# Basel III FRTB Expected Shortfall & ISDA SIMM Quantitative Guide 2026

> Quantitative guide to Basel III FRTB Expected Shortfall (97.5% ES), Liquidity Horizons, NMRF stress calibration, SA-CCR, and ISDA SIMM margin engineering.

Canonical: https://quantcalctools.com/guides/basel-iii-endgame-frtb-ima-expected-shortfall-isda-simm-v2-margin-calibration-2026
Last updated: 2026-10-27T03:56:30.301Z

**Publication:** QuantCalc Tools
**Category:** Quantitative Risk & Capital Modeling
**Published:** 2026-10-22T17:13:00.805Z
**Author:** Elena Rostova, FRM, CQF (FRM, Certificate in Quantitative Finance (CQF))
**Reviewer:** Not recorded

## Answer-first summary

Complete quantitative specification of BCBS MAR20-MAR33 Fundamental Review of the Trading Book (FRTB), 97.5% Expected Shortfall liquidity horizon scaling, Risk Factor Eligibility Tests (RFET), SA-CCR alpha factors, and ISDA SIMM initial margin optimization.

## Key takeaways

- FRTB Internal Models Approach (MAR33) replaces 10-day 99% VaR with a liquidity-adjusted 97.5% Expected Shortfall (ES) cascade scaled across 10, 20, 40, 60, and 120 business-day liquidity horizons via marginal square-root-of-time weights (1.00, 1.414, and 2.449).
- Cross-asset correlation benefits inside the Internals Model Capital Charge (IMCC) are restricted by the supervisory parameter rho = 0.50, blending unconstrained empirical ES with the linear sum of five standalone asset-class ES charges (GIR, CSR, EQ, COM, FX).
- Under the Risk Factor Eligibility Test (MAR31.12), every modellable risk factor bucket requires >=24 verifiable real-price observations per year (with >=4 per 90-day window) or >=100 observations in 12 months; failing factors incur an un-diversifiable Non-Modellable Risk Factor (NMRF) Stressed ES charge with rho_NM = 0.60.
- Trading desks must pass quarterly P&L Attribution Tests (PLAT, MAR32.20) requiring Spearman rank correlation r_s > 0.80 and Kolmogorov-Smirnov distance KS < 0.09 to remain in the Green Zone; falling into the Red Zone (r_s < 0.70 or KS > 0.12) forces immediate fallback to the FRTB Standardized Approach.
- ISDA SIMM v2.6/v2.7 computes 99% 10-day MPOR bilateral initial margin across four strictly additive product silos (RatesFX, Credit, Equity, Commodity), creating strategic cross-margin optimization opportunities against SA-CCR (1.4 * (RC + PFE)) and CCP clearing via multilateral compression.

## Analysis

## 1. Executive Summary: The Basel III Endgame & Cross-Margin Paradigm Shift (2025–2026)

The global prudential and margin architecture governing institutional derivatives trading books, central clearing interfaces, and non-centrally cleared bilateral over-the-counter (OTC) portfolios has reached a decisive structural inflection point across the 2025–2026 supervisory cycle. Under the consolidated Basel Committee on Banking Supervision (BCBS) *Minimum Capital Requirements for Market Risk* framework (`MAR10` through `MAR33`) [[1]](https://www.bis.org/basel_framework/standard/MAR.htm) and the Standardized Approach for Counterparty Credit Risk (`CRE52` / SA-CCR) [[2]](https://www.bis.org/basel_framework/standard/CRE.htm), global systemically important banks (G-SIBs), regional dealers, and quantitative hedge funds confront an unprecedented convergence between regulatory market risk capital, counterparty exposure multipliers, and bilateral initial margin (IM) funding obligations.

Historically, trading desk capital allocation under Basel 2.5 relied on a 10-day, 99th percentile Value-at-Risk ($VaR_{99\%,10d}$) combined with an additive Stressed VaR ($SVaR$) and Incremental Risk Charge ($IRC$). That legacy specification exhibited three fatal structural deficiencies during periods of acute market illiquidity:
1. **Inability to Capture Tail Convexity:** As a quantile boundary metric, $VaR_{99\%}$ ignored the magnitude and distribution of losses beyond the 99th percentile threshold, creating perverse incentives to accumulate deep out-of-the-money (OTM) gamma and correlation skew risks.
2. **Uniform 10-Day Liquidation Assumption:** Assuming all trading book exposures could be delta-hedged or unwound within 10 business days without inducing endogenous market impact violated empirical microstructure behavior across credit correlation, emerging market currencies, and bespoke volatility surfaces.
3. **Permeable Diversification Parameters:** Unrestricted empirical correlation matrices permitted banks to offset idiosyncratic risk factors across disparate asset classes, significantly understating aggregate portfolio drawdown during cross-asset correlation regime breaks.

To remediate these vulnerabilities, the Fundamental Review of the Trading Book (FRTB) replaces $VaR_{99\%}$ with a liquidity-adjusted **97.5% Expected Shortfall ($ES_{97.5\%}$)** metric calibrated to the institution's most severe 12-month historical stress window (`MAR33.5`), enforces a five-tier **Liquidity Horizon ($LH_j \in \{10, 20, 40, 60, 120\}$ days)** scaling cascade (`MAR33.12`), and introduces two stringent desk-level gates: the **Profit and Loss Attribution Test (PLAT)** (`MAR32.20`) and the **Risk Factor Eligibility Test (RFET)** (`MAR31.12`). Any risk factor failing to demonstrate at least 24 verifiable real-price observations per year—or 100 observations over a 12-month period—is carved out of the diversifiable Expected Shortfall engine and subjected to an un-diversifiable **Non-Modellable Risk Factor ($NMRF$)** stress scenario capital charge (`MAR33.16`).

