Notes on whatever I'm currently building or thinking about — a UK election model, a dissertation project, and whatever comes next.
Flows, uncertainty, and tactical voting — opening up the pipeline that turns any group structure into seat-by-seat forecasts.
A dissertation project's baseline house-price model finds sensible, well-behaved relationships everywhere — until it hits the one UK housing market that isn't behaving like the rest of the country.
Why single-category constituency models are running out of road, and what unmixing 632 seats’ census composition into a handful of shared archetypes gets us instead.
Running the full standard toolkit — K-means, HAC, DBSCAN and GMM, across PCA and UMAP, scored on four metrics — on Britain's constituencies, and reporting honestly on where it falls short.