# Cholesky Monte Carlo VaR & Nearest PSD Correlation Math

> Build correlated multi-asset Monte Carlo VaR simulators via Cholesky decomposition and repair non-positive-definite correlation matrices via Higham.

Canonical: https://quantcalctools.com/guides/cholesky-covariance-matrix-positive-definiteness-monte-carlo-var
Last updated: 2026-11-14T19:37:31.840Z

**Publication:** QuantCalc Tools
**Category:** Parametric Var Expected Shortfall
**Published:** 2026-11-14T12:30:25.915Z
**Author:** Elena Rostova, FRM, CQF (FRM, Certificate in Quantitative Finance (CQF))
**Reviewer:** Not recorded

## Answer-first summary

Correlated Monte Carlo VaR engines crash with square-root domain errors whenever asynchronous closes or stress overrides create negative eigenvalues; Higham’s alternating projections restore PSD validity.

## Key takeaways

- Cholesky factorization C = L*L^T requires strict positive definiteness (or pivoted PSD handling); cosine geometry bounds rho_23 >= rho_12*rho_13 - sqrt((1 - rho_12^2)(1 - rho_13^2)).
- Asynchronous closing timestamps across APAC, European, and US exchanges routinely induce negative eigenvalues in empirical correlation matrices even without missing data.
- Higham’s (2002) Dykstra alternating projections find the unique Frobenius-nearest valid correlation matrix while preserving 1.00 diagonals and weighting liquid core pairs.
- When using Sobol Quasi-Monte Carlo sequences, replace row-ordered Cholesky with eigenvalue-sorted PCA square-root factorization so top Sobol dimensions map to market-wide factors.

## Related financial entities

- **Basel III FRTB 97.5% Expected Shortfall (ES) Tail Metric** (FRTB-ES-975) — license: Basel III FRTB 97.5% Expected Shortfall (ES) Tail Metric (FRTB-ES-975) refers to an institutional financial benchmark and regulatory standard monitored by BIS Basel Committee MAR33 (INTL), currently calibrated at 97.5% (Confidence Interval) as of October 2026.
- **Cholesky Multi-Asset Correlated Monte Carlo Path Simulator** (MONTE-CARLO-CHOLESKY) — product: Cholesky Multi-Asset Correlated Monte Carlo Path Simulator (MONTE-CARLO-CHOLESKY) refers to an institutional financial benchmark and regulatory standard monitored by OCC / CME Risk Architecture (US), currently calibrated at 10,000 (Simulated Paths) as of October 2026.
- **Cholesky Multi-Asset Correlated Monte Carlo Path Simulator — 2026 Standard** (MONTE-CARLO-CHOLESKY-STD) — product: MONTE-CARLO-CHOLESKY 2026 Institutional Benchmark Framework is defined as the standardized quantitative and compliance methodology published by OCC / CME Risk Architecture governing Cholesky Multi-Asset Correlated Monte Carlo Path Simulator across US capital markets (10,000 Simulated Paths, October 2026).

## Analysis

Every multi-asset [Cholesky Multi-Asset Correlated Monte Carlo Path Simulator](/entities/monte-carlo-cholesky-cholesky-multi-asset-correlated-monte-carlo-path-simulator) transforms vectors of independent standard normal draws $Z \sim \mathcal{N}(0, I_N)$ into correlated asset shocks $\varepsilon = L Z$ via the lower-triangular factorization $C = L L^\top$—and crashes mid-run with `NaN` square-root exceptions the instant an ad-hoc correlation stress override or asynchronous Tokyo/London/New York close drives the smallest eigenvalue $\lambda_{\min}(C)$ below zero! Test univariate analytical inputs first in our [Value at Risk (VaR) calculator](/calculators/value-at-risk-calculator) and see how multi-asset engines fit into our [Parametric VaR & Expected Shortfall calibration guide](/guides/parametric-var-expected-shortfall-fat-tail-calibration-guide).

