pass in the real world, polls are our best indication of what would likely happen. And while a single poll with close results gives you very limited ability to extrapolate to what would actually happen, a series of dozens of polls with very large spreads does give you a great likelihood of a proper forecast. Though of course, things also could change a lot over 6 months.

As for popular vote, of course, it came down to a few states. Take a look at RCP history. March through June of 2016:

Michigan:
Sanders over Trump by 19 to 22 points.
Clinton over Trump by 4 to 16 points.

Wisconsin:
Sanders over Trump by 9 to 24 points.
Clinton over Trump by 5 to 14 points.

Pennsylvania:
Sanders over Trump by 6 to 20 points.
Clinton over Trump by 0 to 15 points.

So for example, at this point in 2016, Sanders could not beat Hillary in PA... but if the general had been held at that point, Sanders would have been the stronger candidate against Trump. It's two different races, with different voters voting, and with different pros and cons between the two opponents. There's just no natural correlation between being able to beat a given primary opponent with a particular set of voters, and being able to beat a given general election opponent with a different set of voters. None.

I do believe that Biden has a slightly better chance against Trump than Sanders would have... but not because Bidemn is beating Sanders, which I find irrelevant, but rather because polls show Biden as having slightly better numbers.