Finalised Basel III («Basel IV») — Building Blocks, Parameters & Methods · V1

Exposures

Loans & advances
on-balance

Commitments & guarantees
off-balance × CCF

Derivatives & SFTs
SA-CCR / IMM → EAD

Trading positions
boundary → trading book

Income & activity
business indicator

▶

Risk engines — each produces an internal and a standardised answer

03 Credit risk U: IRBS: SA-CR

03 Counterparty credit U: IMMS: SA-CCR

05 CVA U: SA-CVAS: BA-/SA-CVA

04 Market risk U: A-IMAS: (S)SA / A-SA

06 Operational risk U = S: SMA

Securitisation (07) & funds (03·C) U: SEC-IRBAS: SEC-SA

▶

U-TREA
sum of internal approaches where approved

S-TREA
the same book, all-standardised

▶

02 Output floorTREA = max(U-TREA, 55%×S-TREA)
x → 72.5% by 2030

▶

01 Requirements — all % of TREA

Minima 4.5 / 6.0 / 8.0% + P2R

Combined buffer (CCoB · CCyB · G/O-SII · SyRB)

MREL / TLAC risk-based leg

08: SREP · P2R/P2G · IRRBB · Pillar 3 disclosure — steer & verify, don't sum

Leverage exposure measure
non-risk-based lane — bypasses the engines

▶

LREM = on-bal + derivatives (SA-CCR-based) + SFT + off-bal (CCF ≥ 10%)

▶

3% Tier 1
+ 50% × G-SII buffer

▶

Parallel backstop — binds where floored RWA stay low

Capital instruments — eligibility by tier

Art 28 · 52 · 63 CRR

What an instrument must promise — and give up — to count in each tier.

	CET1	Additional Tier 1	Tier 2
Maturity	Perpetual; no incentive to redeem	Perpetual; call ≥ 5y, supervisory approval, no expectation of call	Original maturity ≥ 5y; straight-line amortisation in the final 5y
Distributions	Fully discretionary, non-cumulative	Fully discretionary, non-cumulative; no dividend pushers/stoppers	Coupons payable; deferral not required
Loss absorption	First loss, going concern	Write-down / conversion at CET1 < 5.125% (or higher contractual trigger)	Gone concern — bail-in / insolvency
Subordination	Most junior claim	Senior only to CET1	Senior to AT1; junior to all other liabilities

Repurchase/reduction of any tier requires prior supervisory permission (Art 77–78); grandfathering of legacy instruments has lapsed.

Own funds — and the bridge to TLAC / MREL

Art 26–72b · 92a CRR · Art 45 ff BRRD

Going-concern capital is built left to right; add eligible liabilities and the same stack becomes gone-concern loss absorbency.

+ CET1 items

Capital instruments & share premium
Retained earnings · interim profits (verified)
Accumulated OCI · other reserves
Eligible minority interests (surplus excluded)

− CET1 deductions

Goodwill & intangibles (net of DTLs)
Loss-dependent DTAs
IRB EL shortfall (03·B) · NPE backstop shortfall
Own & cross holdings · DB pension assets · AVAs

Threshold exemption

Significant FS holdings & temporary-difference DTAs: each ≤ 10% of CET1, ≤ 17.65% combined → risk-weighted 250% instead of deducted

=

CET1 capital

+ AT1

Instruments & premium − own/cross holdings
Shortfall spills the deduction into CET1

Tier 1 capital

+ Tier 2

Instruments & premium − own/cross holdings
General provisions ≤ 1.25% SA RWA · IRB EL surplus ≤ 0.6% IRB RWA

Total capital = own funds

＋

+ Eligible liabilities

Issued & fully paid; not funded, secured or guaranteed by the entity itself
Remaining maturity ≥ 1 year; no set-off, no acceleration outside resolution
Subordination: contractual, structural (HoldCo) or statutory — required for resolution entities, limited senior allowance
No incentives to redeem; calls only with authority approval

TLAC / MREL resources
own funds + eligible liabilities

Calibration

G-SII TLAC (Pillar 1): ≥ 18% of TREA and ≥ 6.75% of LREM
MREL (firm-specific, resolution authority): loss-absorption amount + recapitalisation amount + market-confidence charge, set per resolution strategy
Internal MREL: 75–90% pre-positioned from material subsidiaries to the resolution entity
CET1 used for buffers cannot double-count toward MREL/TLAC (stacking order)

Reading the bridge: Pillar 1 capital protects a going concern; MREL/TLAC ensures that, in resolution, the same balance sheet holds enough bail-in-able claims to absorb losses and rebuild the capital stack without public money. Cross-holdings of other G-SIIs' TLAC are deducted from own funds to stop contagion loops.

Capital ratios & minima

Art 92 CRR

Three fractions over one denominator — the floored TREA from section 02.

CET1 ratioCommon Equity Tier 1TREA≥ 4.5%

Tier 1 ratioCET1 + AT1TREA≥ 6.0%

Total capitalTier 1 + Tier 2TREA≥ 8.0%

+ P2R — firm-specific, same fraction structure: ≥ 56.25% CET1, ≥ 75% T1.
+ Combined buffer (CET1 only) on top; breach → MDA ladder restricting distributions, AT1 coupons, variable pay.
Denominator discipline: TREA is the floored figure — the output floor propagates into every ratio on this chart.

Buffers (CET1)

Art 128–142 CRD

Usable-in-theory cushions stacked on the minima; all sized on floored TREA, all CET1.

SyRB uncapped · sectoral possible · additive if different risk

max( G-SII 1.0–3.5% , O-SII ≤ 3.0% )

CCyB 0–2.5%+ · exposure-weighted avg of national rates

CCoB 2.5% · always on

Sum of the four layers = combined buffer requirement (CBR), sitting on top of 8% + P2R.
Dip into the CBR → automatic MDA restrictions scale with the depth of the breach; using buffers is legal, but the payout penalty makes them sticky in practice.
CCyB rates are set per country by macroprudential authorities (NL: DNB) and bind via the location of exposures, not of the bank.

Leverage ratio

Art 429–430 CRR

The non-risk-based backstop — same capital, no risk weights.

3%T1 / LREM

G-SII add-on = 50% × G-SII buffer; leverage MDA applies to it.
LREM: on-balance + SA-CCR-based derivatives + SFT add-on + CCF-weighted off-balance (min 10%).

Large exposures

Art 387–403 CRR

Concentration limits — a cap per name, not a risk weight; and "name" is defined widely.

Single client / group of connected clients	25% Tier 1
G-SII exposure to another G-SII	15% Tier 1
Large-exposure definition & reporting threshold	≥ 10% Tier 1

Group of connected clients (GCC)

Control: clients linked by a control relationship (parent–subsidiary, dominant influence) form one risk regardless of sector.
Economic interdependency: clients so linked that the failure of one would likely cause funding or repayment difficulties for the other (single dominant customer, common funding source, guarantees) — grouped even without ownership links (EBA GL on connected clients).
Look-through of funds, securitisations and other structures to underlying obligors; unknown underlyings pool into a single "unknown client".

