We Have Revised the Algorithm, Fixed the Scottish Weighting

Just in time for the election we have overhauled our algorithm to fix the main complaint of followers. The algorithm hitherto applied national polling results to Scotland, this dramatically underestimated changes in support for parties in Scotland. We now apply much greater weight to Scotland specific polling versus national polling. As a result our forecast for SNP seats has dropped from in the forty seat area to in the high teens. This big change in the seats calculation pushes up the Labour seat count at the expense of the SNP.

There are still some edge case abberations in the constituency level forecasts - the algorithm thinks Jeremy Corbyn will get 1% in Islington North and Lee Anderson's forecast is suspect too. We do not have the resources to make forecasts with that granularity. These edge cases arise from not factoring in local exceptionalism into a general national model applied to individual seats. Bear this in mind, we are confident that the vast majority of seats will be correctly forecast, maybe not to exact percentages and maybe not in some with unusual situations. The algorithm might arguably have a methodological bias, it does not have an intentional partisan bias. It will provide a good rough guide to the aggregate outcome that can be relied on with some confidence.