Portfolio Risk Analytics

Last reviewed: by .

Professional-grade risk analytics including Value at Risk, stress testing, and scenario analysis to understand and manage portfolio risk exposure. Available with a Professional subscription.

VaR vs Stress vs Tail Risk: When Each One Matters

The three risk measures answer different questions. VaR answers "in a normal-ish day, how bad could the typical worst case be?" A 95% VaR of $4,000 means that, on roughly 1 day in 20, the portfolio would lose at least that much, estimated from historical returns (with a Monte Carlo fallback). It is calibrated to recent regime data, so it is most useful when conditions resemble that regime. Stress tests answer "what if conditions do not resemble that regime?" They let you bolt on a 2008-style equity drop, a 2020-style vol spike, or a custom rate-hike scenario and see how the book responds. Tail risk and expected shortfall answers "given that I am in the worst 5% of outcomes, what is the average loss?" This is important because two portfolios can have identical VaR but very different expected losses in the tail.

Workflow: Sizing a New Position

Before adding a position, the typical pattern is: open the Risk page with the existing book loaded, add a hypothetical leg, and watch how VaR and expected shortfall change. If the marginal contribution to risk is small, the position is well-diversified relative to the rest of the book; if it is large, you are concentrating risk in whatever factor that position carries (vol, rates, single-name beta). The correlation matrix helps explain why: if the new position has high correlation with several existing ones, the marginal risk contribution will be larger than the position's standalone risk would suggest.

What This Page Is Not

Risk metrics are model-based estimates, not guarantees. VaR misses regime breaks (the exact moment risk is highest, the historical window is least informative); stress scenarios are only as good as the shocks you choose; the efficient frontier assumes returns and covariances are stationary, which they are not over the timescales most retail trades operate on. Treat these tools as a discipline for surfacing concentrations and asymmetries, not as a "the model says I'm safe" green light.

Choosing the VaR Window: 1 Year vs 90 Days

VaR is computed from your portfolio's reconstructed return history, with a Monte Carlo simulation as the fallback when there isn't enough price history to read an empirical percentile. The one knob you control is the lookback window: 1 Year (252 trading days, the industry-standard window) or 90 Days (more recent, more reactive to the current regime). The tradeoff is the usual one - the longer window is more stable but slower to react to a regime shift, while the shorter window tracks recent conditions but is noisier and more sensitive to a single outlier day. Both the 95% and 99% confidence levels are shown; the 95% number is the more practically useful read, because the 99% number is dominated by data quality at the extreme tail. Historical and Monte Carlo VaR are the two estimators here; there is no separate parametric (Gaussian) mode.

Reading the Correlation Matrix

The correlation matrix is the entry point for diversification decisions. It correlates your holdings by underlying ticker (the largest books are truncated to the top handful of symbols), computed on returns over a 90-day price-history window when history is available, falling back to sector-based estimates when it is not - the panel labels which source it used. High positive correlation between two underlyings (the matrix flags the heavy zone) means they move together in most conditions, so holding both does not reduce risk meaningfully and the book's effective concentration is higher than the position count suggests. Low or negative correlation is what diversification actually looks like. Correlations are not stable over time, so a pair that looked diversifying in one regime can co-move heavily in another (the March 2020 cross-asset spike is the canonical example). Read the matrix as a current-state diagnostic, not a permanent property of the book.

Stress Scenario Design

The stress builder accepts four shock dimensions: an equity move (percentage spot shock), a volatility change (IV shock), a rate change (in basis points), and time decay (days forward). The preset scenarios cover the common shapes - Market Crash, Volatility Spike, Flash Crash, Credit Event, Rate Hike, and Bull Rally, alongside quick templates like Bear Crash, Vol Crush, and Time Decay - and any preset can be loaded and then adjusted by hand. Each shock runs through the position-level Greeks, so the output is a per-position and aggregated portfolio P&L estimate. The estimate is a linear-plus-convexity read off the current Greeks (delta, gamma, vega, theta), so it stays accurate for moderate shocks but drifts at large moves where higher-order effects dominate - which is why full Monte Carlo through the Backtesting page is the right tool for very large hypothetical moves.

Margin and Buying Power

The margin panel estimates the book's margin requirement and the buying power it consumes under Reg T rules (initial 50%, maintenance 25% for equity, with a 2x Reg T buying-power figure). It is an approximation - actual margin varies by broker, and portfolio-margin accounts will differ - so the panel exposes a configurable margin-call threshold (brokers sit anywhere from roughly 50% to 80%) and a usage gauge that bands green, amber, and red against it. It also surfaces maintenance requirement, excess liquidity, and the buffer remaining before a margin call. The most useful read is how close a contemplated new position pushes you toward your own threshold before you confirm the real number with your broker.

This page is part of the Options Analysis Suite features overview. Browse the full documentation.