Further requirements

Exemptions (fully or partly): sovereign exposures, intragroup with waiver, covered bonds (discretion), qualifying CCP trade exposures.
Trading book: the limit may be exceeded only there, against a dedicated own-funds add-on that escalates with the excess.
CRM substitution: protected amounts shift to the protection provider — concentration in guarantors is monitored too.
Reporting: all large exposures, the 10 largest to institutions and the 10 largest to shadow-banking entities; internal limits on aggregate shadow-banking exposure per EBA guidelines.

Mechanism

single stack

Internal models may lower RWA — but never below a fixed share of the standardised answer.

TREA = max( U-TREA , x% × S-TREA )

Compute U-TREA — total RWA using all approved internal approaches (IRB, IMM, SA-CVA, A-IMA).
Compute S-TREA — the same book, all-standardised: SA-CR, SA-CCR, BA-/SA-CVA, SMA, standardised market risk.
Apply the floor percentage x — x is the transitional floor factor set by Art 465(1): 50% in 2025, rising one step per year (55% today) to its permanent level of 72.5% from 2030. Whichever of U-TREA and x%·S-TREA is higher becomes the institution's TREA.
Use the floored TREA everywhere — it is the denominator of every requirement expressed in % of TREA: Pillar 1 minima, P2R, the combined buffer and the risk-based MREL/TLAC leg. Application at each level of consolidation, with a member-state discretion to apply it only at the highest level within a member state.

Worked example — when does the floor bite? IRB mortgage bank: U-TREA = €60bn, S-TREA = €100bn.
2026: TREA = max(60, 55% × 100) = max(60, 55) = €60bn — internal models still bind.
2030: TREA = max(60, 72.5) = €72.5bn — the floor binds: +21% RWA, with every ratio, buffer and MREL requirement scaling up with it.
The lower a bank's model RWA relative to standardised, the earlier in the phase-in this switch happens.

Phase-in & cap

Art 465(1)–(2)

Five annual steps to steady state — the 55% step is the one in force today.

72.5%steady state 2030

55%in force · 2026

50%2025

55%2026 ◀

60%2027

65%2028

70%2029

72.5%2030→

Transitional cap: floored TREA ≤ 1.25 × unfloored during phase-in (member-state discretion).
Supervisors avoid double-counting floored risks in P2R/SyRB; P2R nominal freeze on floor step-ups pending SREP review.

Transitional S-TREA calibrations

Art 465(3)–(5) · to 31·12·2032

Softer standardised inputs — for the floor calculation only — with a hard expiry.

Discretion	Floor-only calibration
Low-risk residential mortgages (NL a.o., strict loss criteria)	10% ≤ 55% of value · 45% remainder
Unrated corporates, own-estimate PD ≤ 0.5%	65% RW
SEC-SA in the floor	p halved → 0.5 · STS 0.25

Cliff: all three lapse end-2032 unless extended by legislation — a low-risk mortgage book's S-TREA roughly doubles overnight in 2033.

03 · A — Determining the exposure value

Exposure value & CRM

Art 111 · 166 · 192–241 · 271–285

Turns every asset, commitment and derivative into the euro amount the risk weights apply to.

On-balance sheet — accounting value net of specific credit risk adjustments.

Off-balance sheet — nominal × CCF:

Unconditionally cancellable commitments	10% PHASED · Art 495d
Other commitments (any maturity)	40%
Transaction-related contingents · NIFs/RUFs	50%
Short-term self-liquidating trade LCs	20%
Direct credit substitutes, guarantees	100%

Derivatives → SA-CCR (the default counterparty credit risk method):

EAD = α × (RC + PFE) · α = 1.4
PFE = mult × AddOn^agg · mult = min(1, 0.05 + 0.95·e^{(V−C)/(1.9·AddOn^agg)})

EAD exposure at default per netting set · α supervisory scalar 1.4 · RC replacement cost (what it costs to replace the trades today) · PFE potential future exposure (how much worse it could get) · V current market value of the trades · C net collateral held · mult multiplier recognising over-collateralisation / negative value · AddOn^agg aggregate add-on across asset-class hedging sets

Supervisory factor	SF
Interest rate	0.5%
FX	4.0%
Credit (single-name, by rating)	0.38–6.0%
Equity — single / index	32% / 20%
Commodity — other / electricity	18% / 40%

RC unmargined = max(V−C, 0); margined = max(V−C, TH+MTA−NICA, 0), where TH = margin threshold, MTA = minimum transfer amount, NICA = net independent collateral amount. Add-ons are built per asset-class hedging set from adjusted notionals × supervisory delta (δ) × maturity factor (MF) × supervisory factor (SF).

Worked examples → section detail ⧉.

Method hierarchy — all three standardised methods share the same skeleton, EAD = α × (RC + PFE); they differ in how finely PFE is measured:

Method (eligibility by derivative business)	Calculation
SA-CCR — default	Full mechanics above: deltas, hedging sets, margin recognition, multiplier
Simplified SA-CCR — ≤ 10% of assets & ≤ €300m	same formula · mult = 1 · coarse add-ons, no supervisory deltas or margin refinement
Original exposure method (OEM) — ≤ 5% & ≤ €100m	EAD = 1.4 × (max(V,0) + N × supervisory % by asset class & maturity)
IMM — internal model, approval	EAD = α × effective EPE (expected positive exposure) · SA-CCR shadow still required for S-TREA

N notional · V market value · EPE model-based expected positive exposure. Simpler methods trade risk sensitivity for operational cost — proportionality for small derivative books.

Result & hand-over — the output of this panel is one number per facility or netting set: the exposure value E. Two things can still change it before capital is computed: eligible collateral reduces E itself (comprehensive method, E*), and eligible guarantees replace whose risk weight applies. Both live in the CRM panel to the right; the mitigated exposure then flows into risk weighting under SA or IRB (group 03·B below).

Credit risk mitigation

Art 192–241 CRR

Two families: collateral you can seize (funded) and promises you can enforce (unfunded).

Classify the protection —
Funded
- Financial collateral (cash, debt, equities, gold, CIU units)
- Receivables · immovable property · other physical (IRB)
- Netting — on-balance & master netting agreements
Unfunded
- Guarantees
- Credit derivatives (CDS, TRS; no restructuring clause → 40% haircut on protection)
- Eligible providers: sovereigns, institutions, rated corporates a.o.
Financial collateral — simple method (SA only): collateralised portion takes the collateral issuer's RW, floor 20% (0/10% for certain repo/sovereign cases).

Financial collateral — comprehensive method:

E* = max( 0 , E·(1+H_E) − C·(1−H_C−H_FX) )

E* exposure value after mitigation · E exposure value · C market value of collateral · H_E haircut inflating the exposure (its value may rise) · H_C haircut deflating the collateral (its value may fall) · H_FX extra haircut where collateral and exposure are in different currencies

Supervisory haircuts H_C
Debt collateral (by issuer / maturity / CQS)	0.5–15%
Main-index equities · gold	20%
Other listed equities	30%
Currency mismatch H_FX	8%

Status under CRR3: own-estimate volatility adjustments are abolished — supervisory haircuts only. The supervisory volatility adjustments for master netting agreements (SFTs: repo, securities lending) remain in force: E* is computed on the net position with add-ons per security class and currency mismatch.

