ETFWealthIQ
Most portfolio advice hands you a static allocation and hopes you stick with it. The hard part was never picking the funds — it’s holding the line when markets move.
This blog is about the other kind of portfolio thinking: construction methodology you can actually understand, backtested on real ETF data with the open-source code that computed it, framed so you can decide for yourself.
Written by someone who spent 25 years inside bank Treasuries, where getting risk wrong was not an option.
Finance research on the gap between finding a pattern and keeping it: trading costs, falsification tests, signal libraries, ETF tax structure and framing bias.
Finance research on hidden dependence: risk models that miss a common factor, correlation used as a proxy for outcomes, and exposure travelling through supply chains.
Finance research on what counts as evidence: reading insignificant results, backtest specification choices, the axioms under systematic investing, and access.
Research on grading backtests for robustness, selection bias in econometrics, decaying model performance, and judging an allocation by what a portfolio already holds.
Research on reconstructing a decision: explaining a single trade, delegated judgment and accountability, auditable AI outputs, and a finding that moved with its rule.
New research on checking a method: simulated lifetime effects of AI financial advice, carbon beta as a measured exposure, and skewness under anomaly returns.
New retirement research: pension allocation and insurance as one decision, guarantees inside the tax code, concentrated care costs, and risk that shifts with horizon.
A large AI fund reportedly lost 67% in a month running 4x leverage. Its long and short legs were the same bet — so the leverage landed on gross exposure, not net.
New research on visibility-biased risk data, product labels covering different contracts, and why a good price fit is not a recovered risk distribution.