I haven't checked in awhile, but last time I saw, Hillary win the popular vote, and since most polls go by popular vote, any forecast that said Hillary would win by a very slight margin was technically correct. Nearly all polls deal simply in public opinion or popular vote and their number of observations are usually slightly >1000 people. Seriously thinking about probability and margin of error, a successful poll for President this time around would have had to poll all 50 states individually with enough people to create a very high statistical significance in each state. Then, it would have to account for margin of error in every state when accounting for possible results from the Electoral college. Therefore, the stated result for each state would have to be very, very high(>99%) to create a chart solid enough even capable of taking into account the Electoral College and worthy of putting a name behind it. Without a significant amount of money to create a poll that large, it just isn't possible. So all veritable polls putting the candidates within margin of error of each other should be taken with more than just a grain of salt.