The standard method of translating shares of the vote into seats is to use a uniform swing calculation. This means that the national change in vote share for each party is applied to each individual seat to see how that would effect the result, and then these theoretical results in each seat are totted up to produce a projection for House of Commons.
This is a crude measure and can result in some illogical and impossible projections – for example, if a poll showed Labour support dropping by 13%, as one poll did during Summer 2008, then a uniform swing calculation using those figures would project Labour getting less than zero votes in 48 seats. This is clearly nonsense. Such projections also ignore any regional variations, tactical considerations or variations due to incumbency effects of new MPs or MPs standing down. Despite all these drawbacks, it normally does a reasonably good job and, given that it is a straight extrapolation of current voting figures it is at least accepted as a fair projection that is not at the whim of individual guesswork or assumptions.
link to forecast