Skip to content
Corporate & Investment Banking My progress Français

Home / Capital Markets & Securities Services / Capital Markets / Module 13

13Derivatives Trading

Listed vs OTC · The two worlds of trading derivatives, how a bank makes markets, clears, margins, and manages the risk of a live book

Running a live book of options, futures and swaps, listed or OTC.

Reading time
2 h 00
Sections
53
Recall questions
32
Quiz questions
20
Simulators
1

Two threads run through everything

The business thread is price and risk. A trading desk earns its living by making a two-way price — a bid it will buy at, an offer it will sell at — and capturing the gap across thousands of trades, while warehousing the risk that results and hedging away the parts it is not paid to hold. Every idea in this module attaches to one of those two verbs: quote the price, manage the risk. Where the money comes from, where it leaks away, and where a book blows up are all consequences of how well a desk does these two things.

The technology thread is that a derivatives book is a computation that never stops. Prices, sensitivities, margin, collateral and limits must be recalculated continuously across every position in both worlds, kept consistent between the front office and independent risk, and pushed cleanly through clearing and regulatory reporting with no re-keying. The value — and most of the pain — lives in the interfaces between these systems and in holding listed and OTC positions in one coherent risk view. Read the business thread and the technology thread together; neither makes sense alone.

How to read this module

Read it front to back the first time. Each part builds on the last: the two worlds and the products in Part 1, the lifecycle and the mechanics of running a book in Part 2, the actors and roles around the desk in Part 3, the systems and data that carry it all in Part 4, and the risks, failures and transformation reshaping the business in Part 5. A running fictional desk — Meridian, a European bank's rates-and-equity-derivatives operation — appears throughout so the abstractions land on something concrete. Every figure attached to Meridian is illustrative.

Coloured boxes carry specific freight. A blue Key insight box crystallises an idea worth remembering. A navy Going deeper box adds detail for readers who want the next layer. A teal IT & data lens box names the systems and data behind a mechanic. An orange Watch out box flags a trap. A green Worked example box runs a number through. A light-blue In a nutshell box recaps each part, and a purple Check your understanding box lets you test yourself before moving on. A recurring Europe / France note grounds each theme in EU market structure — EMIR, MiFID II, Eurex, LCH and the French and European supervisors — because that is the context most readers of this module actually operate in.

What you will be able to do

You will understand what derivatives trading is, and how it differs from structuring

You will be able to say precisely what a trading desk owns — the price and the risk — and why that is a different job from designing a bespoke product, which belongs to a structuring desk. The distinction organises everything that follows.

You will read the two worlds — listed and OTC — as one trade-off

You will understand why the same three products trade in two marketplaces, what each world gives up to get what it wants, and why post-2008 regulation narrowed the gap without closing it. You will stop treating "listed" and "OTC" as jargon and start seeing the single trade-off underneath.

You will follow a trade end to end and know where the risk actually lives

You will be able to trace a derivative from execution through capture, clearing, margin and lifecycle to settlement, and to explain why the risk sits on the warehoused book rather than on any single ticket — and how a desk hedges the Greeks to keep only the risk it is paid to hold.

You will map every actor and system a trade touches

You will be able to name the exchanges, central counterparties, clearing brokers, venues, dealers and regulators that make the market work, and the front-to-back technology stack — pricing and risk engines, booking, clearing and collateral, reporting and surveillance — that a single trade flows through.

You will recognise the risk and transformation landscape

You will know the six ways a derivatives book takes risk, the handful of ways books actually blow up, how to tell a well-run desk from a lucky one, and where automation, cloud and machine learning genuinely move cost or risk — as opposed to merely rearranging the surface.

1The Two Worlds: What Derivatives Trading Is

Before any mechanics, three things must be clear: what a derivative actually is, what it means to trade one rather than design one, and why the same handful of products live in two very different marketplaces. Get these straight and the rest of the module is detail. Miss them and every later mechanic will feel arbitrary.

A derivative is a contract whose value is borrowed from something else

Start from zero: a derivative is an agreement between two parties whose value depends on the price of some other thing. That other thing — the underlying — might be an interest rate, a stock index, a currency pair, a barrel of oil, or the creditworthiness of a company. The contract itself has no independent worth. It is a bet, a hedge, or a transfer of risk, priced off the underlying and settled later. A future on the German government bond is worth something only because German bonds have a price; a EUR interest-rate swap matters only because EUR interest rates move. The value is derived. Hence the name.

Derivatives exist because the world constantly needs to move risk from those who hold it to those who will take it. A French exporter invoiced in dollars does not want to bet on the euro-dollar rate; it wants certainty, so it sells the dollars forward. A pension fund holding German bonds fears rising rates, so it sells bond futures to offset the loss. A hedge fund believes equity volatility is underpriced, so it buys options. The same instrument that removes an unwanted risk for one party is the vehicle a second party uses to express a view for profit. The market is the meeting place of these opposite motives.

Four uses cover almost everything a derivative is for. The first is hedging — removing or reducing an unwanted exposure to gain certainty. The second is speculation — taking a deliberate view for profit. The third is leverage — controlling a large exposure for a small upfront outlay, because you post margin rather than the full value. The fourth is price discovery — the market's live, aggregated opinion of what a rate, index or commodity is worth, visible in the derivative's price. A single contract can serve all four at once: the future the pension fund uses to hedge is the same future the hedge fund uses to speculate, and the price they strike is itself a piece of public price discovery.

The bank is the marketplace, not a spectator

A dealer bank does not merely use derivatives; it *runs the shop* where they trade. It quotes prices to clients and to other dealers, executes on exchanges and bilaterally, and then holds the resulting positions — the book — which it hedges continuously. When a client wants to shed a risk, the bank is the counterparty that absorbs it; when a client wants to take a view, the bank is the counterparty on the other side. The bank warehouses the net risk the market wants to transfer and provides the liquidity that lets everyone else act. That intermediary role — quoting, executing, warehousing, hedging — is the subject of this entire module.

Trading makes markets; structuring designs products

The single most useful distinction in this module is between trading and structuring, because they are constantly confused. A structuring desk designs bespoke solutions: it assembles building blocks — options, swaps, digitals, barriers — into a new, tailored product that answers a specific client problem. That work is covered in Module 4. A trading desk, the subject here, makes markets and manages risk: it quotes prices, executes flow, and hedges the book, in both the listed and OTC worlds. A structurer sells you a suit, cut to your measurements. A trader runs the shop, prices the cloth every second, and keeps the inventory balanced.

The trading desk owns the price and the risk. It quotes a bid and an offer to clients and other dealers, executes on exchanges and bilaterally, and then holds the resulting position and hedges it. Its profit comes from three honest sources — the bid-offer spread, the flow it sees, and skilled risk management — not from designing clever products. This is worth stating plainly because it reframes what "good" looks like: a great trading desk is not one that makes big directional bets and wins, but one that prices accurately, hedges cheaply, and turns steady flow into steady profit while surviving the tail.

Where the desk sits in the value chain

Trading sits in the middle of a chain that turns a client's need into a risk-managed position. At one end is the client or market need — a hedge, a view, or a flow to execute. Sales prices the client, takes the order, and routes it. Trading, in the middle, quotes, executes, warehouses and hedges the risk. Downstream, clearing and margin novate the trade to a central counterparty or keep it bilateral and collect margin daily, while risk and finance monitor limits, value-at-risk, profit and loss, collateral and capital. The desk is the pivot: everything upstream flows into it, and everything downstream depends on what it booked.

Link in the chain What it does Relationship to the desk
Client / market need A hedge, a view, or flow to execute The source of every trade
Sales Prices the client, takes and routes the order Feeds the desk its flow
Trading / market making Quotes, executes, warehouses and hedges risk The subject of this module
Clearing & margin Novates to a CCP or stays bilateral; margins daily Consumes what the desk books
Risk & finance Sets limits, computes VaR, P&L, collateral, capital The independent leash on the desk

Three product families, sorted by the shape of their risk

Everything a derivatives desk trades reduces to three families, and the honest way to sort them is by the shape of their payoff. Two families are linear — their value moves roughly one-for-one with the underlying. One family is non-linear — its value moves in a curved, asymmetric way that depends heavily on volatility. That distinction, linear versus non-linear, matters more than any product label, because it decides how the risk behaves and therefore how it must be hedged.

Futures and forwards — lock a price today for later

A future or forward is an agreement to buy or sell something later at a price fixed today. The payoff is linear: gain or loss moves one-for-one with the underlying. A future is the listed, standardised, centrally cleared version; a forward is the OTC, bespoke, bilateral version. Both do the same economic job — cheap, direct exposure or hedge to a rate, index or commodity — and both are conceptually simple, which is exactly why they are the most liquid and heavily traded instruments in the market.

Options — the right, not the obligation

An option is the right, but not the obligation, to buy (a call) or sell (a put) the underlying at a fixed strike price. The payoff is non-linear and asymmetric: the buyer's loss is capped at the premium paid, while the upside can be large. Because the value bends rather than moving in a straight line, an option is deeply sensitive to volatility — how much the underlying is expected to move. Options trade both listed and OTC. Clients use them for insurance-like protection; desks use them to trade volatility itself, which becomes a central theme in Part 2.

Swaps — exchange one cash-flow stream for another

A swap is an agreement to exchange one stream of cash flows for another over time. The classic is the interest-rate swap: one party pays a fixed rate and receives a floating rate, the other does the reverse. The payoff is a series of exchanges whose net value moves with the rate or the yield curve, so a swap is essentially linear in the rate. Swaps are overwhelmingly an OTC product, though vanilla ones are now largely pushed into central clearing. Clients use them to transform a rate exposure, hedge a loan, or take a view on the shape of the curve.

Product family Payoff shape Listed or OTC Volatility-sensitive? Typical use
Futures / forwards Linear Future = listed; forward = OTC No Cheap directional exposure or hedge
Options (calls / puts) Non-linear, asymmetric Both Yes — strongly Insurance-like protection; trading volatility
Swaps Linear in the rate Mostly OTC, vanilla now cleared No (vanilla) Transform rate exposure; curve views

Building an option from zero — the one product you must feel in your bones

Because options are the reference case and the source of every non-linear risk in the book, they are worth building from nothing — and the way to do that is to draw the payoff. Forget formulas for a moment. A call option is the right to buy at a fixed strike; a put is the right to sell at a strike. At expiry the value is mechanical: a call is worth whatever the underlying is above the strike, and nothing if the underlying is below it; a put is the mirror. The buyer paid a premium up front for this right, so the buyer's profit is the payoff minus the premium. Draw that and you get the shape that explains everything about options — the hockey stick.

Figure 1 Option payoffs at expiry. The buyer's loss is capped at the premium while the upside runs; the seller's gain is capped at the premium while the loss runs. Direction is only half the story.

The picture makes the asymmetry obvious, and the asymmetry is the whole point. A long call loses at most the premium no matter how far the underlying falls, but gains without limit as it rises. A long put is the reverse. The seller of each has the opposite shape: a capped gain (the premium) and an uncapped loss — which is why selling options is a business of collecting small premiums and fearing large moves. A straddle — long a call and a put at the same strike — profits from a big move in either direction and loses if the underlying sits still. That last shape is the cleanest possible statement of the idea that an option is a bet on movement, not on direction.

Moneyness, intrinsic value and time value

Three plain terms describe where an option stands. An option is in the money if exercising it now would pay (a call with the underlying above the strike), at the money if the underlying is at the strike, and out of the money if exercising would pay nothing. Its price splits into two parts: intrinsic value, the payoff if it expired right now, and time value, the extra the buyer pays for the chance that the underlying moves further in its favour before expiry. An out-of-the-money option has zero intrinsic value and is all time value — you are paying purely for the possibility of a move. Time value decays toward zero as expiry approaches, which is exactly the theta of Part 2 made concrete.

What volatility actually is, and why it raises the price

Volatility is simply how much the underlying moves, measured as the annualised standard deviation of its returns — a stock with 20% volatility is expected, loosely, to move about 20% over a year. Here is the Feynman heart of options: because the option's downside is capped at the premium but its upside is not, more movement is worth more. Wider dispersion increases the chance of a big favourable move without increasing the loss on the unfavourable side, which is already capped. So a more volatile underlying makes every option more valuable — and that single asymmetry is why an options book is really a volatility position, why vega exists, and why volatility becomes a product a desk trades in its own right.

Implied volatility, the smile, and put-call parity

Turn the pricing relationship around and you get the number the market actually watches. Implied volatility is the volatility that makes a pricing model reproduce the option's observed market price — it is the market's forecast of movement, read out of the price. Plot implied volatility against strike and it is not flat: out-of-the-money puts usually cost more, giving the downward skew (crash insurance is dear), and the whole shape is the smile. Trading that shape is a discipline in itself.

Figure 2 Implied volatility is not one number: it varies by strike, forming the smile and skew. The steepness of the skew reflects how much the market fears a crash — and it is itself tradeable.

One elegant relationship ties calls and puts together and is worth carrying: put-call parity. A call and a put at the same strike and expiry are not independent — holding a call and selling a put replicates a forward on the underlying, so their prices are locked into a fixed relationship. The practical consequence is that a desk can manufacture one from the others: a call can be built from a put plus the underlying, which is why an arbitrageur can police option prices and why the two payoffs above are really two views of one object. You do not need the algebra to use the idea; you need to know that calls, puts and the underlying are three faces of the same triangle.

The market is not one market — five asset classes

"Derivatives" is a category, not a market; in practice a desk trades one of five asset classes, each with its own products, venues and rhythm. The mechanics this module builds — price and risk, the two worlds, the Greeks, clearing and margin — are common to all five, which is why they are worth learning once. But the products, the dominant venues and the typical clients differ enough that desks specialise. Knowing the five, and where each one's liquidity and risk concentrate, is part of reading any markets business.

Asset class Core products Typical venues / CCPs What drives it
Rates Interest-rate swaps, bond & STIR futures, swaptions Eurex, CME; LCH SwapClear Central-bank policy, the yield curve
FX Forwards, FX swaps, options, cross-currency swaps Largely OTC; CME for futures Rate differentials, capital flows
Equity Index & single-name futures and options, total-return swaps Eurex, Euronext, CME Equity levels, dividends, volatility
Credit Credit default swaps (single-name & index) ICE, LCH; index CDS cleared Default risk, credit spreads
Commodity Energy, metals, agricultural futures and options ICE, CME, LME Supply, demand, storage, weather

Two lessons follow from the table. First, the rates and FX classes are the giants by notional, dominated by swaps and forwards and heavily shaped by central-bank policy; equity and commodity are more volatility- and event-driven; credit trades the probability of default itself. Second, the venue and clearing map is asset-class specific — a rates desk lives at LCH and Eurex, a commodity desk at ICE, CME and the LME — which is why the clearing-connectivity and basis problems of later parts look different in each class even though the principles are identical. Meridian, our running desk, spans rates and equity; a full CIB runs all five.

A worked product — how a credit default swap actually works

Credit is one of the five asset classes, and its signature product deserves building because it looks strange until you see it as insurance. A credit default swap (CDS) is a contract on the default of a reference entity — a company or a government and its debt. One side, the protection buyer, pays a regular premium (the spread); the other, the protection seller, receives that premium and, in exchange, promises to pay out if the reference entity suffers a credit event — typically bankruptcy, a failure to pay, or a restructuring. It is insurance on someone else's default: you pay a spread, and you get paid if the name goes bad.

Figure 3 A credit default swap is insurance on default: the buyer pays a spread each period (the premium leg); the seller pays out on a credit event (the protection leg). The reference entity is the name being insured.

Two legs and one trigger describe the whole product. The premium leg is the stream of spread payments the buyer makes — quoted in basis points per year on the notional, so a 200bp spread on €10m is €200,000 a year. The protection leg is the single contingent payment the seller makes if a credit event occurs: the buyer is made whole for the loss, receiving roughly par minus the recovery value of the defaulted debt, settled through a standardised auction that fixes the recovery price across the market. If no credit event ever happens, the seller simply keeps the premiums; if one does, the seller pays out and the contract terminates.

A CDS lets a desk trade credit risk directly, in both worlds, without ever touching the underlying bond. A bank can buy protection to hedge a loan it has made, an investor can sell protection to earn spread as a synthetic long position in the credit, and a hedge fund can express a view that a name will deteriorate by buying protection cheaply. Index CDS — baskets of many names, such as the iTraxx family in Europe — trade like any liquid derivative and are subject to the clearing obligation, while single-name CDS on a specific issuer are more bespoke. The spread itself is a live, tradeable price of default risk, which is why the CDS market is watched as a barometer of credit stress far beyond the desks that trade it.

The reference case: a volatility market-making book

Throughout the module, the hardest case — an options and volatility market-making book — is the reference against which simpler trades are measured. A futures hedge is a single linear trade; a vanilla swap is a stream of linear exchanges; but a book of listed and OTC options, quoted two-way across dozens of strikes and maturities and hedged continuously, is where every mechanic in this module comes together. If you can follow how a desk makes markets in EURO STOXX 50 index options — pricing the whole volatility surface live, auto-hedging the direction, managing the curvature, and clearing and margining it all — you can follow anything simpler. That book is the module's IPO-equivalent: the complex case that teaches the general lesson.

The market is enormous, systemic, and reshaped by regulation

Derivatives are one of the largest markets on earth, and that scale is why trading them well and safely is core to a bank. Global notional outstanding runs into the hundreds of trillions — a figure that overstates true economic risk (notional is the reference size, not the money at stake) but nonetheless signals how central these contracts are to how the financial system transfers risk. A CIB that cannot make competitive markets in rates, FX, equity and credit derivatives cannot serve its corporate and institutional clients, and forfeits one of the most reliable fee streams in the bank.

The business is intensely cyclical, and the cycle is driven by volatility. Derivatives desks make more money when markets move and clients need to hedge or reposition, and less when markets are calm and flow thins. Volatility itself trades in regimes — long quiet stretches punctuated by sharp spikes — and each spike is simultaneously the desk's best opportunity (spreads widen, flow surges) and its greatest danger (hedges gap, margin calls balloon). The recent cycle made this vivid: the pandemic shock of March 2020 sent volatility and margin calls soaring; the 2021 recovery was calmer; the 2022 rate-hiking cycle and commodity dislocations brought fresh spikes and margin spirals. Understanding the cycle is not optional colour — it decides when a book is most likely to break.

The cycle has a shape, and the shape is volatility

The derivatives cycle is not driven by economic growth in any simple way; it is driven by volatility, which trades in regimes. Long stretches of calm — low volatility, thin spreads, quiet flow — are punctuated by sharp spikes when a shock hits. Each spike is simultaneously the desk's richest opportunity and its greatest danger, which is why the same event that makes a quarter can also break a book. The recent history is a clean illustration: the pandemic shock of March 2020 sent volatility and margin calls soaring; 2021 was comparatively calm; the 2022 rate-hiking cycle and commodity dislocations brought fresh spikes, including the margin spirals that Part 5 dissects.

Figure 4 Volatility trades in regimes — long calm stretches broken by sharp spikes. Each spike is both the desk's best opportunity and its greatest danger.

This regime structure is why technology investment in derivatives is itself cyclical and easy to get wrong. In calm regimes, the risk and margin systems look over-engineered and expensive; in a spike, the same systems are all that stand between the desk and a spiral. Banks that cut resilience and risk-compute spending in the quiet years discover in the loud ones that the missing capacity is exactly what they needed. The lesson threads through Part 4 and Part 5: the systems that matter most are the ones that are only tested when everything else is going wrong. Post-2008 regulation permanently rebuilt how derivatives trade, and that rebuild is itself a defining feature of the business. Before the crisis, most OTC derivatives were bilateral, lightly collateralised, and opaque. The failure of that model in 2008 produced a wave of reform — mandatory central clearing for standardised trades, margin rules for those that stay bilateral, a push onto electronic venues, and vast reporting obligations. The result is a market where even "OTC" is heavily cleared and collateralised, where transparency and reporting are enormous workloads, and where the plumbing — clearing, margin, reporting — consumes as much technology effort as the trading itself. Part 4 and Part 5 return to this repeatedly; for now, hold the shape: regulation is not a footnote to derivatives trading, it is one of its load-bearing walls.

The benchmark that vanished — from IBOR to risk-free rates

No recent change reshaped the rates-derivatives market more than the death of LIBOR, and a reader of this module must know why it happened and what replaced it. For decades, floating rates on swaps and loans referenced IBOR benchmarks — LIBOR, EURIBOR and their peers — rates meant to reflect what banks paid to borrow from each other, set each day from bank submissions. The financial crisis exposed two fatal flaws: the underlying interbank lending that IBOR was supposed to measure had largely dried up, so the rate was increasingly a guess, and the submission process was manipulable, as a series of rate-rigging scandals proved. Regulators concluded the benchmarks were unsound and mandated a transition.