Simultaneously, the uncleared derivatives market operates under the **ISDA Standard Initial Margin Model (ISDA SIMM v2.6 / v2.7)** [[6]](https://www.isda.org/margin-reform/), which computes bilateral initial margin using a variance-covariance sensitivity aggregation calibrated to a **99% confidence interval over a 10-day Margin Period of Risk ($MPOR$)**. When combined with the **72.5% Standardized Output Floor** (`RBC20`), **SA-CCR** ($\alpha = 1.4$ supervisory multiplier), and the resulting **Margin Valuation Adjustment ($MVA$)** and **Capital Valuation Adjustment ($KVA$)** funding drags, quantitative risk engineering teams must solve a coupled, multi-regime optimization problem across every execution venue and clearinghouse.

Institutional risk officers evaluating cross-jurisdictional supervisory transposition can audit the primary rulebooks via our [Global Regulatory Directory](/regulator-index), including the [Basel Committee on Banking Supervision (BCBS)](/regulator-index/bcbs-intl), the [Federal Reserve Board (FRB)](/regulator-index/fed-us), the [European Banking Authority (EBA)](/regulator-index/eba-eu), the [UK Financial Conduct Authority & PRA](/regulator-index/fca-gb), the [Swiss Financial Market Supervisory Authority (FINMA)](/regulator-index/finma-ch), and the [Monetary Authority of Singapore (MAS)](/regulator-index/mas-sg), alongside our institutional execution counterparty benchmarks in the [Prime & Multi-Asset Broker Comparison Matrix](/brokers).

---

## 2. Mathematical Architecture of FRTB Internal Models Approach (IMA) & 97.5% Expected Shortfall (`MAR33`)

![Figure 1: BCBS FRTB Internal Models Approach (IMA) & 97.5% Expected Shortfall Liquidity Horizon Cascade](/api/figure?title=BCBS%20FRTB%2097.5%25%20Expected%20Shortfall%20Architecture&subtitle=Liquidity-Adjusted%20ES%20%2810d%20to%20120d%29%2C%20RFET%20%26%20NMRF%20Capital%20Aggregation&badge=EXHIBIT%2001&steps=RFET%2024-Obs%20Gate|Liquidity%20Horizon%20Cascade|Stressed%20ES%20Calibration|DRC%20%2B%20NMRF%20Capital%20Charge&metrics=ES%20Confidence%20Level:97.5%25%20Tail%20Mean|Liquidity%20Buckets:10d%20%2F%2020d%20%2F%2040d%20%2F%2060d%20%2F%20120d|P%26L%20Attribution%20Spearman:%3E0.80%20Green%20Zone|Output%20Floor%20Phase-In:72.5%25%20of%20SA%20RWA&source=BCBS%20MAR30-MAR33%20%2F%20Federal%20Reserve%20%2F%20EBA%202025-2026)

### 2.1 Continuous-Time Tail Conditional Expectation vs. Value-at-Risk

Under `MAR33.2`, Expected Shortfall at confidence level $\alpha = 0.975$ is defined mathematically as the conditional expectation of portfolio profit-and-loss (P&L) distribution $L$ exceeding the $\alpha$-quantile threshold $VaR_{\alpha}(L)$:

$$ES_{\alpha}(L) = \frac{1}{1 - \alpha} \int_{\alpha}^{1} q_u(L)\,du = \mathbb{E}\left[ L \,\middle|\, L \ge VaR_{\alpha}(L) \right]$$

For a Gaussian reference distribution $\mathcal{N}(\mu, \sigma^2)$, the analytical ratio of the standard normal probability density function $\phi(\Phi^{-1}(0.975))$ to the tail probability $(1 - 0.975 = 0.025)$ yields a tail multiplier of $2.3378\sigma$, closely matching the $99\%$ normal VaR quantile $\Phi^{-1}(0.99) = 2.3263\sigma$. However, for leptokurtic trading book distributions exhibiting fat tails (Student-$t$ with $\nu \le 5$ degrees of freedom or jump-diffusion dynamics), $ES_{97.5\%}$ produces capital charges **18% to 42% higher** than $VaR_{99\%}$, while satisfying the Artzner-Delbaen-Eber-Heath **sub-additivity axiom** ($ES(A + B) \le ES(A) + ES(B)$) required for coherent convex optimization.

### 2.2 The Five-Tier Liquidity Horizon Cascade (`MAR33.4` – `MAR33.12`)

Rather than computing separate multi-month historical simulations—which suffer from overlapping sample autocorrelation and sparse data degrees of freedom—BCBS `MAR33.4` mandates that banks compute base Expected Shortfall shocks over a $LH_1 = 10\text{-day}$ base horizon and scale nested subsets of risk factors $Q(RF, j)$ across five discrete supervisory liquidity horizons: $LH_j \in \{10, 20, 40, 60, 120\}$ business days.

Let $ES_{t}(P, j)$ denote the 10-day Expected Shortfall calculated when holding all risk factors with assigned liquidity horizon $LH_k < LH_j$ constant, shocking only the active risk factor subset $Q(RF, j) = \{RF_k \mid LH_k \ge LH_j\}$. The marginal incremental liquidity-horizon contribution is scaled by the square-root-of-time differential and aggregated via a quadratic root-sum-of-squares cascade:

```formula
ES_LH(P) = sqrt( (ES_10d(P,1))^2 + sum_{j=2}^{5} ( ES_10d(P,j) * sqrt( (LH_j - LH_{j-1}) / 10 ) )^2 ) ; IMCC = rho * IMCC(C) + (1 - rho) * sum_{B=1}^{5} IMCC(B)
- ES_10d(P, j): 10-business-day Expected Shortfall at 97.5% confidence shocking only risk factors in subset Q(RF, j) with supervisory liquidity horizon LH >= LH_j (where LH_j in {10, 20, 40, 60, 120} days).
- sqrt((LH_j - LH_{j-1}) / 10): Marginal square-root-of-time scaling weights equal to 1.000 (j=2, 10d->20d), sqrt(2)=1.4142 (j=3, 20d->40d), sqrt(2)=1.4142 (j=4, 40d->60d), and sqrt(6)=2.4495 (j=5, 60d->120d).
- IMCC(C) and IMCC(B): Unconstrained cross-asset Expected Shortfall charge vs asset-class-constrained Expected Shortfall charges across the 5 broad risk classes (GIR, CSR, EQ, COM, FX) weighted by supervisory correlation parameter rho = 0.50.
```

