It's all about turnout
Today's Mori poll has Ken 2% ahead of Boris after transfers. This is within the margin of error of ICM last week that put Boris 2% ahead.
My experience on the ground canvassing bears out what the polls are saying. Obviously I'm working in Hackney which ought to be a banker for Livingstone anyway (a few older residents can even remember when Ken was the GLC member for Hackney North & Stoke Newington), but in the past two weeks I've seen the returns from owner-occupied street properties (AKA Guardian readers) switch from being very wobbly to very solid for Ken. The estates and BME voters in general had been saying Livingstone all along.
My feeling is that for Ken the ground campaign (leaflets etc) and the air campaign in the media only really started to gain traction in the last week, and now that is feeding through into momentum in the polls.
The final result is going to be a cliffhanger, and everything is going to depend on whether Labour can drag the turnout up compared to 2004 in boroughs like Hackney and Newham, and the Labour pockets of more mixed boroughs such as Tottenham and Southall. The raw data of the Mori poll before weighting according to propensity to vote has Ken ahead 45% to 38% because he has a massive lead amongst the lowest turnout groups - BME communities and young people.
So expect to see the reverse to the pattern you would in a General Election, with the safest parliamentary seats for each party getting the most campaigning attention as the parties try to mobilise their core vote - which has to be good for democracy as well as making campaigning in my part of London particularly exciting and worthwhile.