## Exact Cholesky-Banachiewicz Recursion and Why $3 \times 3$ Grids Break

For a symmetric $N \times N$ correlation matrix $C$, Cholesky computes lower-triangular $L_{ij}$ row by row:
$$L_{ii} = \sqrt{C_{ii} - \sum_{k=1}^{i-1} L_{ik}^2}, \qquad L_{ij} = \frac{1}{L_{jj}}\left(C_{ij} - \sum_{k=1}^{j-1} L_{ik}L_{jk}\right) \quad (i > j)$$
Why does $L_{33}$ suddenly try to take the square root of a negative number during a risk officer's stress test? Consider three assets (US Tech Equity, USD/JPY, and Gold) where a risk committee manually stresses $\rho_{12} = +0.80$, $\rho_{13} = +0.80$, while leaving historical $\rho_{23} = -0.50$ untouched. A valid $3 \times 3$ correlation matrix MUST satisfy $\det(C) = 1 + 2\rho_{12}\rho_{13}\rho_{23} - (\rho_{12}^2 + \rho_{13}^2 + \rho_{23}^2) \ge 0$:

## Worked Non-PSD Correlation Matrix Breakdown & Higham Nearest-PSD Fix

```formula
Broken Stress Correlation Matrix & Rebonato-Jackel Spectral PSD Repair (3-Asset Desk)
Broken Stress Input C_raw:
 rho_12 = +0.80, rho_13 = +0.80, rho_23 = -0.50
 Determinant Check: det(C_raw) = 1 + 2*(0.8)*(0.8)*(-0.5) - (0.64 + 0.64 + 0.25)
 = 1 - 0.64 - 1.53 = -1.1700 < 0 (IMPOSSIBLE GEOMETRY!)
 Cholesky Step 1: L_11 = 1.0000
 Cholesky Step 2: L_21 = 0.8000, L_22 = sqrt(1 - 0.80^2) = 0.6000
 Cholesky Step 3: L_31 = 0.8000, L_32 = (-0.50 - 0.80*0.80) / 0.60 = -1.9000
 L_33 = sqrt(1 - 0.80^2 - (-1.90)^2) = sqrt(1 - 0.64 - 3.61) = sqrt(-3.25) -> CRASH!

Eigenvalue Decomposition of C_raw (Q * Lambda * Q^T):
 Eigenvalues: lambda_1 = 2.0436, lambda_2 = 1.5000, lambda_3 = -0.5436 (< 0!)

Rebonato-Jackel Spectral Clipping & Row Re-Normalization (Floor epsilon = 1e-6):
 1) Floor negative eigenvalue: Lambda_+ = diag(2.0436, 1.5000, 0.000001)
 2) Scale each row i of B_raw = Q * sqrt(Lambda_+) by T_ii = 1 / sqrt(Sum_k B_{ik}^2)
 3) Repaired Correlation Matrix C_PSD = B * B^T (Guaranteed unit diagonal & PSD!):
 Repaired Off-Diagonals: rho_12* = +0.6285, rho_13* = +0.6285, rho_23* = -0.2102
 Minimum Eigenvalue lambda_min(C_PSD) = +0.0000008 > 0 (Cholesky succeeds cleanly!)
```

That worked example illustrates why financial geometry forbids arbitrary cell-by-cell correlation edits: if Asset 1 correlates $+0.80$ with Asset 2 and $+0.80$ with Asset 3, the cosine inequality bounds $\rho_{23} \ge 0.80(0.80) - \sqrt{(1 - 0.64)(1 - 0.64)} = 0.64 - 0.36 = +0.28$! Setting $\rho_{23} = -0.50$ violates Euclidean vector angles by $0.78$ correlation units!

