How OTC Portfolio-Compression Algorithms Shrink Gross Derivatives Notional Without Changing Market Risk: Tear-Ups, Replacement Trades, Tolerances, Multilateral Optimisation and Reconciliation

Reader question: A dealer can have thousands of swaps whose market risks largely cancel each other, yet the trades remain on the books as huge gross notionals and thousands of operational line items. How can an algorithm remove the redundant trades without accidentally changing the dealer’s economic market risk?

The answer is portfolio compression. A compression engine identifies combinations of existing derivatives that can be terminated, reduced or replaced with a much smaller set of trades while keeping each participant’s agreed market-risk, counterparty-risk and cash-payment changes inside explicit tolerances.

The mathematical problem is not simply “find opposite notionals and cancel them”. Real portfolios contain different coupons, maturities, curves, counterparties, collateral terms and clearing states. The engine therefore solves a constrained optimisation problem:

reduce gross notional and/or line items as much as possible, subject to preserving the agreed economic risk state within tolerance.

This is a post-trade efficiency algorithm, not a trading recommendation and not a claim that notional itself is economic loss.

What this page owns — and what it does not

This article owns:

eligible derivatives population → risk representation → compression constraints and tolerances → tear-up/replacement proposal → participant acceptance → booking and reconciliation → reduced gross notional/line items.

It does not replace swap valuation and curve construction, variation-margin calculation, collateral optimisation, or Basel leverage-ratio exposure calculations. Compression changes the trade population while trying to preserve the economic exposure represented by that population.

This is mathematical and infrastructure education, not advice to terminate, novate, compress or rebalance any real derivatives portfolio.

Why redundant derivatives accumulate

Imagine a bank enters a five-year pay-fixed swap today. Six months later it wants to neutralise the exposure and enters a receive-fixed swap with similar economics.

From a market-risk perspective, the two positions may largely offset.

Operationally, however, both contracts remain alive. Each can still require:

• valuation;
• daily collateral calculations;
• cash-flow generation;
• confirmation and reconciliation;
• regulatory reporting;
• capital and leverage exposure calculations;
• lifecycle-event processing.

Repeat this process for years across many dealers and the market can accumulate enormous gross notional even when net risk is far smaller.

Gross notional is not the same as market risk

Suppose a dealer has:

• pay-fixed $100m swap;
• receive-fixed $100m swap;
• same maturity, same coupon, same floating index, same counterparty and same legal netting set.

If the trades are otherwise identical, their first-order and higher-order market risks can cancel nearly perfectly.

Gross notional = $200m.

Net directional notional = approximately zero.

Economic risk can be close to zero while operational gross remains large.

This is the basic opportunity compression tries to capture.

Exact offset is the simplest case

If two trades are exact legal and economic opposites, a bilateral compression can terminate both.

Before:

Trade A + Trade B = 0 economic position.

After:

no trades = 0 economic position.

Gross notional falls from $200m to zero without changing the economic state.

Real portfolios are usually harder because trades are only approximately offsetting.

Compression as an optimisation problem

Represent the original portfolio by trades i = 1…n and a candidate compressed portfolio by trades j = 1…m, usually with m < n.

One stylised objective is:

minimise λ_N × GrossNotional(new) + λ_L × LineItems(new).

subject to constraints such as:

|Risk_r(new) − Risk_r(old)| ≤ ε_r

for each material risk measure r.

Additional constraints can limit:

• mark-to-market change;
• cash settlement required at compression;
• counterparty exposure change;
• currency exposure;
• curve and tenor sensitivities;
• legal-netting-set movement;
• clearing-member or client-account boundaries;
• participant-specific exclusions.

The exact commercial optimisation engines are proprietary. The public mathematical role is clear: minimise redundant trade mass while respecting stated risk boundaries.

Risk-neutral does not mean every trade survives unchanged

A compression exercise can terminate several trades and create one or more replacement trades.

The replacement portfolio can have different individual coupons, notionals or line items while preserving the agreed portfolio-level risk metrics.

This is analogous to simplifying an algebraic expression:

3x − 2x + 5y − 5y = x.

The term count changes dramatically while the mathematical function remains the same.

In derivatives, however, “same” is defined by tolerances over financial risk measures rather than exact symbolic equality.

Step 1: define the eligible population

A compression engine first determines which trades can participate.

Eligibility can depend on:

• product type;
• currency;
• cleared versus uncleared status;
• CCP or legal netting set;
• remaining maturity;
• trade lifecycle state;
• documentation;
• participant opt-ins and exclusions.

A trade that is economically offsetting but legally outside the compression set cannot simply disappear.

Step 2: normalise trade economics

Trades from different systems may encode equivalent economics differently.