### 2.3 Reduced-Set Stressed Calibration & Cross-Asset Constraint ($\rho = 0.50$)

Because full-revaluation simulation across all five liquidity subsets for every historical 12-month window since 2007 is computationally prohibitive, `MAR33.5` permits an Indirect Stressed ES Approximation using a **Reduced Set of Risk Factors ($R$)** alongside the **Full Set ($F$)**:

$$ES = ES_{R,S} \times \max\left(1.0,\, \frac{ES_{F,C}}{ES_{R,C}}\right)$$

Here $ES_{R,S}$ is calibrated over the most severe 12-month **Stressed Window ($S$)** (typically 2008–2009 or March 2020), while $ES_{F,C}$ and $ES_{R,C}$ are calibrated over the **Current 12-month period ($C$)**. By statute (`MAR33.6`), the reduced set $R$ must explain at least **75%** of the variance of the full Expected Shortfall ($ES_{R,C} / ES_{F,C} \ge 0.75$). Furthermore, under `MAR33.13`, the Internals Model Capital Charge ($IMCC$) blends the unconstrained empirical portfolio charge $IMCC(C)$ with the linear sum of standalone asset-class charges $\sum_{B=1}^{5} IMCC(B)$ using $\rho = 0.50$, capping cross-asset diversification at 50%.

| Risk Factor Class / Bucket | Assigned Liquidity Horizon (Days) | Marginal Scaling Weight | Minimum RFET Real-Price Obs / Yr | Max Diversification Benefit (%) | Typical Stressed ES Scaling Multiplier |
| :--- | :---: | :---: | :---: | :---: | :---: |
| **G10 Sovereign Interest Rate Curves & OIS (EUR, USD, GBP, JPY)** | 10 | 1.00 | 24 | 50.0% | 1.18 |
| **Investment Grade (IG) Corporate Credit Spreads & Equity Large-Cap Spot** | 20 | 1.00 | 24 | 50.0% | 1.42 |
| **G10 Interest Rate Volatility, FX Volatility & Equity Small-Cap / Repo** | 40 | 1.41 | 24 | 50.0% | 1.76 |
| **High Yield (HY) Credit Spreads, Sovereign EM Credit & Equity/FX Exotics** | 60 | 1.41 | 24 | 50.0% | 2.15 |
| **Credit Spread Volatility, Base Correlation & Bespoke Commodity Skew** | 120 | 2.45 | 24 | 50.0% | 2.94 |

---

## 3. Desk-Level Model Eligibility Gates: RFET, NMRF & P&L Attribution (PLAT)

Under the Basel III Endgame framework (`MAR30` – `MAR32`), internal models approval is granted, monitored, and revoked at the granular **Trading Desk level** [[1]](https://www.bis.org/basel_framework/standard/MAR.htm). Every trading desk must continuously pass two quantitative verification regimes to avoid mandatory fallback to the Standardized Approach (SA).

```flow
Stage 1: Risk Factor Eligibility Test (RFET) | Audit committed trades, firm executable two-way dealer quotes, and third-party vendor pools (Tradeweb, Bloomberg, ICE) to verify >=24 real price observations per 12-month window with no 90-day gap (<4 obs/quarter failure) or >=100 real observations in 12 months (`MAR31.12`).
Stage 2: Non-Modellable Risk Factor (NMRF) Carve-Out | Isolate any risk factor failing RFET or lacking buckling interpolation validity; compute standalone LH_j Stressed Scenario Loss (SES) calibrated to the most extreme 12-month tail and aggregate with rho_NM = 0.60 constraint (`MAR33.16`).
Stage 3: P&L Attribution Test (PLAT) Dual-Metric Audit | Compare Hypothetical P&L (HPL, stripping intraday trades and fees) against Risk-Theoretical P&L (RTPL, pricing via risk engine) over 250 rolling days using Spearman Rank Correlation (r_s) and Kolmogorov-Smirnov (KS) distributional distance (`MAR32.20`).
Stage 4: Traffic-Light Surcharge & Total IMA Aggregation | Assign Green (r_s > 0.80, KS < 0.09), Amber (0.70 <= r_s <= 0.80 or 0.09 <= KS <= 0.12 with k_amber surcharge), or Red (mandatory 100% Standardized Approach fallback) and sum IMCC + SES(NMRF) + DRC (`MAR33.41`).
```

### 3.1 The Risk Factor Eligibility Test (RFET) & Non-Modellable Risk Factors (`MAR31.12` – `MAR33.21`)

A risk factor mapped to a desk's Risk-Theoretical P&L ($RTPL$) is classified as **modellable** only if it passes one of two quantitative verifiability criteria over the preceding 12-month period:
1. **Criterion A (24-Observation Standard):** At least **24 verifiable real prices** per year, with no 90-day period containing fewer than **4 verifiable real prices** (operationalized under EBA RTS [[4]](https://www.eba.europa.eu/regulation-and-policy/market-risk) as no more than one 30-day gap).
2. **Criterion B (100-Observation High-Frequency Standard):** At least **100 verifiable real prices** over the preceding 12-month period.