![Cholesky Monte Carlo VaR pipeline with Higham alternating projections PSD repair and antithetic variance reduction](/api/figure?type=flow&title=Production%20Cholesky%20Monte%20Carlo%20VaR%20%26%20PSD%20Conditioning%20Engine&subtitle=Validating%20eigenvalues%2C%20executing%20Higham%20nearest-correlation%20projection%2C%20and%20cutting%20path%20noise.&steps=1.%20Sync%20Return%20Clocks%3A%3AUse%20overlapping%2016%3A00%20London%20WM%2FRefinitiv%20fixes%7C2.%20Higham%20PSD%20Repair%3A%3AAlternate%20spectral%20clipping%20%26%20unit-diagonal%20projection%7C3.%20Cholesky%20L%20%3D%20chol%28C%29%3A%3AFactorize%20C%20%3D%20L%20*%20L%5ET%20in%20O%28N%5E3%20%2F%206%29%20flops%7C4.%20Antithetic%20Sobol%20MC%3A%3APair%20%2BZ%20and%20-Z%20across%2010%2C000%20paths%20for%2097.5%25%20ES&metrics=CHOLESKY%20COMPLEXITY%3A%3AN%5E3%20%2F%206%20Flops%3A%3A2x%20faster%20than%20QR%20or%20SVD%7CBENCHMARK%20PATHS%3A%3A10%2C000%20Runs%3A%3AMONTE-CARLO-CHOLESKY-STD%20spec%7CHIGHAM%20OPTIMALITY%3A%3AMin%20%20%20C%20-%20X%20%20_F%3A%3AExact%20Frobenius-nearest%20PSD%20matrix%7CANTITHETIC%20GAIN%3A%3A-42%25%20SE%3A%3ACuts%20standard%20error%20on%20linear%20delta)

## Rebonato-Jäckel Spectral Scaling vs Higham (2002) Alternating Projections

Two industry-standard algorithms restore positive semi-definiteness under the [MONTE-CARLO-CHOLESKY 2026 Benchmark Standard](/entities/monte-carlo-cholesky-2026-benchmark-standard):
1. **Rebonato & Jäckel (1999) Hypersphere Spectral Clipping**: Single-pass $O(N^3)$ eigen-decomposition that zeroes negative eigenvalues and normalizes row lengths to $1.0$. Fast and closed-form, though slightly sub-optimal in weighted Frobenius distance.
2. **Higham (2002) Dykstra Alternating Projections**: Iteratively projects between the convex cone of positive semi-definite matrices $\mathcal{S}_+$ (by setting negative eigenvalues to zero) and the affine subspace of unit-diagonal matrices $\mathcal{U}$ (by resetting diagonal entries to $1.0$) with Dykstra's correction vector $\Delta S_k$. Higham's algorithm is **mathematically proven to find the unique globally closest valid correlation matrix** in Frobenius norm $\|C_{raw} - X\|_F$, and supports weight matrices $W$ so trusted liquid core correlations (like EUR/USD vs GBP/USD) remain anchored while noisy stress overrides absorb the adjustment!

## Comparing Broken-Correlation Repair Algorithms in Production Risk Engines

| Matrix Failure Trigger in Trading Books | Typical $\lambda_{\min}(C)$ | Best Mathematical Remedy | Companion Guide / Tool |
| :--- | :--- | :--- | :--- |
| Asynchronous Tokyo / London / NY closes | $-0.02$ to $-0.08$ | Synchronous 16:00 London fix or Hayashi-Yoshida estimator | [FX Forward Calculator](/calculators/fx-forward-calculator) |
| Pairwise deletion with missing IPO histories | $-0.05$ to $-0.22$ | Factor-model backfill + Higham weighted projection | [Filtered Historical Simulation](/guides/filtered-historical-simulation-garch-standardized-residual-var) |
| High dimension $N = 500 > T = 250$ days | $0.00$ ($250$ zero modes) | Linear or nonlinear eigenvalue shrinkage | [Ledoit-Wolf Covariance Shrinkage](/guides/ledoit-wolf-covariance-shrinkage-high-dimensional-portfolio-var) |
| Manual stress correlation block overrides | $-0.15$ to $-0.65$ | Higham $W$-weighted alternating projections | [Correlation Breakdown Stress Math](/guides/fx-equity-correlation-breakdown-stressed-beta-var-amplification) |

