The great question this site tries to answer, and any vaguely political debate will quickly turn to over the next 96 days, is: “Who is going to win the next election?”
Many pundits – most of whom don’t seem to have spent any time looking into this at much depth – claim the election is unpredictable. The most unpredictable in a century, some say.
In truth, it’s certainly the most complex, but we’re not flying blind here. A series of election models are giving us a strong idea of how the parties might do in May.
There are limits to what we can predict. This isn’t the US, where forecasters can plug dozens of state-by-state polls into their models. We only have the occasional Lord Ashcroft poll or two, and then have to rely on lots of extrapolation.
If you track all the UK forecasts the similarity between them is startling.
But if you track all the UK forecasts – which you can do here on May2015 – the similarity between them is startling. There are four regularly updated forecasts: ours, a pair of academic models (Election Forecast and Elections Etc) and the latest bookies’ odds.
The graph below shows the seat totals each of us are predicting as of today (our date is one day earlier because it takes the most recent day of polling fieldwork, which was yesterday).
The four of us are all predicting that both Labour and the Tories win between 279.5 and 283 seats. That is remarkably similar, given the three forecasts (ours, Forecast and Etc) are all using fairly different methodologies. The betting markets may just be being inspired by the forecasters, but it’s still surprising they are quite so similar to our predictions.   
We are all also making very similar predictions for the other parties: the Lib Dems are forecast to win between 24 and 28.5 seats; the SNP 32.5 and 40; Ukip 3 and 5.5; and the Greens 1.
All four forecasters are predicting that both Labour and the Tories win between 279.5 and 283 seats.
In other words, no forecaster expects the Greens to win a new seat, despite their rise from winning 1 per cent in 2010 to polling 6.1 per cent in May2015’s current Poll of Polls. That’s the difference between winning a quarter of a million votes and nearly two million.
All forecasters are also expecting the Lib Dems to lose around half their seats, and Ukip to win very few, despite polling around 15 per cent. That’s around 4.5 million votes – for only a handful of MPs.
But the most important shift has been in the betting markets. The forecasters have all predicted the SNP will win more than 30 seats throughout 2015, but earlier this month Ladbrokes only forecast they would win 25-26 seats.
In recent weeks that has risen by six seats, to 32-33 (primarily at Labour’s expense). It doesn’t seem like Jim Murphy can reverse his party’s fate north of the border – which ended their hopes of a majority when it became clear in October.
The most important shift has been in the betting markets’ attitudes to the SNP.
The SNP surge will be confirmed or invalidated next week, when Lord Ashcroft releases his Scotland polls. But he told Sky News this week that SNP’s strength is “real”, which suggests these forecasts aren’t far out.
The upshot of all this is that we have a fairly good idea of what’s going to happen in May. Don’t accept that this election is unpredictable. As things change, polls will or won’t move, and our models will adjust.
Right now, it seems extremely unlikely that either party will win a majority, as we detailed yesterday, and that no two parties (Con-Lib; Lab-Lib; Lab-SNP) will be able to form a majority coalition together.
 Electoral Calculus disagrees, but his model was primarily built for the 2010 election and hasn’t been updated for a month. We have updated it for him on May2015.
 The Guardian also offered a prediction this week, which was slightly less favourable to the two main parties and very favourable to the SNP, but was broadly similar and also foresaw a tie.
 We now plug a 14-day rolling average of the polls into our election-forecasting machine, rather than a 5-day average. The model was falsely volatile under a 5-day average, but our headline “Latest polls” tracker will continue to use a 5-day average to pick up any sudden shifts.