Live Multi-Asset Risk Cockpit

This page runs a live, ticking dashboard, correlated prices moving in real time, rolling Value-at-Risk and Conditional VaR recomputed on every tick, and a correlation matrix that updates as the simulated market evolves. Read the note below before you dive in: this is exactly the risk machinery a trading desk cockpit runs, on a feed built for demonstration, not a live market connection.

An Honest Note On "Live"

What This Dashboard Actually Runs On

This is a static site with no backend and no live market data subscription, so the ticking prices here are a simulated correlated market, generated in your browser by correlated geometric Brownian motion via Cholesky decomposition, not a real feed. Every tick is randomly generated fresh, on your device, using the parameters you set.

What is genuinely real is everything downstream of the price feed: the Cholesky decomposition linking the assets, the rolling parametric and historical VaR and CVaR calculations, and the live correlation estimation, are the same formulas and the same code path a real trading risk cockpit would run against an actual data feed. Swap in a real price stream, keep the rest of this page exactly as it is.

Portfolio & Market Parameters

Weight: Equities 40%

Weight: Rates/Bonds 30%

Weight: Credit 30%

Equities Volatility (Annualized) 18%

Rates/Bonds Volatility (Annualized) 7%

Credit Volatility (Annualized) 10%

Stress Correlation Multiplier 1.0×

VaR/CVaR Confidence Level 95%

Weights are automatically renormalized to sum to 100%. The stress multiplier scales all pairwise correlations toward 1 to simulate a crisis-correlation regime; try setting it above 1× and watch VaR and CVaR react even before volatility itself has moved.

Portfolio Risk, Recomputed Every Tick

Simulated Feed: Paused

Portfolio Value (Indexed To 100)

100.00

Parametric VaR (1-Day)

Historical VaR (1-Day, Trailing)

Historical CVaR (1-Day, Trailing)

Simulated Live Portfolio Value

Real-Time Portfolio Risk Monitoring

The Method

Σ = D·R·D (D=diag(volatilities), R=correlation), Σ = LL′ (Cholesky)
r_t = L·z_t, z_t~N(0,I) (correlated shocks from independent draws)
VaR_β^param = −(μ_p + z_βσ_p), CVaR_β^param = −(μ_p − σ_pφ(z_β)/(1−β))
VaR_β^hist = −Percentile_1−β(trailing simulated portfolio returns)

Correlated asset shocks are generated by Cholesky-decomposing the target covariance matrix and multiplying independent standard normal draws through the resulting lower-triangular matrix, verified in this build to reconstruct the exact target covariance matrix (L·L′) to machine precision. Parametric VaR/CVaR use the closed-form normal formulas; historical VaR/CVaR are computed directly from the empirical distribution of trailing simulated portfolio returns. On long simulated samples, the two methods were verified to converge closely to each other, exactly as theory predicts when returns are genuinely normal, which is also precisely why real portfolios, whose returns are not perfectly normal, show a persistent gap between the two that itself carries information.

When To Actually Use This Approach

Intraday or daily risk monitoring dashboards where stakeholders need to see VaR, CVaR and correlation move as new data arrives, not just at end-of-day batch close.
Comparing parametric and historical VaR side by side as a built-in model-risk check, large, persistent divergence is itself a signal that the normal-distribution assumption is failing.
Demonstrating, live, how a stress-correlation regime (all pairwise correlations rising toward one) degrades diversification and inflates tail risk, even before any change in individual asset volatility.

Key Assumptions & Limitations

The feed on this page is simulated, not connected to any real market data source, a limitation of this static, backend-free environment rather than of the risk methodology itself.
Parametric VaR/CVaR assume normally distributed returns; real asset returns exhibit fat tails, so parametric VaR reliably understates true tail risk relative to a well-estimated historical or Monte Carlo approach.
Historical VaR/CVaR here are computed on a rolling simulated window and inherit whatever sampling noise that window contains, a real system would typically blend multiple estimation windows and methodologies.

Foundational References

Jorion, P. (2006). Value at Risk: The New Benchmark for Managing Financial Risk (3rd ed.). McGraw-Hill.

Rockafellar, R. T., & Uryasev, S. (2000). Optimization of Conditional Value-at-Risk. Journal of Risk, 2(3), 21-41.

J.P. Morgan/Reuters (1996). RiskMetrics Technical Document (4th ed.).

Want to build the portfolio behind this cockpit with tail risk in mind?

Open The CVaR Portfolio Optimization Engine →