For an interest-rate swap the normalised state can include:

• pay/receive direction;
• notional schedule;
• fixed coupon;
• floating index and spread;
• effective and maturity dates;
• payment frequency;
• day-count conventions;
• business-day conventions;
• collateral/discounting context.

If date conventions or curve identifiers are mismatched, two trades that look offsetting by notional may not be economically offsetting.

Step 3: calculate the portfolio risk state

Before proposing any tear-up, the engine needs a risk representation of the original portfolio.

A common educational representation is a sensitivity vector:

R = (PV, DV01₁, DV01₂, …, FXDelta, Vega, …).

For cleared interest-rate swaps, risk can be represented by sensitivities across curve tenors and currencies. Other products require different factors.

The compressed portfolio must remain close to the original risk vector according to agreed tolerances.

Step 4: set participant tolerances

CFTC rules explicitly recognise that participants can establish counterparty, market, cash-payment and other risk tolerances for compression exercises.

This is important because exact equality can make compression nearly impossible.

For example, a participant might permit:

• PV change no greater than $5,000;
• DV01 change per tenor no greater than $500;
• no increase in exposure to a named counterparty;
• compression cash payment no greater than $2m;
• no creation of trades beyond a particular maturity.

These are illustrative teaching tolerances, not market recommendations.

Why tighter tolerances reduce compression capacity

If every risk measure must match to the cent and every old trade term must be preserved, only exact offsets can be removed.

If small economically immaterial differences are permitted, many more combinations become feasible.

There is therefore a mathematical trade-off:

compression efficiency ↑ as tolerance flexibility ↑, but residual model/market differences can also ↑.

A good system exposes this trade-off instead of hiding it.

Step 5: search for a smaller feasible portfolio

At a high level, the optimisation asks:

Which subset of old trades can be terminated, and which replacement trades—if any—can be introduced, so all participant constraints remain satisfied?

This can become a large optimisation problem because:

• many institutions participate simultaneously;
• each institution has different tolerances;
• bilateral legal exposures matter;
• one participant’s acceptable replacement can be another’s unacceptable exposure;
• trade sizes and coupons may need discrete or market-compatible values.

Commercial multilateral compression systems solve this at industrial scale. Public education should explain the constrained optimisation structure without pretending to reveal a vendor’s private solver.

Bilateral compression versus multilateral compression

Bilateral compression considers trades between two counterparties.

Multilateral compression considers a network of counterparties simultaneously.

Multilateral compression can find cycles that bilateral compression cannot.

Example:

• Bank A pays fixed $100m to Bank B;
• Bank B pays fixed $100m to Bank C;
• Bank C pays fixed $100m to Bank A;

Each bilateral pair has one directional trade, so no pair has an exact offset.

Across the network, however, the circular market risk can be close to redundant if all other terms align.

A multilateral engine can potentially collapse the cycle while preserving each participant’s net economic state.

Graph interpretation

Represent each institution as a node and each economically compatible derivative exposure as a directed edge.

Compression searches for:

• opposite edges;
• cycles;
• redundant paths;
• replaceable edge combinations;

subject to risk and legal constraints.

This creates a bridge between financial mathematics and network-flow optimisation.

Tear-up-only compression

The most conservative form terminates existing trades without creating new ones.

This works well for exact or near-exact offsets.

Advantages:

• simple lifecycle result;
• no new trade terms;
• easy audit trail.

Limit:

the achievable notional reduction can be much smaller than with replacement-trade optimisation.

Replacement-trade compression

A more flexible process can terminate many old trades and replace them with fewer trades that recreate the residual risk.

Example:

A portfolio contains five swaps with nearby maturities and coupons. Their aggregate curve sensitivity can perhaps be represented by two new swaps.

Before:

5 line items, gross notional $500m.

After:

2 line items, gross notional $180m.

if the new portfolio satisfies all agreed tolerances.

Coupon blending in cleared swaps

Clearing houses can support compression methods that combine trades with compatible economics and replace them with fewer positions, sometimes using blended coupons or revised trade structures.

LCH SwapClear publicly describes solo compression, blended-rate compression and multilateral compression. Eurex describes functions including downscaling, upscaling, trade revision and trade refactoring within its cleared OTC environment.

This shows that “compression” is a family of algorithms rather than one single cancellation rule.

Cleared compression and uncleared compression have different constraints

Inside one CCP, counterparty structure is standardised around the clearing house and account architecture.

In uncleared portfolios, bilateral legal agreements, thresholds, collateral terms and counterparty limits are more heterogeneous.

The optimisation problem is therefore structurally different even if both aim to reduce notional and line items.

