The 1998 LTCM collapse, the 2007 “Quant Meltdown,” and the March 2020 COVID crash have repeatedly demonstrated: quantitative strategies perform well in normal markets but often suffer far worse losses than expected in extreme conditions. Reasons: correlations spike during crises (“correlations go to 1”), liquidity suddenly dries up, and multiple quant funds deleveraging simultaneously creates cascading selling pressure.
## Market Risk Measurement
**VaR (Value at Risk)**: the maximum loss a portfolio might suffer over a given holding period at a specified confidence level (e.g., 95% or 99%). Example: “Daily VaR (95%) = 1 million RMB” means “under normal conditions, 95% of trading days, single-day loss does not exceed 1 million RMB.” VaR criticisms: it cannot capture tail events (losses in extreme situations can far exceed VaR); historical VaR models broadly failed during the financial crisis.
**CVaR/ES (Conditional VaR/Expected Shortfall)**: an improvement over VaR — the expected value of losses exceeding the VaR threshold. CVaR better captures extreme tail risk; Basel III has shifted bank risk capital requirements from VaR to ES.
## Factor Risk Decomposition and Stress Testing
**Factor risk decomposition**: breaking a portfolio’s total risk into contributions from different factors (market factor, sector factors, style factors). BARRA (MSCI subsidiary) multi-factor risk models are the industry standard; leading domestic quant institutions use similar frameworks for risk attribution and position management.
**Stress Testing**: subjecting the current portfolio to historical extreme scenarios (2008 financial crisis, 2015 A-share crash, March 2020 black swan) or hypothetical extreme scenarios to estimate potential losses. Stress testing complements VaR/CVaR limitations; together they form a complete risk management framework.
See [Quantitative Investing Intro](https://sunqi.org/quantitative-investing-intro-en/) and [MSCI Barra risk model documentation](https://www.msci.com/barra-models).