LGD for funded credit protection (F-IRB, Art 230) — still in force under CRR3, recalibrated: for fully secured exposures (after haircuts) the supervisory LGD* replaces the unsecured 40%:

Collateral Financial Receivables Resi / comm. RE Other physical

LGD* 0% 10% 10% 15%

Partially secured exposures split into a secured part at LGD* and an unsecured part at 40% (25% under A-IRB floors).
Unfunded — substitution: the protected portion takes the protection provider's risk weight (SA) or PD/LGD (IRB); the uncovered remainder stays with the obligor. Adjust the recognised protection amount for mismatches:
maturity: G_A = G × (t − 0.25)/(T − 0.25) · currency: G* = G × (1 − H_FX)

G nominal protection amount · G_A / G* recognised amount after adjustment · t residual maturity of the protection (years) · T residual maturity of the exposure, both capped at 5y · H_FX 8% currency-mismatch haircut. Protection with t < 3 months, or t < T for exposures where that matters, is not recognised.
IRB interaction — F-IRB: collateral lowers supervisory LGD (secured LGD* by collateral type); A-IRB: own LGD estimates, subject to the 03·B floors.

Collateral	Financial	Receivables	Resi / comm. RE	Other physical
LGD*	0%	10%	10%	15%

03 · B — Determining risk-weighted assets

RWA under the Standardised Approach

Art 111–141 CRR

Fixed risk weights from tables — the calibration every bank now shadow-runs for the floor.

Assign exposure class — sovereign · institution · corporate (incl. specialised) · retail · real estate · equity · subordinated · defaulted · other.

Rated? Apply ECRA grid (due diligence may only raise the RW):

CQS →	1	2	3	4	5	6	Unrated
Sovereigns	0%	20%	50%	100%	100%	150%	100%¹
Institutions	20%	30%	50%	100%	100%	150%	SCRA ↓
— short-term ≤ 3m	20%	20%	20%	50%	50%	150%	—
Corporates	20%	50%	75%	100%	150%	150%	100%²
Covered bonds	10%	20%	20%	50%	50%	100%	derived

¹ EU member states in domestic currency 0%. ² SME corporate 85%; 65% investment-grade via floor transitional; EU SME & infrastructure supporting factors on top.

Unrated institution? Grade under SCRA:

Grade A+ / A B C

Base RW 30% / 40% 75% 150%

Short-term ≤ 3m 20% 50% 150%

A = minima + buffers met (A+ if CET1 ≥ 14% & LR ≥ 5%); B = minima only; C = breach or audit doubt.

Grade	A+ / A	B	C
Base RW	30% / 40%	75%	150%
Short-term ≤ 3m	20%	50%	150%

Retail & specialised classes:

Regulatory retail (granular, ≤ €1m)	75%
— transactors (cards/charge, fully repaid)	45%
Specialised lending — project (pre-op / op / high-quality)	130% / 100% / 80%
— object & commodities finance	100%
Subordinated debt / TLAC-eligible	150%
Equity — standard / speculative unlisted TO 2030	250% / 400%
— long-term strategic (discretion)	100%
Defaulted (unsecured, provisions < 20% / else)	150% / 100%

Real estate — completed, enforceable, ETV at origination (upward revaluation constrained):

Residential, general — loan-splitting	20% ≤ 55% of value · rest cpty RW
Residential IPRE — whole loan, ETV ≤50/60/80/90/100/>100	30% · 35% · 45% · 60% · 75% · 105%
Commercial, general — split ≤ 55% of value	min(60%, cpty) · rest cpty
Commercial IPRE — ETV ≤60/80/>80	70% · 90% · 110%
ADC	150% · 100% pre-sold resi

Worked examples → section detail ⧉.

Apply CRM substitution → RWA = RW × exposure value; supporting factors where applicable.

RWA under the IRB Approach (constrained)

Art 142–191 CRR

Own PD/LGD estimates inside a supervisory formula — with floors under every input.

5bpPD floor

25%LGD unsec. corp

×1.06scalar deleted

Check scope:

Class	A-IRB	F-IRB	SA
Large corporates (rev > €500m)	✕	✓	✓
Other corporates	✓	✓	✓
Institutions & other financials	✕	✓	✓
Retail (mortgages, QRRE, other)	✓	n/a¹	✓
Equity	✕	✕	✓ only

Class

A-IRB

F-IRB

SA

Large corporates (rev > €500m)

✕

✓

Other corporates

✓

Institutions & other financials

✕

✓

Retail (mortgages, QRRE, other)

✓

n/a¹

✓

Equity

✕

✓ only

¹ F-IRB does not exist for retail — retail IRB always uses own estimates of PD, LGD and CCF (A-IRB), or falls back to SA. Class-by-class adoption is permitted (partial use liberalised).

Estimate parameters, apply input floors (A-IRB):

PD — corporate & retail	0.05% · QRRE revolver 0.10%
LGD unsecured — corporate / other retail / QRRE	25% / 30% / 50%
LGD — residential mortgages	5%
LGD secured — fin / receivables / RE / other phys.	0% / 10% / 10% / 15%
EAD	≥ on-bal + 50% × SA-CCF off-bal

F-IRB: LGD senior unsecured 40% (was 45%), subordinated 75%; own CCFs only for revolving commitments; M = 2.5y (F-IRB) or effective 1–5y (A-IRB).

Asset correlation R:

Corporates	0.12–0.24 by PD
— SME size adjustment (S = revenue €5–50m)	− 0.04·(1−(S−5)/45)
— large (≥ €100bn assets) / unregulated financials	× 1.25
Residential mortgages / QRRE / other retail	0.15 / 0.04 / 0.03–0.16

Assemble the parameter blocks and run the risk-weight function (99.9% asymptotic single risk factor model):

PD
(floors, step 2)
＋

LGD
(floors / LGD*)
＋

EAD
(03·A, CCF)
＋

M
▶

R (step 3)
▶

K → RWA = 12.5 × K × EAD

K = [ LGD·N( (N⁻¹(PD) + √R·N⁻¹(0.999)) / √(1−R) ) − LGD·PD ] × MA
MA = (1 + (M−2.5)·b)/(1 − 1.5·b) · b = (0.11852 − 0.05478·ln PD)²
RWA = 12.5 × K × EAD

K capital requirement per unit of exposure · PD probability of default (1y) · LGD loss given default · EAD exposure at default · R asset correlation (step 3) · N / N⁻¹ standard normal CDF / its inverse · 0.999 the 99.9% confidence level · MA maturity adjustment · M effective maturity (years) · b smoothed maturity slope · 12.5 = 1/8% converts capital into RWA terms
No maturity adjustment for retail. Defaulted assets: K = max(0, LGD − EL_BE), i.e. capital covers only the uncertainty beyond the best-estimate expected loss EL_BE.
Specialised lending without PD estimates → supervisory slotting:

Category Strong Good Satisf. Weak Default

RW (M ≥ 2.5y) 70% 90% 115% 250% 0%*

RW (M < 2.5y) 50% 70% 115% 250% 0%*

*EL charge applies; categories mapped from supervisory criteria.
Expected loss & provisions — EL = PD·LGD·EAD; shortfall (EL − eligible provisions) deducted from CET1; excess → T2 up to 0.6% of IRB RWA.
Governance — rating-system standards, margin of conservatism, downturn LGD, annual validation; EBA IRB-repair guidelines baked into approvals.