Regulation recognises compression as risk management

U.S. CFTC rules require swap dealers and major swap participants to establish policies and procedures for bilateral and multilateral portfolio compression when appropriate.

EU rules under EMIR require certain counterparties with 500 or more uncleared OTC derivative contracts outstanding with a counterparty to have procedures to analyse the possibility of portfolio compression regularly and at least twice per year, and to engage in compression when appropriate or be able to explain why it is not.

These rules are not mathematical formulas for the optimiser. They define governance around whether compression should be considered and controlled.

EMIR 3 is changing the surrounding PTRR framework

In February 2026, ESMA consulted on regulatory technical standards for post-trade risk reduction services under EMIR 3, covering compression, portfolio rebalancing and basis-risk optimisation and the conditions for a clearing-obligation exemption.

That consultation makes one update trigger explicit: post-trade optimisation rules continue to evolve. Public systems should preserve the rule version rather than treating today’s exemptions and controls as permanent.

Why compression can reduce capital or balance-sheet cost

Some regulatory exposure measures depend partly on gross derivatives positions or trade counts rather than only net market risk.

Reducing redundant notional can therefore reduce:

• leverage exposure in some frameworks;
• counterparty exposure measures in some structures;
• operational and reconciliation load;
• line-item charges at clearing houses;
• collateral and funding complexity.

But the effect is rule-specific. A compression engine should not promise that every dollar of notional reduction creates a fixed amount of capital relief.

A simple compression-efficiency metric

Define:

Notional Compression Ratio = (GrossOld − GrossNew) / GrossOld.

If gross notional falls from $10bn to $3bn:

ratio = (10 − 3)/10 = 70%.

Also define:

Line-Item Reduction = (TradesOld − TradesNew) / TradesOld.

These metrics measure operational compression, not economic risk reduction.

Risk-preservation metrics

For each risk dimension r:

ResidualRisk_r = Risk_r(new) − Risk_r(old).

A successful compression requires:

|ResidualRisk_r| ≤ Tolerance_r.

This creates a clear audit test: the new portfolio is only acceptable if every required residual lies within the participant’s approved boundary.

Compression cash payments

Terminating trades at non-zero market value can create cash payments between participants.

A participant can therefore be economically risk-neutral after compression but still face a large one-time cash movement.

This is why cash-payment tolerance is distinct from market-risk tolerance.

A solver that preserves DV01 perfectly but creates an unacceptable $500m immediate payment can still be operationally unusable.

Counterparty exposure can move even if market risk does not

Suppose Bank A’s rates exposure is unchanged after compression, but replacement trades shift more positive mark-to-market toward Bank B and away from Bank C.

Market risk may be stable while counterparty credit risk changes.

Therefore a multilateral engine must distinguish:

market-risk neutrality

from:

counterparty-exposure neutrality.

They are not the same constraint.

Compression and rebalancing are not identical

Portfolio compression primarily targets redundant gross trades and notional while preserving risk.

Portfolio rebalancing can intentionally redistribute counterparty risk or other exposures while keeping market risk broadly controlled.

ESMA’s 2026 PTRR consultation treats compression, rebalancing and basis-risk optimisation as distinct service types.

This distinction prevents keyword cannibalisation and conceptual confusion.

Evidence polarity: what supports confidence?

Evidence for a reliable compression result includes:

• the post-compression trade population is smaller or lower-notional as intended;
• all original and replacement trades are traceable;
• PV and required sensitivities remain within approved tolerances;
• cash payments reconcile to terminated market values;
• counterparty and legal-netting constraints remain satisfied;
• all participants accepted the proposal under the applicable process;
• clearing-house or bilateral books reconcile after booking.

Evidence against confidence includes:

• large unexplained change in DV01, vega or FX delta;
• reduced notional but increased unauthorised counterparty exposure;
• replacement trades outside the legal or clearing eligibility set;
• cash settlement exceeding agreed tolerance;
• old trades terminated before replacement trades are safely booked;
• different participants holding inconsistent final compression states.

Counterexample: cancelling equal notionals can change risk

Consider:

• pay-fixed $100m five-year swap at 2%;
• receive-fixed $100m ten-year swap at 2%.

The notionals are equal and directions opposite.

The maturities are different, so duration and curve sensitivities are not.

Blind cancellation would materially change market risk.

Counterexample: same maturity can still fail to offset

Two five-year swaps can reference different floating indices, payment frequencies, currencies or collateral discounting curves.

Same notional and maturity are insufficient tests of economic equivalence.

Counterexample: lower line count can increase operational risk

If a complex compression proposal creates replacement trades that downstream systems cannot book correctly, the line-item reduction can make the operational state worse rather than better.

