Lucky or Good? Outcome Noise, Effective Sample Size, and the Attribution of Skill

Battle Count: 50

The paper directly addresses active equity management (long-only mutual funds) and high-frequency systematic trading, placing them on the knowability map. It discusses the statistical impossibility of distinguishing skilled from unskilled fund managers from outcome records alone, which is central to fund-of-funds allocation and manager selection. The Berk-Green equilibrium model and Fama-French findings are discussed. However, the paper is primarily a conceptual/diagnostic framework rather than a trading strategy or quantitative model. It informs the epistemic foundations of performance attribution in asset management but does not propose new trading algorithms or portfolio construction methods.

The core framework is a single inequality (SNR × √N_eff > 2.5/d) requiring only two parameters per domain. Placing a domain on the map requires estimating σ (outcome noise), σ_skill (cross-actor skill spread), and N_eff (effective independent observations). The equicorrelation adjustment N_eff = N/(1+(N-1)ρ) is straightforward. The main complexity lies in the judgment required to calibrate these parameters for a new domain, not in computation. The population-validation methodology (Framingham-style) is more involved but is described as a general prescription rather than a specific algorithm.

The framework is clearly specified with a single inequality and calibration values drawn from published literature. However, the paper does not release code or a computational tool. Calibration parameters (σ, σ_skill, N_eff) for each domain are sourced from prior empirical studies and involve judgment calls (e.g., correlation adjustments). The knowability map placements depend on these calibrations, though the paper notes claims are supported by approximate regions rather than point estimates. No dataset or code repository is provided.