Category	Strong	Good	Satisf.	Weak	Default
RW (M ≥ 2.5y)	70%	90%	115%	250%	0%*
RW (M < 2.5y)	50%	70%	115%	250%	0%*

03 · C — Special exposure types & adjustments

Collective investment undertakings

Art 132–132c · 104(7) CRR

Fund exposures are priced by transparency — and split by which book they sit in.

Banking book

Look-through (LTA): sufficient, verified (at least quarterly) information → risk-weight the underlying exposures as if held directly; fund leverage scales the result
Mandate-based (MBA): no look-through → assume the worst allocation the mandate permits, most punitive assets first
Fall-back (FBA): neither → 1250%
Third-party-computed LTA/MBA: results × 1.2

Trading book

Eligible only with daily price quotes and sufficient information (look-through or qualifying index tracking)
Capitalised in the FRTB engines (04): look-through into the SbM, or index treatment
Funds failing the criteria are mandatorily banking book (04, boundary)

Worked examples → section detail ⧉.

Also in Part 3, small but mandatory: settlement/delivery risk charges for unsettled transactions and free deliveries (Art 378–380).

RWA adjustments & multipliers

Art 123a · 501 · 501a CRR

EU-specific dials applied after the risk weight — down for policy, up for risk.

Adjustment	Factor
SME supporting factor (turnover ≤ €50m) — part ≤ €2.5m / above	× 0.7619 / × 0.85
Infrastructure supporting factor (qualifying projects)	× 0.75
Unhedged retail FX lending — currency ≠ income currency	× 1.5 · RW cap 150%

What CRR3 changed: the SME and infrastructure supporting factors are carried over unchanged from CRR2 (retained against EBA advice); the ×1.5 multiplier on unhedged retail FX lending is new in CRR3 (Art 123a).
Supporting factors are EU deviations from Basel; they apply in both U-TREA and S-TREA, so they survive the output floor.
Specific provisions reduce the exposure before weighting; the defaulted-asset RW then applies to the net amount.

CCP exposures

Art 300–311 CRR

Cleared trades get concessionary weights — if the CCP qualifies.

2%QCCP trade RW

Qualifying CCP? — authorised/recognised under EMIR → concessionary regime; else bilateral treatment.
Trade exposures (incl. posted initial margin, bankruptcy-remote rules apply):

Clearing member → QCCP 2%

Client via member (conditions met / partial) 2% / 4%

Non-QCCP bilateral SA
Default fund contributions — QCCP: capital from the CCP's hypothetical capital K_CCP (SA-CCR-based), allocated pro-rata to prefunded contributions; non-QCCP: 1250%.
Cap — total charge for a QCCP relationship ≤ the bilateral-treatment equivalent.

Worked examples → section detail ⧉.

Trading-book definition & approach choice

Art 94 · 104–106 · 325a

First decide what is trading; then let size decide which engine capitalises it.

Classify positions — by presumptive list, not intent:
Banking book — mandatory
- Unlisted equities
- Real estate holdings
- Retail & SME credit
- Funds without look-through or daily prices
- Securitisation warehousing instruments
- Derivatives & hedges of the above
Trading book — mandatory
- Held-for-trading assets & liabilities
- Market-making positions
- Underwriting commitments
- Listed equities
- Trading-related SFTs
- Correlation trading portfolio
Derogations — approval
- Listed equity held without trading intent → banking book
- Other deviations from the presumptive lists with documented evidence
- Supervisor may compel reallocation
Lock the boundary — reclassification only in exceptional circumstances, with approval and a capital surcharge preventing benefit; internal risk transfers only via dedicated IRT desks (GIRR) or eligible external hedging.

Select the approach by size:

Test (trading + FX/commodity business)	Outcome
≤ 5% of assets & ≤ €50m (Art 94)	banking-book treatment
≤ 10% of assets & ≤ €500m (Art 325a)	Simplified SA eligible
Above · dealer profile	A-SA default · A-IMA per desk (approval)

Simplified SA — the familiar CRR2 standardised charges, multiplied up per risk class:

Class IR EQ FX COM

Scalar × 1.3 × 3.5 × 1.2 × 1.9

The scalars compensate for keeping the old, less risk-sensitive method: they lift CRR2 charges roughly to FRTB levels. IR interest rate · EQ equity · FX foreign exchange · COM commodities.

Worked examples → section detail ⧉.

Class	IR	EQ	FX	COM
Scalar	× 1.3	× 3.5	× 1.2	× 1.9

EU status: CRR2 market risk applies for own funds until 01·01·2027 (FRTB reporting continues). EBA no-action: boundary/IRT not supervisory priorities before then. 3rd Delegated Act (in scrutiny): optional overall multiplier (Art 495v) toward CRR2-IMA levels for adversely affected banks + operational relief, 2027–29.

Alternative Standardised Approach (A-SA)

Art 325c–325ay

Price sensitivities × supervisory shocks — a model-shaped calculation with no model approval.

Own funds = SbM + DRC + RRAO

Sensitivities-based method — delta, vega, curvature per risk class (GIRR, CSR non-sec, CSR sec CTP/non-CTP, equity, commodity, FX).
Weight: WS_k = RW_k × s_k per prescribed bucket weights (GIRR by tenor; equity by market cap & economy; FX 15%, √2 divisor for liquid pairs).
s_k sensitivity of the portfolio to risk factor k (e.g. PV change per 1bp rate move, or the market value of a cash equity) · RW_k supervisory risk weight of that factor's bucket · WS_k weighted sensitivity — the loss assumed if the supervisory shock hits factor k
Aggregate within bucket:
K_b = √( max(0, Σ WS_k² + Σ_k≠l ρ_kl·WS_k·WS_l) )
Aggregate across buckets with cross-bucket correlations γ_bc; run the whole calculation under three correlation scenarios (all ρ, γ × 1.00 / 1.25 / 0.75, capped at 1) and take the highest — hedges are not fully trusted in either calm or stressed correlation regimes. Curvature repeats the exercise with full stressed up/down shifts (CVR = worst-of loss beyond delta), aggregated analogously.
ρ_kl correlation between factors within a bucket · γ_bc correlation between buckets b, c · K_b bucket-level capital · CVR curvature risk measure (non-linear loss)
Default risk charge — gross JTD = f(notional, LGD, P&L) → net by obligor & seniority → hedge-benefit ratio limits offsetting → RW by band:

AAA AA A BBB BB B CCC NR Dflt

0.5% 2% 3% 6% 15% 30% 50% 15% 100%

Securitisation DRC separately (CTP / non-CTP, no cross-offsetting).
Residual risk add-on — gross notionals × 1.0% (exotic underlyings: longevity, weather, natural catastrophe) or × 0.1% (other residual risks: gap, correlation, behavioural). A blunt charge for what sensitivities cannot see.

AAA	AA	A	BBB	BB	B	CCC	NR	Dflt
0.5%	2%	3%	6%	15%	30%	50%	15%	100%

Alternative Internal Model Approach (A-IMA)

Art 325az–325bp

Desk-by-desk permission to model — kept only by passing statistical tests every quarter.

