We ran a polling simulation of 1000 Holyrood elections. Here's the results

THERE is a 99.7% chance of voters returning a pro-independence majority at the next Scottish Parliament election, according to a simulation of 1000 Holyrood contests.

The simulation, run using software developed by the Devolved Elections team, uses polling averages and random variations to predict 1000 different ways the Holyrood election on May 7 could go.

Dr Eoghan Kelly – who created the system alongside his software engineer partner Ayushi Gupta – said it helps give an insight into polling results by simulating things such as regional variations, voting patterns across wealthy or poorer areas or urban and rural ones, or margins of error in polling.

“For each party, sometimes it’ll be 0.5% over, sometimes it’ll be 3% down,” he explained. “All of those things are randomised, and then they’re all done on a bell curve.

“We protect the numbers at all times, so the swings don’t knock the numbers out. The numbers stay the same, but how it’s distributed changes.”

When the Sunday National ran a simulation, it predicted a 99.7% chance of a pro-independence majority, composed of the SNP and Greens. Seat totals of pro-independence MSPs were projected to vary between 80 at the highest and 64 at the lowest, with a mean result of 72.5.

By party, an average of the 1000 simulations predicted that the SNP are set to return 61.5 MSPs, while the Greens return 11. On the Unionist side, Labour were projected to win 17.8 seats, Reform UK 17.3, the Tories 12.2, and the LibDems 9.2.

Possible seat totals from 1000 election simulations (Image: Graph plotted from Devolved Elections’ simulation results)

The 1000-election simulation also projects the probability of each party winning each constituency seat. It said that some seats – such as John Swinney’s Perthshire North and Shirley-Anne Somerville’s Dunfermline – had a greater than 99.9% chance of being won by the SNP.

However, it also predicted an election battle in Edinburgh Central, where Culture Secretary Angus Robertson was given a 31.8% of retaining his seat, while the Greens’ Lorna Slater was given a 57.3% chance of scooping it.

Ettrick, Roxburgh and Berwickshire was projected to be the closest contested constituency in Scotland, with Reform UK given a 50.5% chance of winning it, against the Tories on 47.9%.

Next was Argyll and Bute, where the SNP were given a 55.6% chance of winning against the LibDems on 44.4%.

Elsewhere, the Devolved Elections website also features a “Nowcast”, which uses the last six months of polling, weighted according to recency, to predict how the election might go. Kelly described this as the “baseline” that the other features lean on.

When the Sunday National used the service on Friday, it predicted that the SNP would return 62 MSPs, with Labour a distant second on 18. Reform UK were projected to win 17, the Tories 11, the Greens 11, and the LibDems 10.

Seat counts (Image: A ‘Nowcast’ snapshot from DevolvedElections.co.uk)

The site also allows users to run 1000 simulated Welsh parliamentary elections – which are also due to be held on May 7 this year. When the Sunday National did so, it gave the pro-independence Plaid Cymru an 81.7% chance of emerging as the largest party, against Reform UK on 18.3%.

Labour, who have run the Senedd since devolution was introduced, were given a 0% chance of winning the 2026 elections.

Kelly – a polling expert who lectures at Trinity College Dublin and Queen’s University, Belfast, while also running the Polling London Project at the Mile End Institute of Queen Mary University of London – said he had come up with the idea for the system while working at the Scottish Election Study.

He said: “It’s hard to visualise what poll numbers mean, particularly in [an additional member, proportional representation system] where you’ve got two different ballots and they interact in a very specific way.

“And with the sheer volume of parties, now that Reform are in the mix and everyone’s getting 10 to 30%, it’s very hard to visualise. So I just thought, I’ll go make one.”

Explaining the system, Kelly said: “There are margins of errors for each party in each direction; sometimes all of the things go right or wrong for a party.

“If you do it over and over again, you will get variations, but the core will remain.”

Asked if a 1000-simulation run of a poll could end up more accurate than the poll itself, he said: “That’s the theory. It depends on the polls. We don’t run the polls, so we’re at the mercy of the pollsters.

“That’s also part of the benefit. If the polls are overestimating, sometimes you can still sort of see where that will happen. So if the SNP are being overestimated, you can get a vague idea of where they’ll fall.”

SNP leader John Swinney may lead a pro-independence majority despite not winning the majority of the votes

The polling expert said that polling and simulations consistently showed that a majority of votes in Scotland were set to go to Unionist parties – despite simultaneous projections of a pro-independence majority of MSPs.

“The pro-indy thing in Scotland is being boosted by the electoral system,” he said.

“The Unionists are probably going to win a comfortable majority of the votes, mid-50s, but the system boosts parties that win the constituencies, and the SNP are winning the constituencies.”

Kelly said that designing the Holyrood system to favour the constituency vote was “a deliberate ploy by Labour”.

“That was the plan. It didn’t work out very well, but it was the plan. They could have made it more equal by having the same number of list seats and constituency seats, but they chose not to,” he said.

You can find the election simulation tool, as well as other resources including seat projection calculators and voting system explanations, on the DevolvedElections.co.uk website.

We ran a polling simulation of 1000 Holyrood elections. Here’s the results

Tags: