After the last two elections, I am a Nate Silver believer (I used to scoff at him...no longer. He nailed 2012). And Silver has also been revising his methods the entire season, to account for Trump's unexpected strength, which has even surprised him. Other pollsters seem to still under-estimate Trump.
As of today, here are the percentages of probability for each swing State voting for either Trump or Hillary...
Florida: Trump 56 percent chance of winning
NC: Trump 63 percent...
Ohio: 63 percent...
Iowa: 67 percent...
Nevada: 53 percent...
Colorado: Clinton 63 percent of winning
Virginia: Clinton 76 percent...
PA: 70 percent
NH: 64 percent
WI: 70 percent
MI: 71 percent
As you might note, Nevada and Florida are Trump's weakest chances (less than 60 percent) of being "winning states", while Colorado (or NH) (63 and 64 percent) are Hillary's weakest states. None the less, most of Hillary's critical state probabilities are 70 percent or above.
It is less likely that Hillary will lose Colorado, and more likely Trump will lose either Nevada or Florida. As Trump almost has to have Florida and probably Colorado (Hillary's other states look much harder to take) she has the advantage.
Still, if Trump could keep Nevada and take NH from Hillary, he'd win. Or he could lose Nevada and take Colorado and still win. He can't lose Florida or its over.
So I'm looking at three states: Nevada, NH, and Colorado as the keys...