A **verifiable real price** must originate from an executed transaction or a committed, firm executable two-way quote from an arm's-length market maker; indicative consensus marks without trade tickets do not qualify. When a risk factor fails RFET, it is expelled from $IMCC$ and subjected to a standalone **Stressed Expected Shortfall ($SES$)** scenario calibrated over the most severe 12-month stress window (`MAR33.17`) at a liquidity horizon of $\max(LH_{\text{assigned}}, 20\text{ days})$:

$$SES_{\text{Total}} = \sqrt{\sum_{i=1}^{I} \left(ISES_{NM,i}\right)^2} + \sqrt{\sum_{j=1}^{J} \left(ISES_{NM,j}\right)^2} + \sqrt{ \left( \rho_{NM} \sum_{k=1}^{K} SES_{NM,k} \right)^2 + \left(1 - \rho_{NM}^2\right) \sum_{k=1}^{K} \left(SES_{NM,k}\right)^2 }$$

Under BCBS `MAR33.19`, the correlation parameter for general non-modellable risk factors is fixed at **$\rho_{NM} = 0.60$**, meaning **60% of all general NMRF charges add linearly**. In dealer books, NMRFs regularly account for **30% to 55% of total IMA capital** on exotic rates, inflation, and correlation desks.

### 3.2 The P&L Attribution Test (PLAT): Spearman Rank Correlation & Kolmogorov-Smirnov (`MAR32.20` – `MAR32.43`)

Even if 100% of a desk's risk factors pass RFET, the desk's risk engine ($RTPL$) must accurately reproduce the front-office pricing engine's Hypothetical P&L ($HPL$) across a rolling 250-trading-day sample (`MAR32.24`):

$$r_s = \frac{\text{Cov}\big(R(HPL), R(RTPL)\big)}{\sigma_{R(HPL)} \cdot \sigma_{R(RTPL)}}, \qquad KS = \sup_{x \in \mathbb{R}} \left| F_{HPL}(x) - F_{RTPL}(x) \right|$$

Under `MAR32.39`, trading desks are classified quarterly into three supervisory zones:
- **Green Zone ($r_s > 0.80$ and $KS < 0.09$):** Full IMA capital eligibility with zero capital surcharge.
- **Amber Zone ($0.70 \le r_s \le 0.80$ or $0.09 \le KS \le 0.12$, and neither metric in Red):** The desk retains IMA eligibility but incurs a mandatory **Amber Zone Capital Surcharge** proportional to the standardized capital gap: $\text{Surcharge}_{\text{Amber}} = k \times \max(0, SA_{\text{all}} - IMA_{\text{all}})$, where $k = 0.50 \times \sum_{i \in \text{Amber}} SA_i / \sum_{j \in \text{All IMA}} SA_j$.
- **Red Zone ($r_s < 0.70$ or $KS > 0.12$):** Immediate disqualification from IMA; the desk must compute 100% of its market risk capital under the FRTB Standardized Approach ($SA$) until it passes Green Zone thresholds and backtesting limits (`MAR32.42`).

---

## 4. ISDA SIMM v2.6 / v2.7 Initial Margin Calibration vs. FRTB Sensitivities-Based Method (SbM)

![Figure 2: ISDA SIMM v2.6+ 99% 10-Day Initial Margin vs SA-CCR Alpha Exposure and MVA Funding Drag](/api/figure?title=ISDA%20SIMM%20v2.6%2B%20vs%20SA-CCR%20%26%20FRTB-SbM%20Calibration&subtitle=99%25%2010-Day%20MPOR%20Sensitivity%20Margin%2C%20Concentration%20Thresholds%20%26%20MVA%20Cost&badge=EXHIBIT%2002&steps=Delta%20Vega%20Curvature%20Sensitivities|Concentration%20Threshold%20Scaling|Intra-Bucket%20%26%20Cross-Bucket%20Aggregation|SA-CCR%20Alpha%201.4x%20%26%20MVA%20Optimization&metrics=SIMM%20Confidence%20Horizon:99.0%25%2010-Day%20MPOR|SA-CCR%20Alpha%20Factor:1.40x%20%28RC%20%2B%20PFE%29|SIMM%20Recalibration%20Cycle:Annual%20%2B%20Semi-Annual%20Audit|Cross-Class%20Aggregation:Product%20Silos%20%28RatesFX%20%2F%20Credit%20%2F%20Eq%20%2F%20Com%29&source=ISDA%20SIMM%20v2.6-v2.7%20Methodology%20%2F%20BCBS-IOSCO%20UMR%20%2F%20CFTC%20Part%2023)

While FRTB IMA governs regulatory capital for approved trading desks, virtually all banks (via the **72.5% Output Floor** and fallback desks) and all bilateral OTC derivatives counterparties subject to BCBS-IOSCO Uncleared Margin Rules (UMR Phases 1–6) [[8]](https://www.bis.org/bcbs/publ/d499.htm) must simultaneously compute variance-covariance sensitivity aggregations under the **FRTB Sensitivities-Based Method (SbM)** (`MAR21`) and **ISDA SIMM** [[6]](https://www.isda.org/margin-reform/).

