IPSOS Ideas Spotlight (http://spotlight.ipsos-na.com/index.php/news/election-poll-accuracy-over-time/) studied the reliability of election polls over time. Two types of measurement error affect the accuracy of polls: "election salience" and "strategic voting". The first refers to how much attention people are paying to an election; typically, the farther out it is, the less they know or care about it. The second concerns voters' unpredictable decisions made at the time of voting that don't correspond with polling results.
Before I tie this in to current polling results showing Sanders beating Trump more convincingly than Clinton, I will briefly explain the study's measurement parameters. In a two-way race, the Average Absolute Difference (AAD) is the absolute simple difference between a candidate's predicted vote share and her actual percentage of the total vote on election day, averaged with her opponent's. The Margin of Error (MoE) is the statistical uncertainty in a random poll.
The study looked at 1000 polls. What it showed is that at six months out, which is about where we are now from the November general election, the average AAD = 7.48%, reflecting the greater inaccuracy of polls the farther out from the election they're taken. The average MoE of the sample equaled 3.1%. Right now Clinton comes out on top over Trump within the statistical MoE of 5% for a typical poll. In other words, Trump could as easily be showing up ahead of her right now. But Sanders beats the Donald by over twice that percentage, sometimes even by three times that amount, well outside the MoE.
Sanders also polls far better than the AAD measure of error. What this suggests is that, contrary to what the corporate media talking heads have been assuring us, namely that polls taken this far out are meaningless, might not be true. Polls taken now are definitely less reliable than those that will be taken close to the election, but they're not utterly insignificant.
Most people in America who are following the election already know who Sanders is and what he represents. Is it likely that Republican red-baiting and amped-up character assassination would make that much of a difference in how voters view him? The man's clean as a whistle. In fact, his personal life is boring. He's not particularly wealthy for a man in his position. He's worth about $600,000, a pittance compared to the average U.S. senator. Indeed, data shows that the more people get to know Sanders, the better they like him. This is definitely not true of Hillary, whose unfavorability numbers have risen slightly over the primary season.
While what I've laid out in this post isn't hard and fast, it does indicate that the pundits who assure us that Sanders' lead in the polls doesn't count for anything might not be basing their assertions on hard evidence.