SOPHIE
SOPHIE Daddy Quant Blog

Command Palette

Search for a command to run...

Overview

A quantitative framework for identifying structural vulnerabilities in systematic strategies and building regime-aware portfolios. This covers the difference between alpha decay and strategy decay, and introduces Minimum Regime Performance (MRP) to solve the blindness of traditional full-sample metrics like the Sharpe ratio.

Core Concepts

  • Alpha Decay vs. Strategy Decay: Alpha decay is the natural half-life of a signal due to crowding. Strategy decay is a structural breakdown in the foundational logic due to macroeconomic regime shifts.
  • The Illusion of Perfection: Full-sample metrics (Sharpe Ratio, Max Drawdown) assume markets are ergodic and hide a strategy's vulnerability to specific hostile environments.
  • Factor Fragility: Empirical demonstration of factor performance across the Investment Clock. For example, Momentum suffers from the "Winner's Curse" during sharp contractionary inflection points, whereas Quality acts as a structural anchor.
  • Minimum Regime Performance (MRP): A dynamic, combinatorial search algorithm that finds the contiguous market era where risk-adjusted compounding was fundamentally weakest. It serves as a pre-optimization threshold filter.

The Mathematics of MRP

MRP actively searches across defined regimes (using HMMs or Macro Clustering) for the lowest Sharpe ratio.

Single Split MRP:

MRP1(x)=min⁡t1∈[d, n−d]{min⁡(S(r1),S(r2))}\text{MRP}_1(x) = \min_{t_1 \in [d,\, n-d]} \{ \min(S(r_1), S(r_2)) \}

Multiple Splits MRP:

MRPs(x)=min⁡T{min⁡(S(r1),S(r2),…,S(rs+1))}\text{MRP}_s(x) = \min_{T} \{ \min(S(r_1), S(r_2), \ldots, S(r_{s+1})) \}

Number of valid splits (combinatorics):

ns=(n−sd−d+ss)n_s = \binom{n - sd - d + s}{s}

Optimization Meta-Risks

  • Look-Ahead Bias: Historical MRP pinpoints exact regime boundaries ex-post, whereas live algorithms suffer statistical lag.
  • Historical Overfitting: Allowing too many regime splits (high ss) data-mines the backtest into transient noise.
  • The Small-Sample Problem: Heavy optimization against rare but severe regimes (the "Peso Problem") forces rejection of long-term robust strategies.
  • Alpha Destruction via Hedging: Over-optimizing for "regime neutrality" strips away compensated structural risk premiums.

Key Takeaways

  • Full-sample Sharpe and Max Drawdown can both mask a strategy that's structurally fragile in specific macro regimes — MRP exists precisely to surface that hidden weakness.
  • Momentum's high full-sample average return hides a deeply negative MRP (the "Winner's Curse" in contractions); Quality's lower average return comes with a strictly positive MRP across all regimes.
  • Use MRP as a pre-optimization threshold filter, not as an input to Mean-Variance Optimization directly — it's a non-linear combinatorial search, not a smooth objective.
  • Don't over-optimize for regime neutrality: factor premiums exist to compensate for un-hedged structural risk, and stripping that away just replicates the risk-free rate.

Related Reading

Back to article