Skip to content

Open Items and Limitations

Jump to bottom
Daniele Penza edited this page Oct 27, 2026 · 9 revisions

🚧 Open Items and Limitations

Section Release Tag Snapshot

📌 Guide, not policy. The CHANGELOG, the support matrix and the linked issues govern; issue state can change after this snapshot, and where this page differs, the repository wins.

📍 Status at release 1.0.2 (v1.0.2, 2026-10-23)

  • Version 1.0.2 is released. The annotated tag v1.0.2 is on commit cfb88e5, and the GitHub Release carries the production workbook K-SACCR_1.0.2.xlsm (SHA-256 1ed00eba852d659b121419846e2c6f1e665ed15353f069bc7ffeea96b8ce4f3f), the user manual K-SACCR_User_Manual_1.0.2.pdf and SHA256SUMS.txt. Check the hash before unblocking the workbook (INSTALLATION.md › Released workbook).
  • What 1.0.2 adds (clearing coverage and CCP trade exposures): a clearing member's trade exposures to a CCP, role CCP with the NettingSets columns R QCCP (Y/N) and S CCP retains VM, collateral not protected (Y/N) (#112, part (a)); settled-to-market trades, with the Trades columns AB Settlement type (N/STM) and AC Next reset date, Comment moving to AD (#360); role MULTI, which rejects a multi-level client structure, and an INFO line on Checks for every CM, CLIENT and CCP netting set (#359); a light-red highlight of INVALID and INCOMPLETE rows on Results; and the clearing coverage specification (#358) with its closing evidence (#361). A portfolio entered as in 1.0.1, with the new columns blank, gives the same EAD; the new columns change the layout of both input tables (see Getting Started › Moving inputs from 1.0.1).
  • Development continues on the active release branch; release/1.0.0, release/1.0.1 and release/1.0.2 are closed. Behaviour on the release branch can change until the next release; use the v1.0.2 tag or its workbook for stable results.
  • The engine implements EU CRR and Basel CRE52; only Basel rules covered by a published CRE99 output are validated (T01–T14). CRR rules, and the client-clearing, CCP and settled-to-market rules of both regimes, are implemented and tested against illustrative values, not validated.

🧾 Release 1.0.2 certification

Item Result at cfb88e5
Host Excel for Microsoft 365, Version 2608 (Build 16.0.20326.20072), 64-bit, Windows 11 Enterprise 25H2 x64
Compile PASS
Full regression All eight suites PASS, each run on its own: 288 cases, 1,283 checks
Evidence validator PASS
Release workbook source round trip PASS: K-SACCR_1.0.2.xlsm (production) holds exactly the source of the tag, 24 components, 0 differences, binding=VERIFIED
Development workbook source round trip PASS: the certified development workbook holds exactly the source of the tag, 35 components, 0 differences, binding=VERIFIED
Deviation A first attempt failed TEST_MainState on a recalculation interrupted in the Excel session; it is kept in the evidence. A second attempt on a fresh workbook from the template passed every suite; no code changed between the attempts

🧾 Release 1.0.1 certification — #308

Item Result at 31222d1
Compile PASS
Full regression All eight suites PASS, each run on its own: 230 cases, 1,033 checks
Evidence validator PASS with the candidate's own policy; no deviation
Release workbook source round trip PASS: K-SACCR_1.0.1.xlsm holds exactly the source of the tag, 24 components, 0 differences, bound to the workbook
Development workbook source round trip PASS: the certified development workbook holds exactly the source of the tag, 35 components, 0 differences, bound to the workbook
Published workbook Downloaded from the release, hash checked, unblocked and used as a user would: Validate, Run, Clear, save and reopen, an evidence export, and a refused export after an edited output

🧾 Release 1.0.0 certification — #46

Item Result at 87194e8
Compile PASS
Full regression All eight suites PASS, each run on its own: 228 cases, 1,001 checks
Release workbook source round trip PASS: K-SACCR_1.0.0.xlsm holds exactly the source of the tag, 24 components (7 standard + 17 document), 0 differences
Evidence validator Declared deviation: against the evidence policy at the tag, the validator rejects the record (EVIDENCE_INVALID) only because that policy names TEST_InputValidation's first case without the input- prefix the suite prints. The same record validates PASS against the policy as corrected in #305, which ships in 1.0.1
Supported-size measurements Done (#291): 1,000 to 500,000 legs measured on the Windows 64-bit host, with an inflated UsedRange scenario; see Getting Started › Portfolio size

The tag messages and the release notes record these certifications. A fix to a released version is published as a new patch release; a published release is never changed.

⚠️ Product limitations

From the CHANGELOG's Known limitations, the README and the methodology's support matrix:

Validation and hosts

  • Only Basel rules covered by a published CRE99 output are validated; no CRR rule is validated in this release, and no client-clearing, CCP or settled-to-market rule in either regime: their values are illustrative, from the derivations in tests/derivations/. There are no independent expected values; they can come only from an external implementation registered as a source (#100).
  • No automated check compiles VBA or runs Excel; Excel evidence is manual and covers one 64-bit Windows host. 32-bit is best effort and untested; Excel for Mac and for the web are not supported.
  • Portfolio size: 500,000 legs calculated in under two minutes on one host; the Trades dropdowns and validation cover rows 5 to 5,000 only, and an inflated used range slows every run (portfolio size).