97.5%ES confidence

≥1.5m_c

Nominate trading desks — strategy, risk management, independent P&L per desk.

Pass the gateways per desk:

Backtesting (desk & top-of-house)	97.5% & 99% VaR exceptions · traffic-light multiplier
P&L attribution (RTPL vs HPL)	Spearman + KS · green / amber (surcharge) / red (→ A-SA)

Nominate desk

▶

Backtesting

▶

PLAT

▶

Green: IMA

/

Amber: IMA + surcharge

/

Red: back to A-SA

Quarterly re-testing — desk eligibility is earned continuously, never granted permanently.

Risk-factor eligibility (RFET) — ≥ 24 real-price observations in 12 months (gap conditions) or ≥ 100; fail → NMRF.
Expected shortfall — 97.5% ES, stressed calibration, scaled across liquidity horizons:

LH (days) 10 20 40 60 120

e.g. IR/FX major, lg-cap eq other IR/eq, energy IG spread HY spread, sm eq vol structural/other
NMRF — stress-scenario capital (SES) per non-modellable factor, extreme shocks, limited diversification.
Default risk charge (IMA) — VaR 99.9%, 1-year, two-factor systematic model, PD floors as IRB.
Aggregate:
OF = max( IMCC_last + SES_last , m_c × IMCC_avg,60d + SES_avg,60d ) + DRC
m_c ≥ 1.5 + backtesting add-on (0–0.5) · + amber-desk surcharge · + A-SA for ineligible desks

OF own funds requirement · IMCC aggregated ES-based charge on modellable risk factors (constrained diversification: weighted mix of full and per-class ES) · SES summed stress-scenario capital for NMRFs · DRC internal default risk charge · m_c multiplier: 1.5 base, up to +0.5 for backtesting exceptions

Worked examples → section detail ⧉.

LH (days)	10	20	40	60	120
e.g.	IR/FX major, lg-cap eq	other IR/eq, energy	IG spread	HY spread, sm eq vol	structural/other

Scope & approach selection

Art 382–383

Capitalises the mark-to-market loss when a derivative counterparty's spread widens.

0.65DS discount BA-CVA

1.0m_CVA SA-CVA

Determine scope — OTC derivatives and fair-valued SFTs, minus EU exemptions (retained): non-financials below the clearing threshold, sovereigns, intragroup, pension scheme arrangements. Exempted trades must still be reported.

Select approach:

Approach	Condition
Simplified: CVA OF = CCR OF	OEM-eligible (≤ 5% / €100m)
BA-CVA — reduced or full	default
SA-CVA	approval · regulatory CVA model + sensitivities

Choice is strategic: BA-CVA punishes unhedged directionality; SA-CVA rewards a genuine CVA desk with model and data investment.

BA-CVA — calculation

Art 384

A closed-form charge from EAD, maturity and a sector risk weight — no sensitivities needed.

Per counterparty:

SCVA_c = (1/α) · RW_c · Σ_NS M·EAD·DF · α = 1.4
DF = (1−e^−0.05M)/(0.05M) · DF = 1 for IMM netting sets

SCVA_c standalone CVA capital for counterparty c · α the SA-CCR scalar, divided back out so it is not charged twice · RW_c supervisory weight for the counterparty's sector & quality (table below) · NS netting sets with c · M effective maturity (years) · EAD the netting set's SA-CCR exposure (03·A) · DF supervisory discount factor

Counterparty sector	IG	HY / NR
Sovereigns, central banks, MDBs	0.5%	2.0%
Local govt · govt-backed non-fin · education · public admin	1.0%	4.0%
Financials (incl. govt-backed)	5.0%	12.0%
Basic materials · energy · industrials · agri	3.0%	7.0%
Consumer goods/services · transport	3.0%	8.5%
Technology · telecoms	2.0%	5.5%
Health care · utilities · professional services	1.5%	5.0%
Other	5.0%	12.0%

Aggregate (reduced — no hedges):
K_red = √( (ρ·Σ_cSCVA_c)² + (1−ρ²)·Σ_cSCVA_c² ) · ρ = 50%

ρ = 50% assumed correlation between counterparties' credit spreads: the first term is the systematic move-together component, the second the idiosyncratic part. With a single counterparty the formula collapses to K_red = SCVA.
Hedged (full) version — recognise eligible single-name CDS (SNH), index CDS (IH) and a misallocation penalty (HMA):
K_hedged = √( (ρ·Σ_c(SCVA_c−SNH_c) − IH)² + (1−ρ²)·Σ_c(SCVA_c−SNH_c)² + Σ_cHMA_c )

Single-name hedge correlation r_hc

References the counterparty directly 100%

Legally related entity 80%

Same sector & region 50%
Combine & discount:
K_BA-CVA = DS × ( β·K_red + (1−β)·K_hedged ) · β = 0.25 · DS = 0.65

β = 0.25 floors hedge recognition — at least a quarter of the unhedged charge always survives, however good the hedges · DS = 0.65 overall discount scalar from the 2020 Basel recalibration, carried into CRR3

Worked example — reduced BA-CVA, one counterparty Unhedged book, one investment-grade financial counterparty: EAD = €20m (SA-CCR, 03·A), M = 4y, RW_c = 5.0%.
DF = (1−e^(−0.05×4))/(0.05×4) = 0.906 · SCVA = (1/1.4) × 5.0% × 4 × 20 × 0.906 = €2.59m
One counterparty → K_red = 2.59 · no hedges → K_hedged = K_red
K_BA-CVA = 0.65 × 2.59 = €1.68m own funds (RWA-equivalent ≈ €21m) — a 4y single-name CDS on the counterparty would remove up to 75% of it (β floor).

Single-name hedge correlation r_hc
References the counterparty directly	100%
Legally related entity	80%
Same sector & region	50%

SA-CVA — calculation

Art 383–383b

FRTB-style sensitivities on a regulatory CVA model — approval required, hedges rewarded.

Regulatory CVA model — exposure model producing CVA and its sensitivities; soundness, governance and validation requirements.

Compute delta + vega sensitivities (no curvature, no DRC, no RRAO) across six risk classes:

1 Counterparty credit spreads	the CVA-specific class
2 Reference credit spreads	exposure-side
3 Interest rate · 4 FX · 5 Equity · 6 Commodity	exposure-side

Weight & aggregate — SbM-style bucket and cross-bucket correlations at reduced granularity; eligible hedges (incl. market-risk hedges of the exposure component) enter with sign.
Multiply:
OF = m_CVA × Σ_{risk classes} K · m_CVA = 1

Legacy note: the m_CVA multiplier was reduced from 1.25 to 1 and the BA-CVA DS = 0.65 introduced in the 2020 Basel targeted revisions — both carried into CRR3. Numbers quoted from pre-2020 impact studies overstate CVA charges.

Calculation

Art 313–315

Capital as a stepped share of income — the bigger the business indicator, the steeper the marginal rate.

