Our publications lay out the philosophy and discipline behind the Arctan product family — how each design choice responds to the fundamental challenges of applying quantitative methods to financial markets.
August 2026 · SSRN Working Paper
Crowding Cycles: Concentration, Relative Run-ups, and the Value of Waiting
Evidence from a century of U.S. equity data (CRSP, 1926–2024). Using a pre-registered research design, the paper asks whether endogenous crowding cycles — the Nifty-Fifty, dotcom and 2021 mega-cap unwinds among them — can be identified early enough to de-risk, and what kind of response survives out-of-sample testing.
A plain-language guide to the discipline behind our research: point-in-time data, pre-registration, held-out testing, placebo benchmarks, data-integrity gates, and treating a well-established “no” as a first-class result.
Why the three structural challenges of machine learning in finance — too many factors, one timeline, non-stationarity — cannot be dissolved, only managed, and Arctan’s three architectural responses: curated features, regime conditionality, and continuous monitoring.