![Monte Carlo 97.5% Expected Shortfall standard error convergence across Crude MC, Antithetic Variates, and Sobol Quasi-MC](/api/figure?type=bars&title=Monte%20Carlo%2097.5%25%20ES%20Estimator%20Standard%20Error%20%28%25%20of%20True%20ES%29&subtitle=Comparing%20simulation%20noise%20at%2010%2C000%20paths%20across%20variance-reduction%20sampling%20schemes.&steps=Crude%20Pseudo-Random%3A%3AMersenne%20Twister%20O%281%2Fsqrt%28M%29%29%20slow%20convergence%7CAntithetic%20%28%2BZ%20%2F%20-Z%29%3A%3ACancels%20odd-moment%20linear%20delta%20sampling%20error%7CImportance%20Sampling%3A%3AShifts%20mean%20toward%20-VaR%20tail%20region%20for%20rare%20events%7CScrambled%20Sobol%20QMC%3A%3ALow-discrepancy%20O%28%28ln%20M%29%5Ed%20%2F%20M%29%20quasi-random%20grid&metrics=CRUDE%2010K%20SE%3A%3A%2B%2F-%203.18%25%3A%3A250%20tail%20paths%20have%20sampling%20jitter%7CANTITHETIC%2010K%20SE%3A%3A%2B%2F-%201.84%25%3A%3AZero-cost%20exact%20zero%20sample%20mean%7CIMPORTANCE%2010K%20SE%3A%3A%2B%2F-%200.62%25%3A%3AGlasserman-Heidelberger-Shah%20shift%7CSOBOL%20%2B%20PCA%2010K%20SE%3A%3A%2B%2F-%200.48%25%3A%3AFeed%20top%20PCA%20factors%20first!)

## Variance Reduction Tip: Why PCA Ordering Beats Cholesky for Sobol Quasi-Monte Carlo

When pairing low-discrepancy **Sobol sequences** with multi-asset Monte Carlo, standard Cholesky assigns Dimension 1 of the Sobol sequence entirely to Asset 1, wasting low-discrepancy uniformity on whichever ticker happens to sit in Row 1! Instead, factorize $C = Q \Lambda^{1/2}$ sorted in descending eigenvalue order ($\lambda_1 \ge \lambda_2 \ge \dots \ge \lambda_N$) so Sobol's cleanest initial dimensions drive the dominant market-wide principal components. Review model risk standards across [UK FCA](/regulator-index/fca) regulated venues like [Axi Trading](/broker-lookup/axi-trading) and explore the [Parametric VaR & Expected Shortfall Topic Hub](/topics/parametric-var-expected-shortfall).

## Frequently asked questions

### Why does pairwise correlation calculation with missing data produce non-positive-definite matrices?

Computing rho_12 over 250 days, rho_13 over 180 days, and rho_23 over 140 days destroys the shared inner-product geometry of complete-case sample covariance matrices, violating triangle/cosine bounds and creating negative eigenvalues.

### What is the difference between Rebonato-Jackel spectral clipping and Higham alternating projections?

Rebonato-Jackel floors negative eigenvalues once and rescales rows to unit length in a single fast pass. Higham adds Dykstra’s correction across alternating spectral and unit-diagonal projections to guarantee the exact minimum Frobenius perturbation.

### How many Monte Carlo paths are needed to estimate 97.5% Expected Shortfall reliably?

At 10,000 crude paths, 97.5% ES averages 250 tail scenarios with roughly +/-3.2% standard error. Combining antithetic variates, Glasserman importance-sampling drift shifts toward the tail, or PCA-ordered Sobol sequences cuts error below +/-0.6%.

### Can Cholesky handle a correlation matrix with an exact zero eigenvalue (positive semi-definite)?

Standard unpivoted Cholesky divides by L_jj and fails if L_jj = 0; pivoted Cholesky (P^T * C * P = L * L^T) or spectral square-root factorization Q * sqrt(max(Lambda, 0)) handles rank-deficient PSD matrices cleanly.


## Source URLs

- https://academic.oup.com/imanum/article-abstract/22/3/329/708688
- https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1969689
- https://www.bis.org/basel_framework/chapter/MAR/33.htm

## Citation

- APA: Elena Rostova, FRM, CQF. (2026). Cholesky Factorization in Monte Carlo VaR: Fixing Broken Correlation Grids. QuantCalc Tools. https://quantcalctools.com/guides/cholesky-covariance-matrix-positive-definiteness-monte-carlo-var
- Canonical article: [Cholesky Factorization in Monte Carlo VaR: Fixing Broken Correlation Grids](https://quantcalctools.com/guides/cholesky-covariance-matrix-positive-definiteness-monte-carlo-var)
- Site research index: [QuantCalc Tools](https://quantcalctools.com/llms-full.txt)
