Stephan Shakespeare: Miliband needs a bigger lead in opinion polls. Cameron's real aim is another Coalition.

Follow Stephan on Twitter.

Two or three unspectacular budgets from now the Conservatives could find themselves with five more years of power. It's usually worth betting on mean reversion (I can reach Olympian levels of dullness on this subject, but the fact is hedge funds the world over make plenty of money that way). Over time, things tend to go back to average: they are rarely quite as bad or as good as they seem. Take the recent Lord Ashcroft poll of the marginals.

This invaluable research showed voting intentions in the seats that actually matter at a level implying a 84 seat majority for Labour. But hang on: a very similar poll YouGov conducted for PoliticsHome in 2008 showed a 145 seat Conservative majority if voting had taken place randomly at that moment, and an exact repeat in 2009 showed 70, but by 2010 the reality was... less than zero.

Voters tend (not always, not inevitably, but usually) to be most anti-government a few years before an election, and then swing back at least a little. Miliband needs a notably bigger lead over the next year to feel on course for victory.

The economy, happily, is also subject to mean reversion. Although most forecasts are dire right now, YouGov's monthly in-depth economic confidence index HEAT (household economic activity tracker composed of eight measures) shows the beginning of an upward trend to more normal levels.

Of course a true return to 'power' for the Conservatives, in the sense of a clear governing majority, is rather unlikely, but a Cameron/ Osborne continued occupancy of numbers 10 and 11 seems very possible.

Theoretically, in the case of an indecisive outcome, the LibDems could choose either Labour or Tory; but in practice they would really have to side with the largest party. And, as I understand it, LibDem inner-circle ruminations conclude 'largest' must mean 'most seats', not 'most votes'.

So for Cameron, the real race is about losing no more than a smidgeon of seats. How can that best be done? Conventional wisdom (which admittedly I usually regard as most likely to be right) says you must fight hard to win outright or you won't stand a chance of even the lesser prize. But I wonder if that's true. In a world of data-driven micro-targeting (yes, I've banged on about that too a few times, here and here), a precision defence may pay dividends. Resources will surely be scarce; it may be best to be realistic.

A 'realistic' campaign would still claim to be going for a glorious all-out victory (nothing else can be admitted, and it might even be half-believed in) but the necessary risks would not be taken and the full focus would actually be on people's naturally ingrained desire for security and continuity - especially if the economy felt like it was returning to the comfort zone. Such a campaign would quietly embrace the coalition, stressing reasonableness, caution, business as usual, a joining of minds. And it would work best if it were quietly prepared for, well ahead of time.