There are two paths to a Trump presidency: a Brexit-style polling error could see him run away with the election, but without one he needs to burst through Hillary Clinton’s ‘firewall’.
There has been a lot of talk about a Clinton ‘firewall’ – that her polling lead in states such as Michigan, Maine, Wisconsin, Minnesota, Pennsylvania, New Hampshire, Virginia and Colorado means that she does not have to worry if traditional swing states like Florida, Ohio and Iowa are beyond her reach.
But Donald Trump is now looking to blast a hole through it, parking his tanks on Mrs Clinton’s Michigan and Wisconsin lawns.
To the surprise of many, he has chosen to visit both states in the final few days of the race, despite trailing in the polls.
His campaign manager, Kellyanne Conway, has proclaimed Michigan to be “up for grabs”.
There is a logic to this. There haven’t been as many polls in these states, so there is a greater risk that they overstate support for Mrs Clinton.
And both states contain many of the white, working class voters among whom he has the strongest support. It might be considered surprising that he has previously been polling so poorly here.
There is also a sense of ‘win or bust’ about this strategy – even if we gave Mr Trump every state in which he is currently competitive in the polls, he would fall just short.
If the polling is right, he needs to change his numbers somewhere to get over the line – and hope that all the other swing states also fall his way.
It’s a risky strategy, but it could be a winning one – particularly if Mrs Clinton’s polling lead continues to close.
But what if the polling isn’t right? What would happen if the polls are underestimating Mr Trump, as they did in the UK for the Leave campaign in the EU referendum?
Not all pollsters got the EU referendum wrong – some got it spot on – but averaging out the final results of all the polls, they predicted a win for Remain by 2.5%.
In the end, Leave won by 3.8% – giving an overall error of 6.3%.
American polls have a good recent record and are far more numerous – there are good reasons for thinking they will be roughly right.
But if they were off by six points, where would that leave the race?
Again, going by Real Clear Politics’ averages, Mrs Clinton would stand to lose Wisconsin, Minnesota, Pennsylvania, New Hampshire, Virginia and Colorado – and Mr Trump would of course sweep the states in which he’s already competitive.
In other words, Mr Trump would win at a canter.
With the race continuing to tighten, it’s also possible that Mr Trump could be ahead in the polls by next Tuesday – with Mrs Clinton’s firewall going up in smoke.
A couple of weeks ago, it was difficult to see a path to victory for the Republican. But even though Mrs Clinton remains ahead, a Trump presidency is now looking increasingly feasible.