An exercise designed to remove every excuse
Under Article 78 CRD, the EBA runs an annual benchmarking exercise where financial institutions calculate the risk weighted assets of various portfolios prescribed in the implementing technical standards. The results are submitted through their supervisor, and the EBA publishes the overall results. For a decade this was a conversation about internal models. With most European trading-book capital moving onto the alternative standardised approach from 1 January 2027, the exercise has pivoted: the current cycle restricts data collection to the ASA information set, and its centre of gravity is a deceptively simple annex.
Annex 10 defines a validation set of 388 portfolios built from over five hundred sensitivity-defined instrument rows in which every instrument is specified directly by its SBM sensitivities. No pricing models, no market data, no interpretation of trade economics. The regulator hands every bank the same delta, vega and curvature inputs; the only thing left to differ is the machinery that turns sensitivities into capital. If two banks disagree on an Annex 10 portfolio, at least one of them has implemented the rulebook incorrectly.
What supervisors have already found
The EBA’s reports on the early ASA submissions read like a checklist of quiet implementation risk: FX sensitivities appearing inside non-FX risk classes, instruments assigned to the wrong buckets, the aggregation formula applied inconsistently, and divergent readings of how SBM, DRC and RRAO provisions interact. None of these are modelling judgements: they are software defects in a calculation banks often assume is too mechanical to get wrong. The standardised approach’s difficulty was never mathematics; it is vocabulary, bucketing and plumbing, executed identically across thousands of risk factors.
The European calibration adds traps of its own. CRR Article 325ah runs twenty credit-spread buckets where Basel’s text has eighteen. Member-State sovereigns and covered bonds are carved out, high-yield sectors renumbered, the index buckets moved to 19 and 20 with several risk weights diverging along the way. An engine validated against the Basel text alone will misprice, or simply reject, the EU validation set.
2027: the exercise comes for everyone
Until now, benchmarking obligations followed internal-model permission. Under the ITS amendments now in consultation for the 2027 exercise benchmarking extends to institutions applying the ASA regardless of whether they use an internal model. The exercise moved to the second half of 2027 expressly to give newly included banks time to prepare. Hundreds of institutions that have never participated in a benchmarking cycle and never had a supervisor line their standardised numbers up against the market’s will submit exactly these portfolios, in the same year the ASA becomes their binding capital requirement and the new COREP templates go live. The institutions most exposed are precisely those that treated the standardised approach as the “safe” option requiring no independent validation.
What good preparation looks like
- Run the set now. The Annex 10 workbook is public. Pricing it through your production ASA engine costs days, not months, and converts an abstract supervisory risk into a concrete list of divergences.
- Reconcile at component level. Match per risk class, per delta/vega/curvature component, per correlation scenario, a total that happens to agree can hide offsetting errors that a supervisor’s decomposition will expose.
- Chase vocabulary before mathematics. Most observed failures are bucket assignment, risk-factor identification and jurisdictional parameter tables, exactly the defects the EBA has already documented.
- Fix before the data flows. Parallel-run output is feeding capital planning and, from 2027, regulatory reporting. An aggregation defect discovered by your supervisor is a finding; the same defect found by you is a maintenance item.
The Practitioner’s Test
LINXS built an independent FRTB-SA engine sensitivities-based method across delta, vega and curvature, plus the default risk charge carrying both the Basel and the EU CRR calibrations, and ran the complete EBA Annex 10 validation set through it: every instrument, all 388 portfolios, across all three correlation scenarios, without error. The same engine powers the free calculator on our website, where the Annex 10 workbook can be uploaded and priced in the browser in under a minute.
Our ASA Pre-Validation service applies that engine to your implementation: your sensitivities and the EBA set, priced independently, with a component-level divergence report against your own system’s numbers before benchmarking season does it for you.
Test the claim yourself — upload the EBA workbook and watch it price.
Open the FRTB-SA Calculator© 2026 LINXS Advisory. For information and discussion only; not regulatory reporting, a capital calculation for supervisory purposes, or advice. References to EBA materials describe publicly available documents and imply no endorsement. Calculations described are produced by LINXS’s independent implementation of the published standards and should be validated against applicable implementing texts.