Market-risk reform is often presented as a new capital formula. In practice it changes the relationship between trading-book boundaries, desk definitions, valuation factors, market data, statistical evidence and fallback calculations. A bank can implement a formula correctly and still fail to produce a defensible capital number if those dependencies are inconsistent.
REFERENCE · NO CALL
This report synthesises introductory reform presentations, a historical Basel comparison chart, a collection of implementation research and a September 2026 validation guide. Their regulatory references span several versions and dates. The report explains the architecture and the distinctions needed to read those materials. Historical implementation timelines are not presented as current law or as evidence of the framework’s present adoption in any country.
Begin with the perimeter
A capital calculation applies to a defined set of positions. The boundary between trading and banking books determines which framework measures particular exposures. If positions can move between treatments opportunistically, apparent model improvement may simply reflect a change in perimeter.
The perimeter should therefore be connected to position records, business purpose and the applicable transfer rules. A desk classification is also an economic and organisational definition, not merely a convenient reporting folder. It identifies the activities whose model performance is assessed together.
Changing that definition can change the statistical result. Combining two desks may conceal one desk’s poorly represented risk; splitting a portfolio may change diversification and fallback capital. These effects need governance rather than being treated as harmless data rearrangements.
Separate the standardised and internal-model routes
The standardised approach provides a prescribed calculation and a common benchmark. The internal-model approach permits more institution-specific modelling, subject to evidence and eligibility requirements. The historical materials emphasise that the standardised calculation remains necessary as a fallback and comparison even where internal models are used.
The decision is not simply that one number is lower. Internal modelling creates obligations: reproducible valuation, approved factor coverage, usable price history, outcomes testing, independent validation and controlled production. The economic benefit should be assessed after considering those costs and the possibility of losing eligibility.
A modelled desk should therefore have a ready fallback calculation. Reconstructing it only after a failed assessment creates an avoidable operational problem at the same moment management needs reliable information.
Figure 01 · Conceptual framework
From trading-book perimeter to capital components
Standardised sensitivities are not total economic risk
The historical standardised architecture separates sensitivity-based risk, default risk and a residual-risk component for instruments whose features require additional treatment. Within the sensitivity-based component, delta, vega and curvature address different responses to market movements.
Sensitivity-based components retain distinct meanings
| Aspect | Retained observation |
|---|---|
| Delta | Delta captures local first-order effects. |
| Vega | Vega represents sensitivity to volatility under a defined convention. |
| Curvature | Curvature captures prescribed nonlinear revaluation beyond the corresponding linear contribution. |
These are inputs to rule-specific weighting, bucketing and aggregation, not interchangeable labels for one Greek calculation.
The quality of the input mapping remains critical. Internal curve nodes, credit spreads and volatility surfaces must be translated into the required factor representation. Unit errors or inconsistent interpolation can distort a standardised result as readily as an internal-model output.
The default-risk component addresses jump-to-default exposure. It should not be confused with the counterparty credit risk of the institution facing the trade. An equity position’s issuer default and the failure of a derivative counterparty are different events with different exposure mechanics.
Internal-model eligibility is a sequence of gates
The teaching framework combines several distinct assessments:
Separate internal-model assessments and their evidence
| Assessment | Principal question | Evidence type |
|---|---|---|
| Factor eligibility | Is there adequate observable-price evidence? | Qualified dated market observations |
| Backtesting | Are losses consistent with the stated percentile forecasts? | Forecasts and actual/hypothetical P&L |
| P&L attribution | Does the risk representation reproduce the portfolio’s value changes sufficiently? | Paired HPL and RTPL |
| Model validation | Are assumptions, implementation and use appropriate? | Independent analysis and outcomes evidence |
| Production controls | Can the process run reliably and be reproduced? | Input lineage, reconciliations and change records |
A favourable answer to one question does not buy exemption from the others. A broad factor set can pass attribution but struggle with observation eligibility. A model with a generous volatility calibration can produce acceptable exception counts while omitting important economic drivers. A technically sound model can produce unusable outputs through uncontrolled market-data snapshots.
The framework should therefore be implemented as linked evidence products. Each gate needs a clear input owner and a defined consequence, with consistent dates and portfolio scope across the chain.
Figure 02 · Qualitative comparison
Two routes and one common control need
| Approach or concept | What it contributes | What still matters |
|---|---|---|
| Standardised approach | Uses prescribed risk sensitivities and charges | It is not a measure of all economic risk |
| Internal-model approach | Requires a sequence of eligibility and validation gates | Approval and ongoing evidence are essential |
| Data governance | Maintains reproducible risk-factor evidence | Economic relevance does not ensure regulatory modellability |
Expected Shortfall changes the capital basis
The historical reform materials move the internal-model capital focus toward Expected Shortfall at 97.5%, while retaining percentile-based backtesting controls. This is not evidence that ES cannot be tested; specialised and joint VaR–ES methods exist. It reflects the combination of regulatory design choices and practical measurement constraints in the framework described by the sources.