### 4.1 Structural Isomorphism and Parametric Divergence: FRTB-SbM vs. ISDA SIMM

Both FRTB-SbM and ISDA SIMM decompose portfolio risk into **Delta ($\Delta$)**, **Vega ($\mathcal{V}$)**, and **Curvature ($CVR$)** across six asset classes. Within each risk bucket $b$, weighted sensitivities $WS_k = RW_k \cdot s_k \cdot CR_k$ are aggregated via an intra-bucket correlation matrix $\rho_{kl}$ and combined across buckets using an inter-bucket correlation matrix $\gamma_{bc}$:

$$K_b = \sqrt{\max\left(0,\, \sum_{k} WS_k^2 + \sum_{k} \sum_{l \neq k} \rho_{kl} WS_k WS_l\right)}, \qquad \text{Capital}_{\text{Class}} = \sqrt{\sum_{b} K_b^2 + \sum_{b} \sum_{c \neq b} \gamma_{bc} S_b S_c} + K_{\text{residual}}$$

Despite sharing this algebraic backbone, quantitative engineers must account for four fundamental calibration divergences between ISDA SIMM v2.6/v2.7 and FRTB-SbM:

1. **99% 10-Day VaR (`SIMM`) vs. 97.5% Variable-Horizon Stressed ES (`FRTB-SbM`):** ISDA SIMM is calibrated to a **99% Value-at-Risk ($VaR_{99\%}$) over a uniform 10-day Margin Period of Risk ($MPOR$)** using a 3+1 year historical lookback (3 recent rolling years plus 1 stress year). Conversely, FRTB-SbM risk weights (`MAR21.39` – `MAR21.95`) replicate **97.5% Stressed Expected Shortfall across variable liquidity horizons (10d to 120d)**, making FRTB-SbM credit and equity risk weights **1.4x to 2.8x larger** than ISDA SIMM weights.
2. **Product Class Silos (`SIMM`) vs. Portfolio-Wide Risk Class Netting (`FRTB-SbM`):** Under BCBS-IOSCO UMR rules [[8]](https://www.bis.org/bcbs/publ/d499.htm), ISDA SIMM partitions every trade into one of **four mutually exclusive Product Silos**: (1) Interest Rates & FX, (2) Credit, (3) Equity, and (4) Commodity. Total SIMM initial margin is the **strict linear sum** across silos ($IM_{\text{SIMM}} = IM_{\text{RatesFX}} + IM_{\text{Credit}} + IM_{\text{Equity}} + IM_{\text{Commodity}}$). Thus, an interest-rate hedge in the RatesFX silo cannot offset interest-rate delta embedded in an equity total return swap (Equity silo) under ISDA SIMM, whereas FRTB-SbM nets all interest rate sensitivities portfolio-wide.
3. **Concentration Thresholds ($T_b$):** ISDA SIMM applies explicit dollar Concentration Thresholds ($T_b$) via $CR_b = \max\left(1.0, \sqrt{|\sum_{k \in b} s_k| / T_b}\right)$, scaling both effective risk weights and cross-bucket correlation penalties when a counterparty's net DV01 or vega exceeds $T_b$.
4. **Three-Scenario Correlation Stress (FRTB-SbM Only):** FRTB-SbM computes capital across **Medium (Base)**, **High ($1.25 \times \rho_{kl}$, capped at $1.00$)**, and **Low ($\max(2 \times \rho_{kl} - 1,\, 0.75 \times \rho_{kl})$)** correlation matrices (`MAR21.6`), taking the maximum charge across the three scenarios.

| Asset Class & Calibration Bucket | FRTB-SbM Risk Weight (bps or %) | ISDA SIMM v2.6+ Risk Weight (bps or %) | SIMM Concentration Threshold ($M/bp or $M) | Intra-Bucket Correlation ($\rho_{kl}$ %) | Cross-Bucket Correlation ($\gamma_{bc}$ %) |
| :--- | :---: | :---: | :---: | :---: | :---: |
| **G10 Major Interest Rates (USD/EUR/GBP 5Y–10Y Vertex)** | 106.0 bps | 61.0 bps | 310.0 | 94.0% | 50.0% |
| **Secondary / Emerging Market Interest Rates (10Y Vertex)** | 150.0 bps | 108.0 bps | 37.0 | 82.0% | 27.0% |
| **Corporate Credit Qualifying IG (Financials & Industrials)** | 100.0 bps | 59.0 bps | 0.95 | 65.0% | 42.0% |
| **Corporate Credit High Yield (HY) & Non-Rated** | 600.0 bps | 210.0 bps | 0.22 | 45.0% | 28.0% |
| **Large-Cap Developed Market Equities (Tech / Financials)** | 55.0% | 25.0% | 3400.0 | 25.0% | 15.0% |
| **FX Spot Delta (Major G10 Specified Currency Pairs)** | 10.6% | 7.4% | 8200.0 | 60.0% | 60.0% |

---

## 5. Counterparty Credit Risk (SA-CCR `CRE52`), CVA (`MAR50`), and the Cross-Margin Optimization Frontier

A trading desk's return on equity ($RoE$) is governed by the coupled interaction of **FRTB Market Risk RWA**, **SA-CCR Counterparty Credit Risk RWA**, **Credit Valuation Adjustment (CVA) RWA**, and the funding cost of posting initial margin (**MVA**).

### 5.1 SA-CCR Exposure at Default (`CRE52`) and the $\alpha = 1.4$ Multiplier

Under `CRE52.1` [[2]](https://www.bis.org/basel_framework/standard/CRE.htm), Exposure at Default ($EAD$) for derivatives netting sets is defined as:

$$EAD_{\text{SA-CCR}} = \alpha \times (RC + PFE), \quad \alpha = 1.40, \quad PFE = m \times \text{AddOn}_{\text{agg}}$$

For margined netting sets, Replacement Cost is $RC = \max(V - C,\, TH + MTA - NICA,\, 0)$, and the PFE multiplier $m = \min\left(1, 0.05 + 0.95 \exp\left(\frac{V - C}{1.90 \times \text{AddOn}_{\text{agg}}}\right)\right)$ dampens exposure when initial margin over-collateralizes the netting set ($V - C < 0$). Because **segregated bilateral initial margin posted under UMR cannot reduce $NICA$ for the posting party** (as it is held in a bankruptcy-remote third-party custodian account under CFTC Part 23 [[7]](https://www.cftc.gov/LawRegulation/FederalRegister/index.htm)), posting $\$1\text{B}$ of ISDA SIMM initial margin suppresses the counterparty's $PFE$ multiplier toward the $5\%$ floor for the receiver, yet imposes a permanent funding drag ($MVA = \int_0^T s_F(t) \mathbb{E}^{\mathbb{Q}}[D(0,t) IM_{\text{SIMM}}(t)] dt$) on the poster.

