Live: Follow results of the 2026 Senedd & Holyrood elections
MRP/Projection Accuracy
This page compares the various MRPs and more traditional seat projections of the 2026 Senedd and Holyrood elections to the actual results. All MRPs and projections with publicly available seat level numbers are included. We use either the final projection as announced by the organisation, or if no final call was made, the final vote projection on the night of May 7th. After polls closed but before counting began. Websites which do not provide a polling average of their own (eg. Cavendish Consulting) or which do not provide regional list numbers (eg. Election Polling) are not included. The only MRP to be excluded is Stonehaven MRP as they did not provide seat level vote shares.
This aims to provide a public record of general accuracy. MRPs collect their own data but projections use averages from pollsters, so if there is a systemic polling error it should impact every projection site. For transparency, our own Nowcast at 11pm on May 7th is included in the same way as every other organisation. All data included will be freely available for download after the last results are in, we only ask that you accredit us and the other organisations if you reuse it.
We track the headline seat numbers and seat winners for every major party for all the projections as well as the constituency level vote shares.
Where possible, all errors are measured by the Absolute Margin of Error (AME) at the national and constituency level, this is how far from being correct a seat projection is. If a party is overestimated by 1 or underestimated by 1, those are considered to be equivalent. We do not weight anything here and our own projection is treated exactly the same as the others. This page will update automatically as results come in.
A note on projections and MRPs
Creating accurate seat by seat level projections is very difficult and most of us are operating on tiny budgets and in our spare time. Most are made by one or two people using their own knowledge and experience.
This page is not meant to take a swipe at any of the projections as being biased or bad, they are not. This is just a sense check of which was closest this year and the order could be entirely different next time or if the polling error had gone in a slightly different direction.
The MRPs are major works by teams of people working for companies which earn millions each year to produce exactly this. If any projection beats any MRP then that MRP has failed to do their job correctly.
The J.L. Partners MRP is particularly terrible because it didn't use the correct electoral system in either Scotland or Wales. This is an indictment of the industry and journalists who do not question the wisdom of MRPs.