Capital also incorporates liquidity horizons and stress calibration. A current one-day risk measure is therefore not the final capital input. The sequence includes multi-day factor changes, longer-horizon factor sets, stressed history and constraints on diversification.
Liquidity horizons in the historical teaching exhibit
| Observation | Retained horizon (days) |
|---|---|
| First horizon | One teaching exhibit uses liquidity horizons of 10, |
| Second horizon | 20, |
| Third horizon | 40, |
| Fourth horizon | 60 |
| Fifth horizon | and 120 days. |
Those buckets are useful for understanding the architecture, but assignment details depend on the risk factor and governing version. It would be misleading to infer a current regulatory treatment for a live trade solely from a general table copied from an old presentation.
Non-modellable factors must remain visible
The evidence requirement creates a distinction between factors included in the modellable ES set and factors receiving non-modellable stress treatment. A factor can remain essential to risk-theoretical P&L even when it does not meet the modellability test.
This closes an important conceptual loophole. Removing a difficult factor from the capital inventory does not make it economically immaterial. The factor should remain represented through the treatment appropriate to its evidence status, and its effect should reconcile with the valuation and attribution processes.
The implementation challenge is particularly strong for finely parameterised surfaces. More dimensions improve fidelity but distribute observable events across more buckets. Coarsening the grid can improve eligibility while degrading attribution. The design decision should evaluate both effects together.
Do not merge capital components by analogy
The source material includes a simplified legacy VaR multiplier example, a modern ES-based capital-stack example and a remediation example using a teaching proxy. Their reported rupee figures cannot be linked into a single continuous calculation without checking the underlying bases.
A useful capital input ledger should identify for every number:
- The portfolio and aggregation level.
- The risk measure, confidence level and horizon.
- Whether it is current, stressed, averaged or assumed.
- The rule version and any applicable minimum or multiplier.
- The separate charges that remain outside it.
An output described as a capital benefit should also identify the comparison. A difference between standardised and modelled measures is not automatically a reduction in the institution’s final binding requirement. Other constraints, floors and aggregation rules may limit the benefit.
Figure 03 · Conceptual framework
Can the desk use internal-model treatment?
How the parts connect
- Desk and model perimeterPath AAll required gates satisfied
- Desk and model perimeterPath BA gate fails
- All required gates satisfiedReviewReconciled capital record
- A gate failsReviewReconciled capital record
Wider Basel reforms have different perimeters
The historical comparison chart places market-risk changes beside credit-risk, counterparty-risk, operational-risk and output-floor reforms. That is useful context, but it can encourage an incorrect impression that all measures share the same implementation calendar or calculation base.
The chart includes a 72.5% fully phased output-floor reference and an older phase-in schedule. The percentage is retained as a historical framework parameter. The phase-in dates are not repeated as current effective dates. An output floor at an aggregate level should not be treated as a generic multiplier applied to each trading desk’s standalone figure.
Likewise, operational-risk reform, credit-risk standardisation and revised CVA capital have distinct mechanics. A market-risk programme needs interfaces with them, but it should not silently absorb their charges into a vaguely labelled FRTB total.
Read historical timelines as versioned evidence
The archive contains presentations referring to 2016 and 2019 standards, subsequent frequently asked questions and implementation dates that were later subject to change. A date on a slide can represent a proposal, publication, expected commencement or teaching assumption. Those meanings should not be conflated.
For a live implementation, the control record needs the governing jurisdiction, legal entity, applicable text, effective date, transition provisions and approved interpretation. An international standard and a national rule are related but not identical. The archive’s observations about Indian sequencing are historical commentary, not a substitute for that determination.
This report deliberately preserves the conceptual lessons without asserting an unverified current implementation status. The distinction protects the usefulness of the historical material: the architecture can remain educational even when its old calendar no longer applies.
Build the programme around reconciliations
The most useful implementation reconciliations connect position inventory to desk scope, pricing factors to risk factors, price evidence to modellability, risk factors to attribution, capital components to their aggregation and modelled desks to standardised fallback results.
Data lineage matters because a change at the beginning can affect several gates. A revised volatility-surface mapping may alter sensitivities, observation counts, attribution metrics and capital. Independent workstreams that optimise only their own output can produce an inconsistent overall system.
The implementation report should therefore show dependencies and controlled change effects, not just a completion percentage for each formula. It should distinguish mathematical implementation, validated economic use and operational readiness.
The conclusion is architectural
The enduring lesson of the historical reform material is that market-risk capital depends on a complete evidence chain. Portfolio definitions, data, factors, valuation and controls must agree before the final arithmetic can be trusted.
A well-designed programme makes every component reproducible and every limitation visible. It also keeps historical standards, local implementation rules and illustrative calculations separate, so that a technically impressive number does not carry a claim its inputs cannot support.