### 5.2 Cleared CCP vs. Bilateral UMR vs. Multilateral Compression Trade-Offs

Quantitative XVA and capital optimization desks continuously re-balance exposures across three structural venues:
1. **Central Counterparty (CCP) Clearing (LCH SwapClear, CME, Eurex):** Carries a preferential **2% Qualifying CCP ($QCCP$) risk weight** (`CRE54.14`) and a 5-day MPOR ($MPOR_{\text{CCP}} = 5\text{d}$ vs. $10\text{d}$ bilateral), lowering initial margin by roughly $1 - \sqrt{5/10} \approx 29.3\%$. However, CCP initial margin must be funded in cash or sovereign collateral, and directional books lose cross-asset netting against bespoke OTC structures remaining in bilateral CSA silos.
2. **Bilateral Uncleared SIMM Netting Sets:** Subject to $10\text{-day}$ MPOR and counterparty credit risk weights ranging from $20\%$ to $100\%$ under Standardized Credit Risk (`CRE20`), plus BA-CVA or SA-CVA capital (`MAR50`).
3. **Multilateral Risk Compression & SIMM Rebalancing Cycles (OSTTRA TriOptima / Quantile):** By injecting zero-NPV market-neutral basis swaps, FX forwards, or swaption straddles between pairs of dealers, compression engines transfer delta and vega sensitivities from counterparty pairs where **SIMM concentration thresholds ($T_b$)** are breached to pairs with offsetting exposures—routinely reducing aggregate bilateral SIMM IM by **25% to 45%** and SA-CCR $\text{AddOn}_{\text{agg}}$ by **30% to 50%** without altering front-office economic positioning.

---

## 6. Jurisdictional Divergence Matrix: US Basel III Endgame vs. EU CRR III vs. UK PRA PS9/24 vs. APAC (2025–2026)

A critical operational challenge for global banking groups in 2026 is the **fragmented timeline and supervisory calibration** across the four major regulatory blocs.

### 6.1 United States (`12 CFR Part 217`), European Union (`CRR III Art. 461a`), UK (`PS9/24`) & APAC

Following extensive industry feedback on the July 2023 Notice of Proposed Rulemaking (NPR) [[3]](https://www.federalreserve.gov/supervisionreg/basel-endgame.htm), the [Federal Reserve Board](/regulator-index/fed-us) outlined structural revisions leading into the 2026 final rulemaking cycle: Category III/IV regional banks ($\$$100B–$\$$700B assets) are largely exempted from internal model overhauls, while US G-SIBs face a recalibrated FRTB implementation targeting an aggregate CET1 increase of roughly **9%** (down from **19%** in the 2023 NPR), alongside targeted SA-CCR client-clearing relief and broader vendor data pool recognition for RFET.

In the European Union, while CRR III (`Regulation (EU) 2024/1623`) entered into force on **January 1, 2025** for credit risk, SA-CCR, CVA, and the **50% Output Floor phase-in** (reaching **72.5% by 2030**), the European Commission invoked **Article 461a of the CRR** [[4]](https://www.eba.europa.eu/regulation-and-policy/market-risk) to postpone binding Pillar 1 own-funds FRTB application until **2026–2027** to preserve transatlantic parity, while maintaining mandatory quarterly **COREP FRTB-SA/IMA** reporting supervised by the [EBA](/regulator-index/eba-eu). Similarly, the UK Prudential Regulation Authority (PRA) published `PS9/24` [[5]](https://www.bankofengland.co.uk/prudential-regulation/publication/2024/september/implementation-of-the-basel-3-1-standards-near-final-policy-statement-part-2) (detailed in our [UK FCA & PRA Profile](/regulator-index/fca-gb)), aligning Basel 3.1 go-live to **January 1, 2027** with bespoke Default Risk Charge ($DRC$) sovereign adjustments. Conversely, [FINMA in Switzerland](/regulator-index/finma-ch) (`ERV`), the [Monetary Authority of Singapore (MAS)](/regulator-index/mas-sg) (`MAS Notice 637`), and the HKMA brought final FRTB Pillar 1 rules into force across **2024–2025**.

| Regulatory Jurisdiction & Supervisory Authority | Binding Pillar 1 FRTB Effective Year | Output Floor Initial Phase-In (%) | Terminal Output Floor Target (%) | SA-CCR Alpha Factor (Standard) | Avg Est. Trading Book RWA Uplift (%) |
| :--- | :---: | :---: | :---: | :---: | :---: |
| **BCBS Global Baseline Standard (MAR10–MAR50)** | 2025 | 50.0% | 72.5% | 1.40 | 51.0% |
| **United States (FRB / OCC / FDIC Re-Proposal)** | 2026 | 60.0% | 72.5% | 1.40 | 38.0% |
| **European Union (EBA / CRR III Art. 461a)** | 2026 | 50.0% | 72.5% | 1.40 | 44.0% |
| **United Kingdom (Bank of England PRA PS9/24)** | 2027 | 55.0% | 72.5% | 1.40 | 39.0% |
| **Switzerland (FINMA Revised CAO / ERV)** | 2025 | 50.0% | 72.5% | 1.40 | 46.0% |
| **Singapore & Hong Kong (MAS Notice 637 / HKMA)** | 2025 | 50.0% | 72.5% | 1.40 | 42.0% |

