A fund running a single static VaR model in February 2020 got the same position-size signal in week one as it did in week four. By the time historical volatility caught up with reality, the portfolio was already down 34%. The model wasn't broken — it was answering the wrong question. It calculated risk for one average regime, not for the regime actually in progress.

Value at Risk assumes a single return distribution. In practice, markets run three distinct distributions. Bull regimes produce tight, positively skewed returns. Bear regimes fatten the left tail. Crisis regimes — defined by correlation spikes and liquidity collapse — produce distributions that bear no statistical resemblance to the other two. Blending them into one model is like designing a bridge for average weather and ignoring cyclones.

CONCEPTRegime-conditional VaR calculates separate loss estimates for each distinct market state, then applies the estimate matching the current regime.
WARNINGA single-regime 99% VaR calibrated on mixed data will systematically underestimate tail losses during crisis by 2x–4x — the exact moment accuracy matters most.
KEY IDEARegime detection (Hidden Markov Models, volatility thresholds, or spread regimes) must precede VaR calculation — not follow it.

The mechanics work as follows. Traders first classify historical dates into regimes — typically using a Hidden Markov Model on realised volatility and cross-asset correlation, or simpler threshold rules such as VIX below 18 (bull), 18–30 (bear), above 30 (crisis). Each labelled subset then trains its own VaR model. Bull-regime VaR at 99% on a $500,000 equity position might produce a one-day loss limit of $8,500 (1.7% of position). Crisis-regime VaR on the identical position might produce $31,000 (6.2%) — a 3.6R difference. Position size scales accordingly.

99% VaR — One-Day Loss as % of Position1.7%Bull3.1%Bear6.2%Crisis0%

The practical implementation requires three decisions: regime classifier, lookback window per regime, and a transition rule. Many quantitative desks use a 60-day lookback within each regime rather than a fixed calendar window, ensuring the estimate reflects regime-specific volatility density rather than diluted mixed-sample data. The transition rule is critical — position resizing triggers when the classifier confirms a regime shift for two consecutive sessions, not on a single reading, to avoid whipsaw. Foundational research on distributional regime behaviour appears in the academic literature on Value at Risk and is formalised through frameworks like Hidden Markov Models, while the statistical properties of fat-tailed crisis distributions are detailed under tail risk research.

Regime-conditional VaR doesn't eliminate drawdowns. It ensures the size of each position matches the statistical character of the environment it's actually operating in — not the average of all environments ever observed.

A risk model calibrated on everything is optimised for nothing.

This content is for educational purposes only and does not constitute financial product advice. Past performance is not indicative of future results. Profit Logic Ltd (ACN 688 669 936) accepts no responsibility for errors or omissions in this content or anywhere on this website. Always seek advice from a licensed financial adviser before making investment decisions.