The replacement is a family of *risk-free rates* built from actual overnight transactions rather than estimates. In the euro area this is €STR (the euro short-term rate); in the US, SOFR; in the UK, SONIA. These are overnight rates compounded over a period, anchored in real transactions, and therefore far harder to manipulate — but structurally different from IBOR, because they carry almost no bank-credit component and are known only in arrears rather than set in advance. The transition, largely completed between 2021 and 2023, forced the entire market to re-paper contracts, rebuild pricing and discounting curves, and adjust systems to compound overnight rates — a vast, coordinated change.

Why a benchmark change is a systems project

Switching benchmarks is not a legal footnote; it is a deep systems and data change. Every pricing and risk engine had to build new €STR and SOFR curves, every booking system had to represent compounded-in-arrears floating legs, every valuation had to move to the new discounting basis (the "big bang" switch of CCP discounting from EONIA to €STR), and every downstream margin and reporting flow had to follow. Reference-data systems had to map old contracts to new ones and handle fallback language for trades that referenced a benchmark that no longer existed. It was one of the largest coordinated data-and-systems migrations the derivatives market has ever run — a concrete lesson that a change in a single number can cascade through the entire stack of Part 4.

The same products trade in two worlds

The organising fact of the entire module is that these three products trade in two marketplaces — listed and OTC — and the only real difference is *how* they are executed, cleared and margined. The economics of a Bund future and a bespoke bond forward are cousins; the plumbing around them is completely different. Learn the two worlds as a single trade-off and you have the spine of derivatives trading.

Figure 5 The two worlds are the same trade-off on every dimension: standardised safety versus bespoke flexibility.

The listed world: standardised, cleared, margined

Listed derivatives buy safety, transparency and liquidity by giving up flexibility. A listed contract trades on an exchange — Eurex, CME, ICE, Euronext — where anyone with access sees the same order book and the same prices. The contracts are standardised: fixed size, fixed expiry dates, fixed tick size and terms, which makes them fungible and therefore liquid. Every matched trade is cleared by a central counterparty (CCP), which steps between buyer and seller so that each faces the CCP rather than the other trader, and is margined daily — losses are collected as they occur, so credit risk is tiny. The price of all this safety is that you cannot customise anything: you take the contract as it is. In return you can trade huge size in and out in seconds, with almost no counterparty credit risk.

The OTC world: bespoke, bilateral, collateralised

OTC derivatives buy precise, tailored risk transfer by taking on complexity and some credit risk. An OTC trade is negotiated directly between two parties, usually through a dealer, with no exchange order book — it is a private contract between them. Any size, date, underlying or payoff can be tailored to the client's exact need, which is the source of OTC's value and its complexity. The relationship is governed by an ISDA Master Agreement plus a Credit Support Annex (CSA), the legal framework for netting, default and collateral between the two parties. Standardised OTC trades, such as vanilla swaps, are now pushed into CCP clearing; genuinely bespoke ones stay bilateral, exchanging margin under the Uncleared Margin Rules. Prices are not on a public screen — they are dealer-quoted and relationship-driven — though the trades are reported to repositories.

One trade-off, on every dimension

Read listed versus OTC as the same choice repeated across every attribute of a trade. Standardisation buys liquidity and safety; customisation buys precision and complexity. The table below is the single most important reference in Part 1: every row is the same trade-off wearing different clothes.

Dimension Listed (exchange) OTC (bilateral)
Standardisation Fixed, standard contracts — fungible Fully bespoke — any terms
Liquidity Deep and continuous in liquid contracts Varies; can be thin for bespoke
Credit risk Minimal — CCP plus daily margin Counterparty risk unless cleared
Flexibility Low — take the contract as-is High — tailor size, date, payoff
Transparency High — public order book and prices Lower — dealer-quoted, reported
Cost / margin Exchange and clearing fees; IM + VM Spread plus collateral (CSA / UMR); XVA
Best for Fast, liquid hedging and views Precise, tailored risk transfer

Four motives share one market, and the bank sits in the middle

Every participant in the derivatives market is playing one of four roles, and the dealer bank faces all of them. Hedgers — corporates, asset managers, insurers, pension funds — reduce an unwanted risk to gain certainty; they are the economic backbone of the market. Speculators — hedge funds, proprietary traders, directional investors — take a view for profit and provide much of the volume and liquidity. Arbitrageurs trade price discrepancies between related markets — cash versus futures, listed versus OTC, one maturity versus another — and their activity keeps prices consistent across the system. Market makers, the dealers this module is about, quote continuous two-way prices, warehouse the resulting risk, and earn the spread. Part 3 dissects these participants in detail; for now, the key point is structural.

The dealer is the counterparty to all three of the others, and nets their risk internally. A hedger wants to shed a risk; a speculator wants to take one; an arbitrageur wants to trade a small gap. The dealer stands ready to trade with each, warehouses the net exposure that remains after their offsetting interests cancel, and hedges only that residual. Crucially, the same participants populate both worlds: a pension fund might hedge with listed Bund futures and an OTC swap at the same time, and a hedge fund might arbitrage the basis between them. The bank makes markets across both worlds and nets the risk in one book — which is exactly why the two worlds cannot be understood separately.

  1. Why is a derivative's value described as "derived"?

  2. State the difference between trading and structuring in one sentence each.

  3. Which product family is non-linear, and why does that matter?

  4. Give the one trade-off that distinguishes listed from OTC. — Listed buys safety, liquidity and transparency by surrendering flexibility; OTC buys precise, tailored risk transfer by taking on complexity and some credit risk.