---

## 7. Production Risk Engine Implementation: Adjoint Algorithmic Differentiation (AAD) & Capital Allocation

### 7.1 Adjoint Algorithmic Differentiation (AAD) for Real-Time SIMM & PLAT

Under FRTB PLAT (`MAR32.20`), any discrepancy between the front-office pricing library ($HPL$) and the risk engine's Taylor-expansion or grid-interpolation approximation ($RTPL$) drives the Kolmogorov-Smirnov ($KS$) metric above the **0.09 Green Zone ceiling** or the **0.12 Red Zone cliff**. Simultaneously, ISDA SIMM requires exact first-order Delta ($\partial V / \partial r_i$), Vega ($\partial V / \partial \sigma_k$), and second-order cross-gamma sensitivities across thousands of vertices within intraday pre-trade check latencies ($<50\text{ ms}$).

Replacing finite-difference ("bump-and-revalue") Greek calculations—which scale linearly $\mathcal{O}(N)$ with risk factor count $N$—with **reverse-mode Adjoint Algorithmic Differentiation (AAD)** computes the complete gradient vector $\nabla V = (\partial V / \partial x_1, \dots, \partial V / \partial x_N)$ in a **single backward pass** bounded by $\le 4 \times$ a single forward valuation ($\mathcal{O}(1)$ complexity relative to $N$). Embedding AAD directly into front-office C++ / CUDA Monte Carlo pricing kernels enables unified front-to-back valuation ($HPL \equiv RTPL$, securing $r_s > 0.94$ and $KS < 0.04$) and real-time pre-trade $\Delta IM_{\text{SIMM}}$, $\Delta EAD_{\text{SA-CCR}}$, and $\Delta KVA$ routing.

### 7.2 Total IMA Desk Capital Aggregation & Euler Allocation (`MAR33.41` – `MAR33.43`)

For trading desks maintaining Green or Amber PLAT status and passing backtesting (`MAR32.5`), aggregate Internal Models Approach capital ($C_{\text{IMA}}$) is assembled via `MAR33.43`:

$$C_{\text{IMA}} = \max\left(IMCC_{t-1} + SES_{t-1},\, m_c \cdot \overline{IMCC}_{60d} + \overline{SES}_{60d}\right) + DRC + \text{Surcharge}_{\text{Amber}}$$

where $m_c \ge 1.50$ is the supervisory multiplier, $\overline{IMCC}_{60d}$ and $\overline{SES}_{60d}$ are 60-day moving averages, and $DRC$ is the **Default Risk Charge (`MAR33.18`)** computed via a two-factor Gaussian copula Monte Carlo simulation at **99.9% 1-year confidence** with a **0.03% (3 bps) PD floor**. Because $ES_{97.5\%}$ is homogeneous of degree one, desks allocate $IMCC$ down to individual trades using **Euler's Homogeneous Function Theorem**: $\text{Euler } ES_i = \mathbb{E}[L_i \mid L_P \ge VaR_{97.5\%}(L_P)]$. For execution teams comparing prime brokerage margin models against dealer balance sheet costs, consult our [Institutional Broker & Clearing Comparison Hub](/brokers) and [Global Regulatory Directory](/regulator-index).

---

## Primary Regulatory Filings, Statutory Citations & Empirical References

- **[1]** Basel Committee on Banking Supervision (BCBS), *Minimum Capital Requirements for Market Risk (Consolidated Basel Framework MAR10–MAR50)*, Bank for International Settlements (BIS): [https://www.bis.org/basel_framework/standard/MAR.htm](https://www.bis.org/basel_framework/standard/MAR.htm)
- **[2]** Basel Committee on Banking Supervision (BCBS), *Standardised Approach for Counterparty Credit Risk (SA-CCR, CRE52)*, Bank for International Settlements (BIS): [https://www.bis.org/basel_framework/standard/CRE.htm](https://www.bis.org/basel_framework/standard/CRE.htm)
- **[3]** Board of Governors of the Federal Reserve System, FDIC & OCC, *Regulatory Capital Rule: Large Banking Organizations and Banking Organizations with Significant Trading Activity (Basel III Endgame)*: [https://www.federalreserve.gov/supervisionreg/basel-endgame.htm](https://www.federalreserve.gov/supervisionreg/basel-endgame.htm)
- **[4]** European Banking Authority (EBA), *Regulatory Technical Standards (RTS) on FRTB Risk Factor Eligibility Test (RFET), P&L Attribution (PLAT), and CRR III Market Risk Implementation*: [https://www.eba.europa.eu/regulation-and-policy/market-risk](https://www.eba.europa.eu/regulation-and-policy/market-risk)
- **[5]** Bank of England Prudential Regulation Authority (PRA), *PS9/24 — Implementation of the Basel 3.1 Standards Near-Final Policy Statement (Market Risk & Output Floor)*: [https://www.bankofengland.co.uk/prudential-regulation/publication/2024/september/implementation-of-the-basel-3-1-standards-near-final-policy-statement-part-2](https://www.bankofengland.co.uk/prudential-regulation/publication/2024/september/implementation-of-the-basel-3-1-standards-near-final-policy-statement-part-2)
- **[6]** International Swaps and Derivatives Association (ISDA), *ISDA SIMM Methodology v2.6 / v2.7 & Uncleared Margin Reform Governance Framework*: [https://www.isda.org/margin-reform/](https://www.isda.org/margin-reform/)
- **[7]** Commodity Futures Trading Commission (CFTC), *17 CFR Part 23 — Margin Requirements for Uncleared Swaps for Swap Dealers and Major Swap Participants*: [https://www.cftc.gov/LawRegulation/FederalRegister/index.htm](https://www.cftc.gov/LawRegulation/FederalRegister/index.htm)
- **[8]** Basel Committee on Banking Supervision & IOSCO, *Margin Requirements for Non-Centrally Cleared Derivatives (Consolidated Framework)*: [https://www.bis.org/bcbs/publ/d499.htm](https://www.bis.org/bcbs/publ/d499.htm)