Compression is only successful when the new state is both economically acceptable and operationally valid.

Counterexample: risk-neutral today can drift tomorrow

If the compression engine preserves only today’s PV but ignores curve sensitivities, two portfolios can have the same current value and very different future responses to rates.

A meaningful tolerance set must preserve the risk dimensions relevant to future valuation, not only the current price.

Weak links in implementation

Trade normalisation error. Equivalent fields are mapped inconsistently across source systems.

Stale valuation snapshot. Participants optimise against market data from different times.

Risk-basis mismatch. One participant uses OIS DV01 while another compares par-rate DV01.

Tolerance-unit error. A basis-point tolerance is interpreted as a percentage-point tolerance.

Legal-set leakage. Trades from non-compatible netting sets are compressed together.

Replacement-booking race. Tear-ups become effective before replacement trades are confirmed.

Cash-payment sign error. A compensation amount is reversed.

Incomplete downstream cancellation. Old trades survive in one ledger after being terminated elsewhere.

Diagnostics: how to test the engine

• exact-offset test: two identical opposite trades should compress to zero where legally eligible.
• maturity-mismatch test: opposite notionals with different maturities must not be treated as exact offsets.
• risk-vector test: independently calculate pre- and post-compression PV, DV01, vega and other required factors.
• tolerance-boundary test: move one proposed replacement just inside and just outside a tolerance and verify acceptance changes correctly.
• cash-cap test: preserve market risk while forcing excessive termination cash and require rejection.
• counterparty test: hold total market risk constant but shift bilateral exposure beyond a participant limit and require rejection.
• cycle test: use a three-bank circular exposure that cannot be compressed bilaterally but can be reduced multilaterally.
• atomic-booking test: simulate replacement-trade booking failure and verify old trades are not irreversibly terminated into an inconsistent state.
• reconciliation test: participant, CCP and trade-repository populations must agree after the cycle.
• independent rerun test: the same inputs, tolerances and optimisation version should reproduce the same accepted economic state or a documented equivalent optimum.

What would falsify confidence?

Confidence should be withdrawn if a compressed portfolio breaches agreed risk tolerances, if gross notional falls only because trades disappeared from one system but not their legal records, if cash payments cannot be reconstructed from trade values, if replacement trades change unauthorised counterparty exposure, or if downstream books cannot reconcile to the compression result.

Alternatives and limits

Simple bilateral termination is easier but finds fewer opportunities.

Multilateral compression can find network-wide reductions but requires more coordination and constraint handling.

Portfolio rebalancing intentionally changes counterparty distribution rather than only eliminating redundant trade mass.

Basis-risk optimisation can change the risk-factor composition within specified bounds.

Compression cannot eliminate genuine directional market risk without changing the portfolio’s economics. If a bank is truly long rates duration, some residual trade must continue to represent that exposure unless the bank chooses to hedge or unwind it economically.

How this connects to the surrounding knowledge estate

Multi-curve swap valuation provides the PV and sensitivities that a rates compression engine must preserve. Variation-margin mechanics explain one operational cost of carrying many live trades. Leverage-ratio algorithms explain why gross derivatives structures can matter even when net market risk is small. Collateral optimisation is a different constrained optimisation problem applied after margin obligations exist.

Verification and update triggers

Preserve:

• compression-service and algorithm version;
• eligible trade population;
• valuation timestamp and market data;
• risk-factor definitions;
• participant tolerances;
• legal/clearing exclusions;
• proposed tear-ups and replacement trades;
• cash payments;
• participant approvals;
• post-cycle reconciliation evidence.

Revalidate after pricing-model changes, regulatory capital changes, EMIR/CFTC rule updates, clearing-house compression changes, new product types, counterparty agreement changes or any failed post-compression reconciliation.

Primary and high-quality references

• U.S. Commodity Futures Trading Commission, Portfolio Reconciliation and Compression Requirements, including the final rule discussion of bilateral and multilateral portfolio compression.
• European Union, Commission Delegated Regulation (EU) No 149/2013, Article 14 on portfolio compression.
• European Securities and Markets Authority, ESMA consults on post-trade risk reduction services under EMIR 3, 26 February 2026.
• LCH SwapClear, Enhancing Efficiency and Risk Management, describing cleared compression methods.
• Eurex Clearing, OTC derivatives transaction management, describing downscaling, upscaling, revision and refactoring compression functionality.
• OSTTRA, Portfolio Compression, describing multilateral compression across cleared and uncleared OTC products.

Educational boundary: This article explains public portfolio-compression mathematics and controls. It does not reveal any vendor’s proprietary optimisation algorithm, advise a real compression exercise, or determine the legal or regulatory treatment of any derivatives portfolio.