12·15·18%bucket coefficients

ILM=1EU fixed

Business indicator (3-year averages):
BI = ILDC + SC + FC
ILDC = min(|II − IE|, 2.25%·IEA) + dividends · SC = max(OI, OE) + max(FI, FE) · FC = |P&L_TB| + |P&L_BB|

BI business indicator — an income-statement proxy for operational scale · ILDC interest, lease & dividend component: II / IE interest income / expense, capped at 2.25% of IEA interest-earning assets · SC services component: OI / OE other operating income / expense, FI / FE fee income / expense — taken as max, not net · FC financial component: absolute net trading- and banking-book P&L. All three-year averages.
Business indicator component — marginal buckets:

BI ≤ €1bn 12%

€1bn < BI ≤ €30bn 15%

BI > €30bn 18%
Apply the multiplier: ORC = BIC × ILM, ILM = 1 — historic losses do not scale the charge in the EU.

Worked examples → section detail ⧉.

Loss data regime

Art 316–323

Losses no longer drive capital — but must still be collected, governed and disclosed.

Mandatory loss data set where BI ≥ €750m: events ≥ €20,000, ten-year history, annual disclosure of net losses ≥ €100,000.
Exclusion of divested activities / merger adjustments with supervisory permission.
Governance: op-risk framework, ICT & outsourcing risk aligned with DORA; loss-register completeness is an audit topic, not a modelling one.

Consequences

design intent

What the switch from models to income actually changes.

Capital is a function of income size, not loss history — fee-rich models (custody, payments, asset servicing) structurally penalised vs former AMA outcomes.
Marginal design: BI growth compounds at 15→18% for the largest.
ILM = 1 is an EU deviation — Basel permits loss sensitivity; the EU chose comparability over risk sensitivity.

Securitisation — hierarchy & steps

Art 254 ff

Tranche capital interpolated from pool capital — steeper the closer you sit to first loss.

15/10%senior RW floor / STS

p 1.0/0.5SEC-SA / STS

Choose the approach (hierarchy): SEC-IRBA if K_IRB computable → SEC-SA (EU preference) / SEC-ERBA / IAA where applicable → else 1250%.
Compute pool capital: K_IRB (IRB pool) or
K_A = (1−W)·K_SA + W·0.5 · W = delinquent share

K_SA capital ratio of the underlying pool under SA (avg RW × 8%) · K_IRB same under IRB, incl. expected loss · W share of the pool that is delinquent — delinquents assumed to need 50% capital · K_A the blended pool capital anchor
Set supervisory p — SEC-IRBA: formula on N, LGD, tranche maturity (min 0.3) · SEC-SA: p = 1.0, STS 0.5 · resecuritisation: SEC-SA only, p = 1.5.
Apply the SSFA-type formula between attachment A and detachment D: tranches with D ≤ K_A → 1250%; tranches above K_A → supervisory exponential decaying over p·K_A; straddling → thickness-weighted blend. SEC-ERBA where used: RW = f(CQS, seniority, maturity interpolated 1–5y, tranche thickness) — from 15% senior CQS1 up to 1250% at the bottom of the scale (grids: Art 263).
A attachment point — losses (as % of pool) at which the tranche starts absorbing · D detachment point — where it is wiped out · p supervisory parameter spreading capital up the structure: higher p = more capital pushed into senior tranches

Worked examples → section detail ⧉.

Apply floors & caps:

RW floor — senior / STS senior	15% / 10%
Resecuritisation floor	100%
Overall cap	≤ K of underlying pool (senior look-through)
Output-floor S-TREA TRANS	SEC-SA p halved: 0.5 · STS 0.25

SRT assessment — significant risk transfer tests (mezzanine/first-loss) unchanged in principle; EBA SRT framework under review.

Crypto-assets

Art 501d TRANS

A bridge regime: traditional-in-token-form gets normal weights, everything else near-deduction.

Category	Treatment
Tokenised traditional assets	as underlying
ARTs backed by traditional assets	250%
Other (incl. unbacked)	1250% + exposure limit

Pending permanent BCBS SCO60-based implementation; MiCA governs issuance/conduct; aggregate exposure limits and disclosure apply.

Worked examples → section detail ⧉.

Parameter calendar

what changes when

Every date on this chart that moves a number, in one table.

Date	Change
01·01·2027	Floor x → 60% · FRTB own funds apply (Art 495v multiplier option to end-2029) · CRD6 third-country branch regime
01·01·2028	x → 65%
01·01·2029	x → 70%
01·01·2030	x → 72.5% steady state · equity RW phase-in complete · FRTB multiplier expires
31·12·2032	Floor transitionals lapse: mortgage 10/45% split · unrated-corporate 65% · halved p-factors

SREP & the full capital stack

Art 97–104b CRD · EBA SREP GL

Supervision turns the rulebook into a firm-specific number — and stacks it into a tower.

Four assessment blocks, scored 1–4: business model viability · internal governance & risk management · risks to capital (vs ICAAP) · risks to liquidity (vs ILAAP).
Set P2R — binding, for risks not (fully) in Pillar 1; same tiering as minima (≥ 56.25% CET1). Floored risks stripped out to avoid double-counting.
Set P2G — sized from EU-wide stress-test depletion; sits on top of everything, CET1-quality expectation; breach → supervisory dialogue, not automatic MDA.

The stack, bottom to top

P2G guidance · CET1 · stress-test sized

Combined buffer CCoB 2.5% + CCyB + G/O-SII + SyRB · MDA trigger

P2R firm-specific · binding · ≥56.25% CET1

Tier 2 2.0%

Additional Tier 1 1.5%

CET1 minimum 4.5% — Pillar 1 total 8.0%

Worked examples → section detail ⧉.

Every layer is a % of the floored TREA (02) — which is how the output floor quietly raises the whole tower, not just the minimum.

IRRBB — supervisory outlier tests

Art 84, 98 CRD · EBA RTS/GL

Banking-book rate risk is capitalised through supervision, not Pillar 1 — but the thresholds are hard numbers.

15%ΔEVE / Tier 1

5%ΔNII / Tier 1

EVE SOT — run six supervisory rate scenarios (parallel ±, steepener, flattener, short-rate ±) with prescribed currency-specific shocks and a post-shock floor rising from −150bp; decline in economic value of equity > 15% of Tier 1 → supervisory outlier.
NII SOT — two parallel scenarios, one-year horizon, constant balance sheet; large decline in net interest income (> 5% of Tier 1 per EBA RTS) → outlier.
Consequence — outlier status triggers supervisory scrutiny and typically P2R/P2G measures or hedging expectations; CSRBB assessed qualitatively alongside.

Worked examples → section detail ⧉.

Pillar 3 — disclosure

Art 431–455 CRR

The market-discipline leg: what must be published, how often, by whom.

Institution class	Frequency
Large — listed	annual full · semi-annual subset · quarterly key metrics
Large — non-listed	annual full · semi-annual subset
Other — listed / non-listed	annual full · key metrics semi-annual / annual
Small & non-complex (SNCI)	annual key metrics only

Content set (EBA ITS): own funds & ratios, RWEA by risk type, output-floor split (U-TREA vs S-TREA — new under CRR3), leverage, liquidity, credit quality & forbearance, remuneration.
ESG disclosure for all institutions, proportionate by size — transition & physical risk, taxonomy alignment (GAR), scope 3 phased.
EBA Pillar 3 data hub centralises publication — disclosures filed to EBA and published centrally, replacing scattered PDFs; SNCIs first.
ESG risks also enter SREP scoring and transition-plan review (CRD6) — supervision leads, Pillar 1 treatment still absent.

