Concept · The RiskModels engine (ERM3)
L-star rule
The L\* rule: a cost-aware, learned selector that picks the shallowest level whose expected out-of-sample variance reduction outweighs the added hedging cost of going deeper — describing a name at the coarsest level that still captures its risk. Scored point-in-time, so the dispatch is backtest-safe.
In depth
Picks the simplest level whose marginal explained return clears a threshold θ (default 1%, tuned on 275 tickers over 60 months). Keeps descriptions at the coarsest level that still captures a name's risk.
Formula
L* = L3 if ER^L3_sub ≥ θ; L2 elif ER^L2_sec ≥ θ; else L1
In the methodology
Choosing a level: the L* rule →Referenced by (5)
- NVIDIA: The Economic-Profit Burden of an AI Bottleneck
A RiskModels.app visual case study in market-implied expectations
- The Persistence of Stock-Selection Residuals
A point-in-time, holdings-based decomposition of mutual fund performance — the stock-selection residual persists out of sample; style timing and sector timing, measured the same way, do not.
- Every Position Has a Level Too
How RiskModels picks the right hedge depth automatically, per stock, per day
- Cascade Hedging and the Cost of Interpretability
Subsector ETF value, joint optimization, and executable hedge layers across 9,074 US mutual funds
- ERM3 Cascade-Residual Persistence and the Allocator Skill Signal
Top-decile rank persistence, active-share comparison, and tail-stratified inference across 1,000 top-AUM US mutual funds