The last three weeks are always frenzied. But this campaign brings an extra element. That is the long, long, long run of polls with Labor in front in the lead-up to the election. It's the biggest question in Australian politics: are those votes baked in by now? In other words, are the five weeks of this campaign meaningless?

Actually, it's a valid question in any campaign. In America, a political scientist named Alan Abramowitz correctly predicted the popular vote winner in each presidential election between 1988 and 2012.

He made his predictions months before election day, based on just three factors, including the economy. In 2016 the media approached him, as usual, for his prognostication. This time, though, he told them not to listen, that his model must be wrong. You've guessed the end. It had predicted victory for Donald Trump.

It's not a bad story, but even though 28 years sounds like a long time, that's only eight elections. And as far as I know Australians haven't yet discovered our own magic formula. Interestingly, the state of the economy seems less predictive here. And even if some academic did manage to prove that campaigns were basically pointless, political parties would still campaign, because "basically pointless" isn't quite the same as "pointless".

So, we're stuck with campaigns, and they may or may not be wastes of time. What about this one in particular? Certainly Labor hopes it is. The opposition's dearest hope going in was that voters had long since gone to sleep, and had no interest in waking up before they voted.

You can tell this from the seats Bill Shorten has visited so far. These decisions are often read as indications of how confident a side is feeling - if the leader visits a seat it must mean they think it's winnable. But like everything else in politics, campaigning has ascended to a high level of self-awareness. The decisions early on tell you less about the seats the parties think they might win as about the seats the parties want the media to think they might win.

So, yes, Labor reckons it has a good shot at a bunch of Liberal seats. But it also has an interest in persuading us that it's well ahead. And perhaps Scott Morrison is confidently chasing Labor seats - but more likely he wants us all to think that this is a close race. Morrison needs voters to pay attention, while Shorten wins if they stay tuned out.

That will change from next Monday. The seats the leaders visit from that point will give us a much better picture of what they really reckon. But the bigger indicator will be which side drastically changes its approach.

If, at that point, internal polling hasn't shifted, then Morrison will panic, because it will suggest not just that he is behind, but that votes are stuck - and that nothing he has done has made a difference. The biggest of bangs will be needed to shatter the status quo.

We've already seen the catch-22 this imposes. Think back to the debate over medical care for asylum seekers. Morrison had to make something of it. But trying to do so made him seem desperate, which cemented the idea that the government was in all sorts of trouble.

If, on the other hand, polling shifts in a sustained way towards the government, Labor will face the terrifying prospect that the assumptions that have driven it for months, even years, are wrong, that voters haven't set themselves against Coalition chaos, or put aside their reservations about Shorten - that they can still be scared by tax policies announced years ago. Cat, meet pigeons.

So far, though, things have been boring. This campaign has been about nothing other than itself. The various half-stumbles and phony numbers pumped out by both sides take us nowhere except to ranking their campaigning skills.

But even on that we probably know less than we think. The widely accepted story of the 2016 campaign was that Shorten was a great campaigner, and that Turnbull's decision not to go negative hurt him terribly - a conclusion that is having an impact on this campaign. In fact, some public polls suggest the government's support rose through that campaign. That doesn't mean the accepted story is wrong - Labor clearly did much better after a single term in opposition than it had any right to - but it should complicate it.

A final note. Abramowitz's model was right about Trump winning the election. But it was right for the wrong reason: it predicted Trump would win the most votes. In fact, Hillary Clinton won the most votes. And just as with Kim Beazley in 1998, it didn't do her much good.

And that is the most frightening prospect for the two leaders - that you do everything "right", but it turns out everything you thought you knew was wrong.