About ReadingSignal
ReadingSignal is an independent UK election forecaster built around three principles:
- Show the working. Every projection comes from a transparent model pipeline. No black boxes.
- Be honest about uncertainty. Confidence intervals are wide because the underlying reality is uncertain. Forecasts are probability distributions, not predictions.
- Go beyond uniform swing. Averaging polls and applying one national swing to every seat misses how unevenly votes are actually distributed. The model converts vote intention into seats using demographic structure and vote-switching dynamics, with empirically calibrated geographic variance.
Where it's heading: the goal is a genuinely multi-signal forecast — combining polling with independent sources like prediction markets, leader approval, and economic fundamentals. Independent signals with uncorrelated errors, properly combined, should beat any single one. That's the roadmap.
The site covers UK general elections, devolved elections (Scotland and Wales), and major local election cycles.
Current calibration
The model was recalibrated in late May 2026 against the local election results and the YouGov / More in Common / Electoral Calculus MRPs. The previous version projected Labour roughly 130–150 seats higher than the MRPs — an 8pp over-projection in Labour's national vote share. The recalibration closes about three-quarters of that gap. The current headline is Reform as the largest party, with Labour meaningfully below the published MRP consensus.
Three issues are worth surfacing rather than leaving implicit. First, Labour's projected seat count sits below both YouGov (138) and More in Common (165). The two obvious candidate causes — the differential turnout assumption (Conservative voters more reliable than Labour) and a possible over-reduction of anti-Reform tactical voting — were both investigated with explicit counterfactuals and ruled out. The remaining gap is in the seat-conversion layer, which is the next thing under review. The honest read is that the model is more bearish on Labour than the published MRPs and we don't yet have a definitive explanation for which side is closer to right.
Second, Labour's national vote share projects about 2pp above the turnout-adjusted polling target. Third, Liberal Democrat vote share projects about 2.7pp above its target. Both stem from the same mechanism — the way the model routes Labour's collapse through tactical voting, which spills some Labour voters into Liberal Democrat and Abstain rather than Green where the IPF calibration expected. Neither materially distorts the seat counts (Labour seats are low not inflated; Liberal Democrat seats are in the MRP band), but they are publicly checkable shares that don't quite match the polls.
Both vote-share residuals and the Labour seat conversion question are queued as the next pieces of calibration work. The model that's on the site is materially better calibrated than the previous version, but not finished.
Who built it
ReadingSignal is built and maintained by Jim Moodie. Jim spent 7.5 years as COO of Prolific, the online research platform used worldwide by academic researchers and social scientists — work that involved many of the same methodological challenges pollsters face: building representative samples, controlling for response biases, and assessing data quality at scale.
He has a long-standing interest in forecasting and prediction markets. ReadingSignal is one of several prediction modelling projects he's working on. More background at jahm.me.
Get in touch
Questions, corrections, or methodological challenges welcome via Twitter/X: @yojimbo23.