  5. What does the dealer bank actually own in a market of hedgers, speculators and arbitrageurs?

  6. Why is "hundreds of trillions of notional" a misleading measure of risk?

1

  1. Why is a derivative's value called “derived”?

  2. Which product family is non-linear?

  3. What does listed trading surrender in exchange for safety and liquidity?

  4. What separates trading from structuring?

2The Trade Lifecycle and Running the Book

This is the longest and most important part of the module, because it is where trading actually happens. We follow a derivative from the moment a price is struck through capture, clearing, margin and daily management to settlement — and, more importantly, we build the machinery a desk uses to run a live book: the Greeks, hedging, volatility trading, collateral and the arbitrage that keeps the two worlds aligned. The single most useful idea in Part 2 is that the risk does not live on the trade ticket; it lives on the book, and managing that book is the job.

Every trade follows one lifecycle, in both worlds

Strip away the execution step and the lifecycle of a listed trade and an OTC trade is nearly identical, because post-2008 regulation pushed both toward clearing and daily margin. A price is struck, the trade is captured and confirmed, it is cleared or kept bilateral, margin flows, the position is managed for its life, and finally it settles or expires. The big divergence is at the very start — an anonymous exchange order book versus a bilateral negotiation — and at clearing, where a listed or cleared-OTC trade novates to a central counterparty while a bespoke OTC trade stays bilateral. Everything in between rhymes.

Stage What happens Typical timing Who leads
1. Execute A price is struck — on the exchange order book (listed) or bilaterally by RFQ (OTC) Seconds Trader / auto-quoter
2. Capture & confirm The trade is booked once into the risk system; terms are confirmed Minutes Trader + middle office
3. Clear / give-up Cleared trades novate to the CCP, often via a clearing broker; bespoke OTC stays bilateral Trade date to T+1 Clearing operations
4. Margin Initial margin posted up front; variation margin exchanged daily against mark-to-market Daily, for life of trade Collateral operations
5. Manage & hedge Net the Greeks across the book, hedge the residual, run within limits Continuous Trader / risk
6. Settle / expire Cash- or physically-settle, unwind or roll; collateral released At maturity Operations

The intensity of a derivatives trade is inverted from an issuance deal: the pressure is not at execution but afterward, while the position sits on the book. Executing a Bund future takes a second. What takes the desk's attention is the days, weeks or years that follow, during which the position must be marked, hedged, margined and defended against every move in the market. The effort-and-risk curve peaks not at the trade but at the warehousing stage — which is exactly why the whole discipline of Part 2 is about managing a live book, not about pressing the button.

Figure 6 The trade lifecycle with an effort-and-risk curve. Unlike an issuance deal, the pressure peaks after execution, while the position is warehoused, margined and hedged.

Meet Meridian — the running example

Throughout Part 2 we follow Meridian, a fictional European bank's rates-and-equity-derivatives desk, and two of its trades. The first is an OTC trade: Garonne Industries, a French industrial with a €200m floating-rate loan, fears rising rates and asks Meridian for a five-year interest-rate swap to fix its cost. The second is a listed operation: Meridian is a registered market maker in EURO STOXX 50 index options on Eurex, streaming two-way prices across dozens of strikes and expiries. One bespoke swap and one screen market-making book — together they exercise every mechanic in this part. Every number attached to Meridian, Garonne or these trades is illustrative, chosen to make the mechanics legible, not to represent any real market level.

Stage 1 — Execution: two ways a price is struck

In the listed world, execution happens on a central limit order book, and it is anonymous, transparent and mostly automated. An exchange runs a central limit order book (CLOB) that matches bids and offers by price first, then time priority. A trader can post a limit order and wait to be filled, or cross the spread with a marketable order to trade immediately. Everyone sees the same book — pre-trade and post-trade transparency is the point — and matching is anonymous, so you face the CCP, not the trader on the other side. Most flow is electronic: algorithms and low-latency systems quote and hit continuously, and large blocks may be executed off-screen through block-trade facilities and then reported. Access is through an exchange member.

The order types that actually populate a book

A trader does not simply "buy" or "sell"; the order type encodes intent. A market order takes immediacy, crossing the spread to trade now at the best available price. A limit order posts a price and waits, providing liquidity and earning the spread if filled but risking not trading at all. A stop order triggers only once the market reaches a level, used to cap a loss or chase a breakout. An iceberg shows only part of a large order to avoid signalling size. On a market-maker's screen, the resting bids and offers are limit orders; the client who "lifts" or "hits" is sending a marketable order against them. Knowing which order type is in play tells you who is paying the spread and who is earning it — the single most important economic fact in any listed trade.

In the OTC world there is no public screen, so a client requests a quote. The request-for-quote (RFQ) workflow is the OTC equivalent of the order book. A client asks one or more dealers for a price on a specific trade — a given size, tenor and set of terms. Each dealer prices it from its models and adds a spread for risk, capital and the relationship. The client deals with one dealer or passes, often putting several dealers in competition. Because the dealer knows the client, the quote reflects the relationship, the two-way flow the client brings, and the balance-sheet cost of the trade — not just a screen level. Once the client deals, the dealer warehouses the risk and must hedge it, which is where the real work begins.