---

### Cross-Desk Institutional Companion Research (October 2026)

For institutional risk committees evaluating cross-regional equity contagion and derivative volatility surface convexity alongside this dossier, compare the primary empirical findings from our specialist desks:

- **Asia-Pacific Equity & Semiconductor Capex Desk (APAC Capital Brief)**: *APAC Equities & Semiconductor Capex Repricing (October 2026): Nikkei 225 Breaches 200-Day Moving Average, KOSPI -1.7% HBM Unwind, and USD/JPY Sub-156.68 Earnings Compression* and *KOSPI Closes at 5,550.30 After Opening 2.36% Lower*.
- **Quantitative Options & Volatility Surface Desk (VolSurface Desk)**: *The October 2026 Volatility Surface Dislocation: VIX/SPX 20x Convexity Squeeze, VIX1D Surges Above 23, Dealer Negative Gamma, and Single-Stock Dispersion Unwinds* and *Four Dollars of Disagreement in Brent, and Gold Back Under $4,000*.

---

> **Flagship Cornerstone Authority Reference**: For the master institutional framework, 2026 benchmark matrix, and mathematical calibration governing this domain, consult our flagship pillar guide: **[Quantitative Options Margin & Kelly Criterion Calculator (QUANT-MARGIN-KELLY)](/guides/options-margin-kelly-criterion-calculator-guide)** and the canonical entity dossier **[Options Margin & Kelly Criterion Calculator](/entities/options-margin-kelly-criterion-calculator-guide)**.

## Frequently asked questions

### How does the BCBS MAR33 Liquidity-Adjusted 97.5% Expected Shortfall formula avoid running separate 120-day historical simulations?

Under MAR33.4, banks compute all Expected Shortfall shocks over a uniform 10-business-day base horizon (ES_10d) and partition risk factors into five nested subsets Q(RF, j) corresponding to supervisory liquidity horizons LH_j in {10, 20, 40, 60, 120} days. For each horizon j >= 2, risk factors with shorter liquidity horizons are held constant while longer-horizon risk factors are shocked over 10 days, scaled by sqrt((LH_j - LH_{j-1}) / 10), and aggregated via root-sum-of-squares.

### What triggers a Non-Modellable Risk Factor (NMRF) capital charge and why is it punitive for exotic derivatives desks?

A risk factor becomes an NMRF if it fails the Risk Factor Eligibility Test (RFET) under MAR31.12—requiring at least 24 committed trades or firm executable two-way quotes per year (with at least 4 per 90-day period) or 100 observations per year—or if it represents uncaptured interpolation basis between observable curve buckets. NMRFs are carved out of the diversifiable ES engine and stressed over a 12-month worst-case window with a 0.60 linear correlation constraint, often accounting for 30% to 55% of exotic desk capital.

### Why can't an equity index total return swap offset the interest rate delta of an interest rate swap under ISDA SIMM v2.6+?

Under BCBS-IOSCO Uncleared Margin Rules, ISDA SIMM partitions all trades into four mutually exclusive Product Silos: RatesFX, Credit, Equity, and Commodity. While sensitivities net across risk classes within a single product silo, total SIMM Initial Margin is the strict linear sum of the four product silos. Therefore, interest rate delta inside an Equity silo trade cannot offset interest rate delta inside a RatesFX silo trade unless the exposures are restructured or cleared.

### How does posting segregated ISDA SIMM Initial Margin impact SA-CCR Exposure at Default (EAD) and Margin Valuation Adjustment (MVA)?

Under SA-CCR (CRE52), EAD equals 1.4 * (RC + PFE). For the party receiving initial margin, segregated collateral reduces the PFE multiplier toward the 5% floor. However, for the posting party, bankruptcy-remote third-party custodial initial margin does not reduce Net Independent Collateral Amount (NICA) against Replacement Cost and simultaneously generates a permanent Margin Valuation Adjustment (MVA) funding drag over the life of the netting set.


## Source URLs

- https://www.bis.org/basel_framework/standard/MAR.htm
- https://www.bis.org/basel_framework/standard/CRE.htm
- https://www.federalreserve.gov/supervisionreg/basel-endgame.htm
- https://www.eba.europa.eu/regulation-and-policy/market-risk
- https://www.bankofengland.co.uk/prudential-regulation/publication/2024/september/implementation-of-the-basel-3-1-standards-near-final-policy-statement-part-2
- https://www.isda.org/margin-reform/
- https://www.cftc.gov/LawRegulation/FederalRegister/index.htm
- https://www.bis.org/bcbs/publ/d499.htm

## Citation

- APA: Elena Rostova, FRM, CQF. (2026). Basel III Endgame FRTB & ISDA SIMM v2.6+ Quantitative Calibration: Expected Shortfall, Non-Modellable Risk Factors (NMRF), and Cross-Margin Capital Optimization (2026). QuantCalc Tools. https://quantcalctools.com/guides/basel-iii-endgame-frtb-ima-expected-shortfall-isda-simm-v2-margin-calibration-2026
- Canonical article: [Basel III Endgame FRTB & ISDA SIMM v2.6+ Quantitative Calibration: Expected Shortfall, Non-Modellable Risk Factors (NMRF), and Cross-Margin Capital Optimization (2026)](https://quantcalctools.com/guides/basel-iii-endgame-frtb-ima-expected-shortfall-isda-simm-v2-margin-calibration-2026)
- Site research index: [QuantCalc Tools](https://quantcalctools.com/llms-full.txt)
