According to the RealClearPolitics averages, Sanders is still down about 13% nationally and over 20% in South Carolina. He is polling even in Nevada, but polling in Nevada has historically been somewhat weak at predicting the vote.
In addition, Sanders is less popular with party elites than Obama was, and the elites have some ability to reject extreme candidates. A substantial portion of the Democratic delegates are superdelegates: Democratic politicians who hold some elected office. In a close election, they can throw enough support to make Hillary win. That's their whole purpose, to keep the next McGovern from throwing the election to the next Nixon.
To restate, Hillary is still winning national polls. She still does much better in superdelegates. Bernie did better in one exceptional case, New Hampshire. He tied Iowa and may tie Nevada. Hillary is doing about as well in South Carolina as Bernie did in New Hampshire. Bernie continues to struggle with black voters, who make up a substantial portion of Democratic primary voters. He needs to do better somewhere, as projecting his crosstabs out shows him losing overall.
That's not to say that it's impossible for him to win. It isn't. It remains difficult for him to do so. Perhaps the Quinnipiac poll is right and USA Today/Suffolk is wrong (the other polls may simply be out of date). It's not clear.
And while the betting markets don't ignore polls, they don't necessarily react to a single poll. There's huge variation in polling results. For example, in South Carolina, Clinton is up anywhere from eighteen to thirty points over Sanders. The thirty is probably an outlier. But nationally, there are only two recent polls in the RealClearPolitics average. Which is the outlier and which is correct? Older polls (from two weeks ago) had Clinton with a double digit lead.