Execution

Listed execution runs over the FIX protocol into exchange gateways, with market-data feeds returning prices and an auto-quoting engine streaming bids and offers; latency and co-location decide who wins the flow. OTC RFQ runs over multi-dealer platforms — Tradeweb, Bloomberg — or, for standardised swaps, regulated venues (SEFs in the US, OTFs and MTFs in the EU). In both cases the executed trade must land in the booking system accurately and instantly: the golden trade record created here is the atom that every downstream system — risk, margin, reporting — depends on. An error at the point of capture propagates everywhere.

Stage 2 — Capture and confirm: book it once, book it right

The trade must be recorded exactly once, accurately, and flow straight through to every downstream system. The moment of capture creates the golden record: the single authoritative version of the trade's economics — product, notional, rate or strike, maturity, counterparty, clearing route. From that record flow the risk engine's Greeks, the margin calculation, the profit-and-loss, and the regulatory report. Re-keying a trade into a second system, or capturing it with a wrong field, is one of the most common and most expensive operational failures in the business, because the error is invisible until a downstream number is wrong.

Confirmation is the formal agreement that both sides booked the same trade, and it differs by world. A listed trade is confirmed by the exchange automatically — the matching engine is the record. An OTC trade is confirmed against the ISDA Master Agreement: the two parties (increasingly via electronic affirmation platforms) agree the precise economics before the trade is locked in. Catching a discrepancy here — a mismatched rate, a wrong date — is cheap; catching it after margin has been exchanged and the position hedged is not.

The internal gauntlet before a trade goes on

Before a bank commits its balance sheet to a client trade, the trade passes through a set of internal gates — the trading equivalent of a deal-governance process. These controls are largely invisible to the client and mostly automated, but they are the reason a desk cannot simply trade anything with anyone. Four gates matter most.

Counterparty onboarding and KYC

Before Meridian can trade with Garonne at all, Garonne must be onboarded: know-your-customer and anti-money-laundering checks, legal capacity to trade derivatives, an ISDA Master Agreement and CSA negotiated and signed, and a Legal Entity Identifier on file. For a new client this can take weeks and is a genuine gate — no ISDA, no OTC trade. This is why relationship and onboarding work precedes flow.

Credit and limit checks

Every trade consumes credit and risk limits, and the system checks them pre-trade. Is there enough counterparty credit line for an uncleared trade with Garonne? Does the trade fit within the desk's risk limits — net delta, vega, notional, value-at-risk? A trade that would breach a limit is blocked or must be escalated. These checks run in milliseconds for listed flow and pre-quote for OTC, but they are hard constraints, not suggestions.

New-product and mandate approval

A genuinely novel product — a new payoff, a new underlying, an exotic structure — cannot simply be traded. It passes a new-product approval process that confirms the bank can price it, risk-manage it, book it in its systems, and report it correctly. This is the gate that stops a desk from warehousing risk it cannot actually model or hedge.

Best execution and conduct

Under MiFID II a desk owes clients best execution and must be able to evidence it, and all trading is subject to market-abuse controls. The obligation to quote fairly, avoid front-running client flow, and record communications is a live, systematised constraint on how the desk behaves — not a compliance afterthought.

Stage 3 — Clearing and give-up: who stands behind the trade

For cleared trades, a central counterparty novates the contract and becomes the counterparty to both sides — the single most important piece of post-crisis market structure. Novation means the CCP legally replaces the original bilateral contract with two new ones: the CCP is buyer to every seller and seller to every buyer. After novation, Meridian faces the CCP, and the client faces the CCP; neither faces the other. Bilateral counterparty credit risk — the fear that the party across the trade defaults — is removed and replaced by exposure to the CCP, which is heavily margined and regulated. This is why a cleared trade carries far less credit risk and far less capital than a bilateral one.

Figure 7 Novation: before clearing, two parties face each other and each other's credit risk; after clearing, each faces the CCP. Behind the CCP stands the default waterfall.

Most clients reach the CCP through a clearing broker, and the mechanism is the "give-up". A client that is not itself a clearing member of the CCP uses a clearing broker — in the US, a Futures Commission Merchant (FCM) — who carries its cleared positions at the CCP and manages its margin. The executing trade is given up to that clearing member, who steps in to clear it. Give-up to a clearing broker is standard for both listed and cleared-OTC trades, and it is why the chain from execution to clearing touches several institutions, not just two.

Clearing connectivity

Clearing demands reliable, compliant connectivity to multiple CCPs (LCH, Eurex Clearing, ICE Clear, CME Clearing) and clearing brokers, each with its own interfaces, message formats and rulebooks. The desk's systems must route trades to the right clearing destination, receive novation confirmations, and reconcile positions with the clearing member daily. Maintaining these links is a large, never-finished engineering job — a recurring pain point we return to in Part 4 — because every new venue, CCP or product adds another interface to build and keep alive.

Stage 4 — Margin: how the market removes credit risk

Margin is the mechanism that makes clearing safe, and it comes in two forms with two jobs. Initial margin (IM) is a buffer posted up front to cover the potential future loss if a member defaults and its position must be closed out over a few days — set by the CCP's risk model as a fraction of notional, which is where the leverage in derivatives comes from. Variation margin (VM) is the daily, sometimes intraday, mark-to-market settlement: the position is revalued, and the loser pays the winner in cash, so exposure never accumulates. IM is a standing cushion; VM keeps the running score settled. Together they mean that even if a counterparty vanishes, the losses are already collateralised.