Not calculated

Item Treatment Planned
Digital options No digital instrument; option legs of one economic trade are an ordinary combination — neither the CRR payoff-sized collar nor the BCBS cap of the effective notional is applied #120 (v1.4.0)
Optional reductions for sold options and sold credit protection (CRR Art. 274(5), (7); CRE52.1 FAQ1–2) Not applied, which gives an exposure not lower than with them (equal where they would not bind) —
Several margin agreements in one netting set; one agreement over several netting sets Declared as MIXED / SHARED and rejected —
Settled-to-market trades in a margined netting set The netting set is rejected (INVALID: STM requires unmargined treatment; …): an all-STM set is misclassified, a mixed set needs the CRE52.74 treatment, not implemented —
Multi-level client structures Declared with role MULTI and rejected; entered as CM or CLIENT, they cannot be detected —
Risk weights of CCP trade exposures Stated on Checks (2% for a QCCP; the standardised weight, to be determined outside K-SACCR, for a non-qualifying CCP), not applied —
Default-fund contributions to a CCP, their own funds requirements and the cap of CRR Art. 303(2) / CRE54.40 Cannot be entered Outside K-SACCR: K-Risk Engines
Simplified SA-CCR, original exposure method Cannot be entered #107 (v1.5.0), #111 (v1.6.0)
SFTs, CVA, RWA Cannot be entered Outside K-SACCR: SFTs in K-SFTs, RWA in K-Risk Engines, CVA risk in an engine of its own
The 1904 date system The run stops —

Inputs that remain the user's responsibility

  • Trade-level data, the mapping of risk drivers and of legs of multi-risk trades (CRR-RTS Articles 1–4 are applied by the user, not checked), and input completeness.
  • Collateral eligibility and haircuts: VM and NICA are entered after haircuts, in reporting currency.
  • FX rates: user-supplied; K-SACCR sources no market data (the template's rates are illustrative).
  • One leg per FX trade: the notional is the leg you enter; choosing it when both legs are foreign is yours.
  • FX options use a fixed quote order (EUR, GBP, AUD, NZD, USD, CAD, CHF, JPY; others alphabetically), which may differ from local market practice.
  • Variable notionals: one notional and one maturity per row; averages and leverage are entered by the user. Settled-to-market trades are declared with STM and their next reset date; whether a contract qualifies is the user's (daily variation margin alone is not STM), and no exchange calendar is derived.
  • Unrated single names under CRR: enter CQS3, or CQS5 where Art. 128 applies, or the mapped step.
  • Client-clearing eligibility, the large-or-illiquid flag and legal netting.
  • For a CCP: the role CCP, its QCCP status and whether it retains VM without protecting the collateral are the user's declarations. Client trades the clearing member need not reimburse on the CCP's default are not entered. A CCP entered with role N is calculated as bilateral and cannot be detected.

Conservative choices

  • One large or illiquid flag covers more than 5,000 trades and illiquid collateral, so it also sets 20 days where the trade-count floor does not apply, for example for a CCP netting set or a clearing member's eligible client netting set.
  • Under BCBS a client facing its clearing member keeps the 10-day floor, a conservative interpretation of the Basel text.
  • After margin disputes the engine doubles F + N − 1 in both regimes (see Margining and MPOR).

Workbook behaviour to keep in mind

  • INVALID and INCOMPLETE netting sets have no EAD (their rows are highlighted in light red from 1.0.2); the totals cover each regime's VALID sets only and are never added across regimes. CCP trade exposures stay in their regime's total; Checks gives their share.
  • A workbook built from a 1.0.1 template stops at the header check with the 1.0.2 VBA: rebuild it from the new template.
  • The synchronizer changes code only; a template change needs a rebuild.
  • A comparison with a bank's figures is not independent validation.

🗓️ Later milestones

Milestone assignment is planning, not delivery. The milestones hold the current scope; the table follows the README roadmap on main. The v1.0.2 milestone was delivered with this release, with part (a) of #112. K-SACCR's milestones stay with the derivatives exposure value and the workbook that produces it. Securities financing transactions are developed in K-SFTs; risk-weighted assets, capital and default-fund contributions to CCPs in K-Risk Engines; CVA risk in an engine of its own.

Milestone Scope Issues
v1.1.0 CSV import and staging for Trades and NettingSets; workbook protection #206 CSV contracts and templates · #207 deterministic CSV parsing · #208 staging worksheets · #209 validation linked to Checks · #210 corrections and repeatable Validate/Commit · #211 atomic commit with rollback · #420 protection of cells and structure · #421 protection lifecycle for macros · #422 protected workflows and maintenance unlock · #212 regressions, evidence and docs
v1.2.0 Excel Ribbon interface #213 RibbonX tab contract · #214 RibbonX XML in the build · #215 callbacks · #216 command availability · #217 remove the sheet buttons after parity · #218 certification · #330 Home landing page · #329 worksheet navigation sidebar · #357 Restore Parameter Defaults command
v1.3.0 Diagnostics and independent validation #220 engine separated from Excel I/O · #81 stable Checks codes · #100, #352 independent second-source cross-check · #165 randomised FormulaTest inputs · #259 formatted run report (XLSX or PDF) next to the CSV evidence
v1.4.0 Advanced products and FX provenance #120 digital options · #157 FX quote import with provenance · #328 TARGET holiday calendar
v1.5.0 Simplified SA-CCR #107
v1.6.0 Original exposure method #111
v1.7.0 Standalone Python engine: the full SA-CCR without Excel, with its own reference tests, and a reconciliation of its intermediates and EAD with the workbook's #450 to #459

🔗 Related

🏦 K-SACCR · ⚖️ EU CRR · Basel CRE52 · 🧪 Published CRE99 checks T01–T14; CRR implemented, not numerically validated · 📦 Release 1.0.2 published 2026-10-23 · 🧮 No real portfolio data · ⚡ 500,000 synthetic trade legs in under two minutes · Repository · Validation Approach · Open Items

🏦 K-SACCR

⚖️ Regulation

🧮 Workbook

📌 Guide, not policy: the repository documents govern.

Clone this wiki locally