Notes
Working notes on portfolio construction, risk and applied modelling — published here first, then shared on LinkedIn. Three a week: a module note on Monday, an analysis piece on Wednesday, and something from the data on Friday.
Published
Diversification failed exactly when it was needed, because the correlations everyone had modelled were estimated over a period that did not contain the event. My own multi-asset work found the same pattern in 2022.
What mean-variance optimisation produced on my own data, and where post-modern portfolio theory argues the framework misreads risk by treating upside and downside volatility as the same thing.
Equities and long bonds moved in opposite directions on average; gold showed little relationship with anything and was the most useful diversifier. Then 2022 broke the equity-bond relationship.
All notes are my own work and do not represent the views of any employer or client.
What I write about
Construction, diversification and what happens to both under stress. Where the standard framework holds and where it quietly stops working.
What models actually deliver once tested out of sample — including credit default prediction and the trade-offs between accuracy and interpretability.
What has to change before imported models work on thinly traded markets. The constraint is usually the data, not the method.
Elsewhere
Posts go out on LinkedIn three times a week. Longer pieces live here.