← notes

research study · complete

Systems Thesis

Portfolio allocation treated as layered system optimization — walk-forward evaluation with ablations and negative results.

Scope 60-arm study Method walk-forward Stance negative results included

What was studied

Walk-forward evaluation across three portfolio universes (Top-100, Top-250, Top-500). Screening and weighting choices under explicit turnover penalties. Ablations that remove or swap components to identify what actually moved results. Deployment limits — where a clean backtest stops being an honest production story.

Results

Top-100 best 1.47 Sharpe · 21d momentum + equal weight
Top-500 best 1.76 Sharpe · 63d vol-adj momentum + 21d rank-and-hold
Research philosophy

Understanding system behavior under constraints beats maximizing a single return chart. If an arm looks strong only because turnover was free, or because leakage snuck in, it fails the study.

Connection to other work

Fractal inherits the journal instinct from this work. AuditLayer inherits the production skepticism. The spatial work inherits the "report conditions separately" habit. The thesis is older, but the standard stuck.