Statistical proof — Out-of-Sample
The figures below are out-of-sample (OOS), measured on data the model never saw. We also publish what does not work.
−15.4 pts
Max drawdown reduction on STOXX (35.5% → 20.1%)
0.62
STOXX Sharpe
vs 0.49 buy & hold
20.1%
STOXX max drawdown
vs 35.5%
0.72
S&P 500 Sharpe
vs 0.74 — no edge
3
Regime states
Risk-On / Neutral / Risk-Off
Validation protocol Walk-forward · Gaussian HMM refit at each date · defensive overlay validated on non-US equities, honest about the absence of S&P 500 edge.
Access on request
Paper reserved — research desk
This research paper is not for public sale. Full access (methodology, code and results) is granted on request, subject to conditions, by our research desk.
Contact the research desk Keywords:
Exclusive Signal
Ready for the next step?
Don't miss any signal. Join our research community for unrivaled macro analysis and quantitative models.
Léo Lombardini
Trader, Economics & Quant
Passionate about market analysis and statistical modeling, Léo oversees the strategic allocation of the model portfolio and the development of Horacle Capital's quantitative frameworks, as well as writing weekly articles.
Learn more about the author →