SA-CCR — Counterparty Credit Risk¶

Specifications for the Standardised Approach for Counterparty Credit Risk (SA-CCR), the regulatory method for translating OTC derivatives and long-settlement transactions into a single exposure-at-default (EAD) figure that then feeds the SA / IRB credit-risk exposure ladder.

Primary regulatory source: PRA PS1/26 Part Three Title II Chapter 6 (Art. 271–311) — the UK SA-CCR regime is largely a verbatim re-export of the onshored CRR text with the Basel 3.1 alpha (1.4) and supervisory-factor calibration retained. References on each child page follow the PRA-priority convention: PRA Art. numbers first, BCBS CRE codes as a secondary cross-reference.

Engine entry point: src/rwa_calc/engine/ccr/pipeline_adapter.py::ccr_rows_to_exposures — shapes SA-CCR netting-set EADs into synthetic exposure rows compatible with RAW_EXPOSURE_SCHEMA, so the downstream Classifier / CRM / SA Calculator chain consumes a CCR row identically to an on-balance-sheet loan.

Pipeline shape¶

SA-CCR is a strictly ordered chain. Each stage consumes the prior stage's LazyFrame and emits a new one — there are no back-edges. The orchestrator at engine/ccr/pipeline_adapter.py::ccr_rows_to_exposures chains the per-trade stages over the legally-enforceable trades in a RawCCRBundle, then aggregates to one synthetic exposure row per netting set:

ccr_rows_to_exposures(raw_ccr, config_ccr, reporting_date)
  │
  ├─ 1. Adjusted notional               (Art. 279b)
  │     ├─ compute_adjusted_notional_ir         (Art. 279b(1)(a))
  │     ├─ compute_adjusted_notional_fx         (Art. 279b(1)(b))   [if fx_rates supplied]
  │     ├─ compute_adjusted_notional_credit     (Art. 279b(1)(a))   [placeholder]
  │     ├─ compute_adjusted_notional_equity     (Art. 279b(1)(c))   [placeholder]
  │     └─ compute_adjusted_notional_commodity  (Art. 279b(1)(c))   [placeholder]
  │
  ├─ 2. Supervisory delta               (Art. 279a)
  │     └─ compute_supervisory_delta_option     (±1 linear / Black-Scholes Φ(d1) for options)
  │
  ├─ 3. Maturity factor                 (Art. 279c unmargined, Art. 285 margined)
  │     └─ compute_maturity_factor_unmargined   (margined branch pending)
  │
  ├─ 4. Hedging-set partition           (Art. 277, 277a)
  │     └─ assign_hedging_set                   (IR per currency / maturity bucket,
  │                                              FX per currency pair, credit/equity per ref,
  │                                              commodity per type)
  │
  ├─ 5. Asset-class add-on              (Art. 277a(2)–(3))
  │     └─ compute_addon_per_asset_class        (SF · |D_HS|, per HS then per NS)
  │
  ├─ 6. Replacement cost (RC)           (Art. 275)
  │     └─ unmargined: RC = max(V − C, 0)       (margined RC pending)
  │
  ├─ 7. PFE multiplier + add-on         (Art. 278)
  │     └─ compute_pfe                          (mult = min(1, 0.05 + 0.95·exp((V−C)/(2·AddOn))))
  │
  └─ 8. EAD                             (Art. 274(2))
        └─ EAD = α · (RC + PFE),  α = 1.4

The resulting per-netting-set EAD is written to drawn_amount on a synthetic exposure row with risk_type = "CCR_DERIVATIVE", ccr_method = "sa_ccr" and source_netting_set_id for downstream reconciliation. From that point on the row is treated as any other unsecured exposure — Classifier resolves the counterparty class, the SA calculator looks up the risk weight from Art. 112–134, and the aggregator rolls the resulting RWA into the firm-level totals.

Specification index¶

The "Pending" pages will land as the SA-CCR engine batches P8.35–P8.38 (credit / equity / commodity add-ons, margined RC, WWR, CCP, failed-trade ladder) ship — see the project root IMPLEMENTATION_PLAN.md for the ordering. Each page, when written, mirrors the structure of the existing adjusted-notional and fx-treatment pages: regulatory citation, formula, engine entry point, pipeline ordering note, and a worked acceptance scenario.

Asset-class coverage¶

SA-CCR partitions every trade into one of five asset classes (Art. 277(1)); each has its own supervisory factor, hedging-set rule and add-on aggregator. The engine ships the IR + FX branches; the remaining three classes carry placeholder no-op functions in engine/ccr/adjusted_notional.py and emit null adjusted_notional until the engine batches P8.35–P8.38 land.

Citation note: the supervisory factors above are illustrative orders-of-magnitude. The authoritative per-row values live in the regulatory table at PRA PS1/26 Art. 280 Table 1 and will be reproduced verbatim on hedging-sets.md when that page lands.

Scenario coverage — CCR-A¶

SA-CCR scenarios use the CCR-A prefix, matching the CRR-* / B31-* convention documented on the specifications index. The table below is a placeholder roadmap — to be finalised as scenarios are implemented in the engine batches P8.35–P8.38 and earlier. Scenario IDs that are already shipped link to the corresponding acceptance test; unshipped IDs are reserved.

Cross-references¶

• Specifications index — full list of CRR / Basel 3.1 / CCR scenario prefixes and per-spec regulatory references.
• Overview — project overview with the SA-CCR pipeline positioned in the broader RWA calculation.
• SA risk weights (CRR) — risk-weight lookup the CCR EAD flows through downstream of pipeline_adapter.py.
• Adjusted notional — per-asset-class d formula and the IR + FX worked-through reference.
• FX treatment — end-to-end CCR-A2 worked example showing the full chain from input row to RWA.

References¶

• PRA PS1/26 Part Three Title II Chapter 6 (Art. 271–311) — UK SA-CCR regime; primary regulatory source for every page in this directory.
• BCBS CRE52 — underlying methodology and Basel-level calibration of supervisory factors, correlations and the α = 1.4 scalar.
• src/rwa_calc/engine/ccr/pipeline_adapter.py — orchestrator that chains the per-trade SA-CCR stages and emits the netting-set-grain synthetic exposure rows.