TREA: — total risk exposure amount: the RWA total all ratios divide by.
CET1 / AT1 / T2: — capital tiers by loss-absorbing quality.
P2R / P2G: — Pillar 2 requirement (binding) / guidance (expected).
MDA: — maximum distributable amount: payout cap on buffer breach.
AVA: — additional valuation adjustment (prudent valuation, CET1 deduction).
OCI: — other comprehensive income.
LREM: — leverage ratio exposure measure.
TLAC / MREL: — gone-concern loss-absorbing capacity (global / EU firm-specific).
CCoB / CCyB: — capital conservation / countercyclical buffer.
G-SII / O-SII: — global / other systemically important institution (buffers).
SyRB / CBR: — systemic risk buffer / combined buffer requirement (the four together).
DTA / DTL: — deferred tax asset / liability.

U-TREA: — TREA using approved internal approaches.
S-TREA: — the same book on all-standardised approaches.
x: — transitional floor factor (Art 465(1)): 50% → 72.5% by 2030; 55% in 2026.
Phase-in cap: — optional limit: floored TREA ≤ 1.25 × unfloored during transition.

EAD: — exposure at default: the amount at risk when default happens.
CCF: — credit conversion factor: how much of an undrawn line counts as drawn.
CRM: — credit risk mitigation (collateral, guarantees).
RC / PFE: — replacement cost / potential future exposure (SA-CCR).
TH · MTA · NICA: — margin threshold, minimum transfer amount, net independent collateral.
V / C: — market value of the netting set / collateral held (SA-CCR).
SD · δ · MF · SF: — supervisory duration, delta, maturity factor, supervisory factor (SA-CCR add-ons).
OEM / EPE: — original exposure method / expected positive exposure (IMM).
SFT: — securities financing transaction (repo, securities lending).
LGD*: — supervisory LGD for collateralised F-IRB exposures.
GCC: — group of connected clients (control or economic interdependency).
ECRA / SCRA: — external ratings approach / standardised grading of unrated banks.
CQS: — credit quality step (1 = best) mapped from ratings.
ETV: — exposure-to-value at origination.
IPRE / ADC: — income-producing real estate / land acquisition–development–construction.
PD / LGD / M: — probability of default, loss given default, effective maturity.
QRRE: — qualifying revolving retail exposures (cards).
R: — asset correlation in the IRB formula.
EL / EL_BE: — expected loss / best-estimate EL on defaulted assets.
Slotting: — supervisory categories replacing PD estimates for specialised lending.
QCCP: — qualifying central counterparty (EMIR-authorised/recognised).

SbM: — sensitivities-based method.
WS: — weighted sensitivity (RW × s).
DRC / JTD: — default risk charge / jump-to-default amount.
RRAO: — residual risk add-on.
ES / LH: — expected shortfall / liquidity horizon.
RFET / NMRF: — risk-factor eligibility test / non-modellable risk factor.
SES: — stress-scenario capital for NMRFs.
IMCC: — internally modelled capital charge (modellable factors).
RTPL / HPL: — risk-theoretical vs hypothetical P&L (attribution test).
m_c: — IMA multiplier, ≥ 1.5 plus backtesting add-on.
GIRR / CSR: — general interest rate risk / credit spread risk classes.
s_k / RW_k / K_b: — sensitivity, its supervisory weight, bucket capital (SbM).
ρ / γ: — intra-bucket / cross-bucket supervisory correlations.
CVR: — curvature risk measure: stressed non-linear loss beyond delta.

CVA: — credit valuation adjustment: expected loss from counterparty spread risk.
SCVA: — standalone CVA capital per counterparty (BA-CVA).
DF: — supervisory discount factor on maturity.
SNH / IH / HMA: — single-name hedge, index hedge, hedging misallocation penalty.
β: — 0.25 floor on hedge recognition.
DS: — 0.65 discount scalar on the BA-CVA charge (2020 recalibration).
m_CVA: — SA-CVA multiplier, set at 1.

BI / BIC: — business indicator / its coefficient-weighted component.
ILDC · SC · FC: — interest-lease-dividend, services, financial components of BI.
IEA: — interest-earning assets (caps ILDC at 2.25%).
II / IE · OI / OE · FI / FE: — interest, other operating, and fee income / expense pairs in the BI.
ORC: — operational risk capital requirement (= BIC in the EU).
ILM: — internal loss multiplier; EU fixes it at 1.

K_IRB / K_SA / K_A: — pool capital under IRB / SA / delinquency-adjusted SA.
W: — share of delinquent underlying exposures.
p: — supervisory thickness parameter spreading capital up the structure.
A / D: — tranche attachment / detachment points.
SRT: — significant risk transfer.
STS: — simple, transparent, standardised label.
LTA / MBA / FBA: — look-through / mandate-based / fall-back fund approaches.
ART: — asset-referenced token (MiCA).
CIU: — collective investment undertaking (fund).

EVE: — economic value of equity: PV of banking-book assets minus liabilities.
NII: — net interest income.
SOT: — supervisory outlier test.
CSRBB: — credit spread risk in the banking book.
SREP: — supervisory review and evaluation process.
ICAAP / ILAAP: — internal capital / liquidity adequacy assessment processes.
SNCI: — small & non-complex institution (≤ €5bn a.o. criteria).

Block	Direction of travel
Credit SA	More risk-sensitive (to serve as a credible floor)
Credit IRB	Constrained: scope cut, inputs floored
Market	Rebuilt (FRTB), own funds from 2027
CVA	Internal models abolished
Op risk	One formula, models abolished
Aggregation	Output floor, single stack
Numerator	Essentially untouched

Area	Change
Own-funds criteria	Substantively unchanged — the 2010 architecture held
Interaction	Single-stack output floor: floored TREA now drives P2R, buffers, MREL alike
Perimeter	Consolidation updates: fintech/ancillary undertakings, holding-company approval regime refinements

Topic	Where the answer is
Instrument terms	EBA Own Funds RTS + the standing EBA Q&A practice — de facto case law
MREL calibration	SRB MREL policy (annual); LAA + RCA + market-confidence charge
AT1 post-2023	Trigger and write-down mechanics now a disclosed pricing factor; EU statements reaffirmed creditor hierarchy

Topic	State
Positive-neutral CCyB	Spreading practice, not EU law — rates are national decisions with 12-month implementation lags
SyRB	Sectoral use (RRE) established; overlap control with Art 458 measures
Usability	Macroprudential review's standing item: shift weight from fixed to releasable

Before	Now
Basel I floor (Art 500, largely spent) — no meaningful RWA floor	Output floor on total TREA, phase-in 50→72.5%
IRB benefit unlimited in EU-specific requirements	Single stack: floor propagates into P2R, buffers, MREL
—	Transitional S-TREA calibrations to 2032 (mortgages, unrated corporates, p-factors)

Clearing member → QCCP	2%
Client via member (conditions met / partial)	2% / 4%
Non-QCCP	bilateral SA

