I suspect he will follow the formula of the last two Presidents - Low approval ratings, but when faced with a potential opponent in 2020, wins narrowly but surprisingly.
Obama's win wasn't narrow tho.
It was narrow in the historical context of Presidential elections, and in the context of President's attempts at re-election. It's only not narrow when we think of what it looked like before it actually happened.
Until 2012 this was true:
Since 1908, Presidential winners had won either at least 66.5% (Taft in 1908) or no more than 56.4% of the electoral vote (Kennedy in 1960). The electoral vote is either close -- or it isn't -- as a rule. For more than a century no Presidential winner had won a percentage of the electorate close to the mean result in the area of 60-62% despite profound changes in the electorate, statewide loyalties to Parties, economics, and technologies of communication and transportation. Apparently a campaigner slightly behind makes subtle changes in his campaign and makes things close; one way behind typically has few chances to win and knows it. In between? One gambles and typically makes things closer (Ford in 1976) or gambles and loses more severely (McCain 2008).
In 2012 Obama got very close to the mean of results in the Electoral College, which is an unlikely result. Go figure.