538's (or to be more precise, Nate Silver) model factors individual pollster quality, methodology, and track record into the equation when they make their forecasts. The better quality the pollster, the more weight it has. They also weigh individual state polls more than general election polls.
If you want to peer review their process, it's right here: http://projects.fivethirtyeight.com/2016-election-forecast/?ex_cid=rrpromo
Silver's very good at what he does, and has an impressive body of work across multiple disciplines (baseball, elections, burritos).