Transitional (S-TREA only)	Calibration	Until
Residential mortgages (conditions met)	10% / 45% split at 55% ETV	31·12·2032
Unrated investment-grade corporates	65% RW	31·12·2032
Securitisation p-factors	0.5 non-STS / 0.25 STS	31·12·2032
Cap on floored TREA	≤ 1.25 × unfloored	member-state option

Area	Change
SA real estate	Loan-splitting & ETV whole-loan tables; IPRE and ADC own classes; valuation capped at origination
SA counterparties	SCRA; CQS2 institutions 30%; due-diligence override; equity 250/400% (phase-in); subordinated 150%; UCC CCF 10% (phased)
IRB scope	A-IRB withdrawn for large corporates & institutions; equity out; partial use liberalised
IRB parameters	Input floors; F-IRB senior unsecured LGD 45→40%; 1.06 scalar deleted
CRM	Own-estimate volatility adjustments abolished; F-IRB secured LGD* recalibrated
New	Retail FX multiplier ×1.5 (Art 123a); crypto transitional (Art 501d)

Before	Now
CEM / Standardised Method	SA-CCR only (below thresholds: simplified/OEM)
UCC CCF 0%	10% (phased)
Own-estimate haircuts	Abolished — supervisory haircuts only

Area	CRR2 → CRR3
Mortgages	35% flat → loan-splitting 20/75 or ETV whole-loan; IPRE own table
Banks	Unrated: flat by sovereign → SCRA A/B/C; CQS2 30%
Corporates	Unrated IG 65% only inside the floor transitional; SME 85% retained via factor
Retail	Transactor 45%; currency-mismatch ×1.5
Equity/sub-debt	100% → 250/400% (phase-in) / 150%
Off-balance	UCC 0% → 10% (phased); CCF ladder rebuilt

Change	Effect
A-IRB withdrawn (large corp., institutions)	Own-LGD advantage gone where defaults are unobservable
Equity → SA	All internal equity models retired
Input floors	PD 5bp (10bp QRRE revolvers), LGD/EAD floors per class
1.06 scalar deleted	−5.7% RWA on all IRB books, mechanically
F-IRB senior unsecured LGD 45→40%	Partial offset to the constraints
Partial use liberalised	Permanent SA carve-outs by class without losing IRB elsewhere

Before	Now
VaR 99%, 10-day + stressed VaR + IRC	ES 97.5% stressed, liquidity horizons 10–120d, IMA-DRC
Bank-wide model approval	Desk-level approval, earned quarterly (PLAT, backtesting)
Intent-based boundary	Presumptive lists, reclassification lock, IRT desks
Old SA (building-block)	Sensitivities-based A-SA + DRC + RRAO; SSA for small books
All risk factors modellable	RFET; NMRFs charged via stress scenarios

Old SA	A-SA
Notional blocks, crude offsets	Sensitivities, correlated aggregation, three scenarios
No curvature concept	Explicit curvature add-on
Specific risk charges	DRC on net JTD + RRAO

Basel 2.5 IMA	A-IMA
VaR + sVaR + IRC, bank-wide	Stressed ES + SES + DRC, per desk
10-day everything	Liquidity horizons 10–120d
All factors modellable	RFET-gated; NMRFs to SES
Approval sticky	PLAT/backtesting quarterly; fallback priced

Before	Now
Advanced CVA (VaR-based, own models) + old standardised	Internal models abolished: SA-CVA (approval) / BA-CVA / simplified
Exemptions invisible	Exemptions retained but computed & reported
Basel 2017 calibration	2020 recalibration: DS = 0.65 · m_CVA = 1.0 · broader index hedges

BI ≤ €1bn	12%
€1bn < BI ≤ €30bn	15%
BI > €30bn	18%

Old (CRR2 Art 384)	BA-CVA
Standardised CVA: bond-equivalent VaR formula	Portfolio-consistent SCVA aggregation, explicit hedge terms
Hedges crudely recognised	r_hc ladder, index recognition, HMA, β floor

Before	Now
BIA (15% GI) / TSA (12–18% by line) / AMA (own models)	One SMA: BIC on the BI, ILM = 1
Loss data an AMA input	Loss data a universal obligation ≥ €750m BI, disclosed
Op-risk modelling for capital	Modelling survives only in Pillar 2 / ICAAP

Repealed	Replaced by
BIA (15% of gross income)	One SMA: BIC × ILM, ILM = 1 (Art 314(2))
TSA (12–18% by business line)
AMA (internal models)

Area	Change
Framework	Hierarchy unchanged (2019 architecture)
Output floor interaction	p halved in S-TREA: 0.5 non-STS / 0.25 STS — transitional to 2032
Crypto	Art 501d transitional weights; disclosure & limits

Area	Change
SREP scope	ESG risks and transition plans enter the methodology; supervisory powers extended
Pillar 3	Centralised EBA data hub; floor-split disclosure (U-TREA vs S-TREA); ESG templates
Perimeter	Third-country branch regime harmonised (CRD6)

The capital pipeline — how the blocks connect

Regulatory capital & overall requirements

Capital instruments — eligibility by tier

Own funds — and the bridge to TLAC / MREL

Capital ratios & minima

Buffers (CET1)

Leverage ratio

Large exposures

Output floor — the binding aggregation rule

Mechanism

Phase-in & cap

Transitional S-TREA calibrations

Credit risk

Exposure value & CRM

Credit risk mitigation

RWA under the Standardised Approach

RWA under the IRB Approach (constrained)

Collective investment undertakings

RWA adjustments & multipliers

CCP exposures

Market risk — FRTB OWN FUNDS FROM 01·01·2027

Trading-book definition & approach choice

Alternative Standardised Approach (A-SA)

Alternative Internal Model Approach (A-IMA)

CVA risk

Scope & approach selection

BA-CVA — calculation

SA-CVA — calculation

Operational risk

Calculation

Loss data regime

Consequences

Securitisation · crypto-assets · calendar

Securitisation — hierarchy & steps

Crypto-assets

Parameter calendar

Supervision, disclosure & Pillar 2 NOT PART OF TREA

SREP & the full capital stack

IRRBB — supervisory outlier tests

Pillar 3 — disclosure

Common terms — by section

What the framework is for

How EU law is layered

Five accords in one line each

Framework-level obligations

Reading the pipeline — one bank, end to end

CRR3 in one table

Sources

2008: capital that did not absorb losses

The lineage

Operating the numerator

What CRR3 changed here

Where the law lives

Open items

Why tiering, why deductions

Instrument-by-instrument discipline

Practice notes & sources

Capital you are allowed to lose

The measure that cannot be modelled down

The one limit that is not capital

The failure it answers

Intent, in one line

How it was fought over

What a bank must actually operate

When the floor binds — the switch year, visually

The cap, and the cliff

CRR2 → CRR3

Where the law lives

Open items

Single stack, and why it was the fight

What S-TREA actually contains

The Basel II bargain — granted, then constrained

The lineage

Standardised approach — what "no model" still requires

IRB — the approval burden, grouped

The same mortgage, three ways — the floor debate in one chart

Corporate, counterparty & funds

CRR2 → CRR3, the working list

Where the law lives

Common findings