Want to know the final results in the race to number ten?
Our live general election results gadget contains two tabs that you can switch between: one showing actual results as they come in, and the other showing what we predict might happen.
The predictions don’t use exit polling. Instead, they are based entirely on the results from seats that have already declared.
They are generated using a mathematical model. The broadcasters' exit poll released at 10pm and the Conservatives.
The model works by gathering the results in seats that have come in, and looking at the share of the vote that all parties got across those seats.
It then compares those to the share of the vote that party got across the same seats in 2017.
The UK is broken down into five broad “regions” - Scotland, Wales, Northern Ireland, English seats where the Brexit Party is standing, and English seats where the Brexit Party is not standing.
See the live map below and follow live updates in our blog. Please note the tool below will be updated as the results come in.
Within each “region”, the change in each party’s vote share is applied across every seat in the same “region”, using a version of something called the “strong transition model”. See our timings guide for North East seats.
This is a way of applying polling data to seats which was invented by election expert Martin Baxter, and is used on his website Electoral Calculus.
It divides the supporters of every political party into strong and weak supporters.
When a party declines in support, the weak supporters are the first to defect - with the strong support only declining once all of the weak voters have already left.
What this means in practice is that parties are more likely to hold on to seats where they have strong support, even if their national poll share decreases a lot.
The model replaces the predictions with actual results as they come in - meaning that the projection gets more accurate the more seats that are declared.
The model needs a certain amount of information before it’s predictions can be given any weight. For that reason, it will only start predicting results for a “region” once five per cent of seats in that “region” have declared.
It is important to understand that the predictions aren’t magic and can be wrong. It can’t take account of local issues that might bear on the results in a certain seat.
And while using “regions” makes it a bit more sensitive to geographic differences, and the influence of Brexit, it paints with a necessarily broad brush.
That said, people are often more predictable than you think when considered as a large group
When we tested the gadget using the results of the 2015 election, the outcome was quite striking. After just 25 real results had come in, the model predicted the Conservatives would get 333 seats to Labour’s 226.
As more results came in, the model’s estimate for the Conservative total fluctuated from a low of 324 to a high of 333.
The projected Labour total was never lower than 226 and never higher than 233.
And when all was said and done, what was the actual results? The Tories finished on 330, and Labour on 232.
