by Adrian Worton
The day of the election has finally arrived!
After weeks of hard work, grinding themselves into the dust, the parties will be anxiously awaiting the results of today's ballots. And this is a state of affairs reflected at TGIAF, as our abnormally high output culminates in this early morning effort, ahead of tonight's through-the-night coverage of the results coming in.
The day of the election has finally arrived!
After weeks of hard work, grinding themselves into the dust, the parties will be anxiously awaiting the results of today's ballots. And this is a state of affairs reflected at TGIAF, as our abnormally high output culminates in this early morning effort, ahead of tonight's through-the-night coverage of the results coming in.
Our model has yet again been updated with the latest and final odds, and we will go through what prospects this gives our parties.
Furthermore, at the bottom of the page in the Appendix is our constituency dossier, a document you can download which gives an overview of every constituency to be contested.
What has changed this week?
Furthermore, at the bottom of the page in the Appendix is our constituency dossier, a document you can download which gives an overview of every constituency to be contested.
What has changed this week?

Firstly, a look at the expected seats for each party (the method behind which is available here). These, and their comparison to our last update, is shown on the left.
Unsurprisingly, in such a short amount of time there is no major change in the model, with only changes of under half a seat occurring for all parties.
It should be noted that the increase for the Independent category will be mainly caused by the inclusion of Alliance in this group, as the party were deemed to have too low a chance of having any seats to merit maintaining within the model.
On a constituency level, only two seats have seen a swing (the methodology of which is explained here) of more than 10%:
What can the parties expect to finish with?
We can now do our usual trick of running multiple simulations of our model, this time using 2000 iterations.
For each party we plot their results in a histogram, and consider how our predictions match up to those made by other sources (which have been helpfully grouped on May2015) for those parties where such data is available.
The slideshow below shows the results for each party, you can either use the arrow icons to cycle through the pictures, or the thumbnails at the bottom.
Unsurprisingly, in such a short amount of time there is no major change in the model, with only changes of under half a seat occurring for all parties.
It should be noted that the increase for the Independent category will be mainly caused by the inclusion of Alliance in this group, as the party were deemed to have too low a chance of having any seats to merit maintaining within the model.
On a constituency level, only two seats have seen a swing (the methodology of which is explained here) of more than 10%:
- The high-profile Sheffield Hallam sees Nick Clegg gain a swing of 15.3% over Labour, making his chances of winning almost 90%. This is most certainly due to reports this week suggesting that tactical votes from Conservative supporters will go in his favour.
- Watford, a seat we identified on Tuesday as a marginal to watch, also sees the favourite pull away, this time it is the Conservatives who benefit from a swing of 14.3%, seeing their chances of victory increase to 76.5%.
What can the parties expect to finish with?
We can now do our usual trick of running multiple simulations of our model, this time using 2000 iterations.
For each party we plot their results in a histogram, and consider how our predictions match up to those made by other sources (which have been helpfully grouped on May2015) for those parties where such data is available.
The slideshow below shows the results for each party, you can either use the arrow icons to cycle through the pictures, or the thumbnails at the bottom.
Conclusion
As with all our projections so far, we have underestimated the SNP and overestimated UKIP compared to other election forecasters. This is potentially a flaw in how we calculate our probabilities - certainly we will be reviewing whether we chose the right value of ϕ after the election.
However, as we have regularly criticised the absolutist nature of the way others have calculated their seat projections, it may in fact eventually be vindication of our model.
Needless to say, there is only one way we are going to find out, and that is by seeing the final results. And once again, a reminder that if you are planning on staying up to see the results come in, then we are offering a very unique take on covering the election.
Appendix - constituency dossier
If you wish to have a handy guide to all 650 constituencies in the UK for tonight, then download the file below:
As with all our projections so far, we have underestimated the SNP and overestimated UKIP compared to other election forecasters. This is potentially a flaw in how we calculate our probabilities - certainly we will be reviewing whether we chose the right value of ϕ after the election.
However, as we have regularly criticised the absolutist nature of the way others have calculated their seat projections, it may in fact eventually be vindication of our model.
Needless to say, there is only one way we are going to find out, and that is by seeing the final results. And once again, a reminder that if you are planning on staying up to see the results come in, then we are offering a very unique take on covering the election.
Appendix - constituency dossier
If you wish to have a handy guide to all 650 constituencies in the UK for tonight, then download the file below:

gedossier.xlsx |
Alongside each seat is their region, and the first, second and third favourites under our model. The incumbent party is highlighted in bold and italics. The column on the left ranks the 100 closest seats.
General Election articles
Previous: Marginal constituencies
Next: The Aftermath(ematics)
Previous: Marginal constituencies
Next: The Aftermath(ematics)