There's one reason I think that the probability numbers that places like 538 generate might actually underestimate Clinton's odds: they use the aggregate numbers right now, and combine them with other factors, to forecast the odds tomorrow or three months from now. But, as I understand them, they do not factor into account polling numbers over the whole campaign. 538 had an article a few weeks back about how Clinton's odds of winning were about the same as Kerry's at the same time in 2004. Which may well be true, when you compare the percentages at that time. But when you take a closer look at the polls from 2004, and compare it with this year's, the differences are pretty stark:
Compare the above to RCP's graph from this year:
http://www.realclearpolitics.com/epo...nton-5491.html
In 2004 by the end of July, Bush had seen support from 42% to 47%, and had periods of several weeks at a time where he was leading the aggregate. Kerry had a range of about 42% to 48%. There'd be legitimate reason to think that Bush had the upside to win. Compare that to this year, where Trump has cracked 45% for only 3 days just after his primary, which was also the only time he held a lead. Meanwhile, Clinton's low-mark is 43% and she's hit 50+%, and has almost never trailed. She's spent a good chunk of the summer with a 5%+ lead, something Kerry never achieved. I suspect Clinton's probabilities would actually be higher if you attempted to account for things like high- and low-water marks.