Behind the CCP stands the default waterfall — the layered defences that absorb a member failure without cascading. If a clearing member defaults, the CCP first uses that member's own margin, then its contribution to the mutualised guaranty fund, then a slice of the CCP's own capital (its "skin in the game"), and only then the mutualised default fund contributed by surviving members. This ordered sequence is what lets a CCP guarantee every trade: the defaulter pays first, and mutual loss-sharing is the last resort. It is also why CCPs are themselves systemically critical and heavily regulated — a point Part 3 returns to.

Uncleared trades are not exempt from margin; they fall under the Uncleared Margin Rules. A bespoke OTC trade that cannot be cleared still exchanges IM and VM under UMR, with initial margin posted to a segregated account and computed, for most dealers, using the industry-standard ISDA SIMM model. The post-2008 principle is uniform: credit exposure must be collateralised, whether the route is a CCP or a bilateral CSA. The difference is operational load and cost, not whether margin applies.

The yield curve, discounting and DV01 — how a rate becomes a price and a risk

Before an interest-rate swap can be priced, one idea must be solid: money in the future is worth less than money now, and the yield curve is the price of that difference. The yield curve plots the interest rate for each maturity — the rate for one year, three years, five years, ten years. It is the market's price of time, and it is the single most important input to every rates trade. Reading it is not optional background; it is the raw material a rates desk works with all day.

Figure 8 The yield curve is the price of time. Each future cash flow is discounted at the rate for its maturity; the sum of the discounted flows is present value; and DV01 is how much that value moves per one-basis-point shift.

Discounting turns a stream of future cash flows into a single value today, and that value is what "pricing" means. Each future payment on a swap is worth its face amount reduced by the rate for its maturity — a €1 payment in five years, at a 3% five-year rate, is worth less than €1 today. Summing every discounted cash flow gives the present value, the fair value the pricing engine computes. A swap is priced at inception so that the present value of the fixed leg equals the present value of the floating leg — the par swap rate is simply the fixed rate that makes those two sides balance, which is where Garonne's 3.00% came from.

The rate risk of the whole position collapses into one number the desk lives by: DV01. DV01 (dollar value of a basis point), also called PV01, is how much the present value of a position changes when the whole yield curve shifts by one basis point. It is the rates desk's delta — a single, additive measure of interest-rate exposure that can be netted across every swap, bond and future in the book. When Meridian delta-hedges Garonne's swap, it is really matching the swap's DV01 with an equal and opposite DV01 in listed futures, so the net DV01 of the combined position is near zero. Every rates hedge, every limit, every risk report is ultimately a statement about DV01.

Pricing a trade: fair value plus the cost of carrying it

An OTC price is not a single model number; it is a model fair value plus a spread plus a stack of valuation adjustments that turn theory into a real, chargeable price. The starting point is the model fair value — the risk-neutral present value the desk's models compute, the price at which the trade is "fair" before any costs. To that the desk adds a bid-offer spread for the risk it will warehouse and the relationship, and then the XVAs: a family of adjustments for the real costs the trade imposes over its life. Only after these does "fair value" become the price the client actually pays.

Figure 9 Pricing an OTC trade: model fair value, plus spread, plus the XVAs (credit, funding, margin, capital), equals the price the client actually pays.

The XVAs are the honest accounting of what a trade truly costs the bank. CVA (credit valuation adjustment) prices the risk that an uncleared counterparty defaults; DVA is its mirror, the value of the bank's own default risk. FVA (funding valuation adjustment) is the cost of funding the uncollateralised part of a position — the money the desk must borrow to carry it. MVA (margin valuation adjustment) is the lifetime funding cost of posting initial margin. KVA (capital valuation adjustment) is the cost of the regulatory capital the trade consumes over its life. Each is a genuine cost; ignoring them makes a trade look profitable when it is not.

XVA What it prices Why it matters
CVA / DVA Counterparty default risk (and the bank's own) The credit cost of an uncleared trade
FVA Funding the uncollateralised part of a position Money the desk must borrow to carry the trade
MVA Lifetime funding cost of posting initial margin Margin is expensive; someone must pay for it
KVA Regulatory capital the trade consumes Capital is scarce and must earn a return

Stage 5 — Managing the book: the heart of the job

Once a trade is on, the desk stops thinking about the trade and starts thinking about the book — the netted sum of every position it holds. A market maker does not manage tickets one by one; it aggregates the risk of every listed and OTC position into a single net risk profile and manages that. The tools for describing how the book's value will move are the Greeks, and the discipline of keeping the book safe is hedging. This is where a derivatives desk earns or loses its living.

The Greeks — the dials that tell a trader how the book will move

The Greeks are the sensitivities of the book's value to the things that can change, and each one is a dial the desk watches. They are named for Greek letters and they are universal maths; the derivation belongs to Module 4, but the operational meaning is what matters here.

Figure 10 The Greeks are the dials of a derivatives book. The desk nets them across every position and hedges the net, not each ticket.
Delta — sensitivity to the underlying's price

Delta measures how much the book's value changes when the underlying moves. It is the first thing a desk neutralises: delta-hedging offsets directional risk so the book's profit does not depend on which way the market goes. A delta-neutral book is one that has stopped betting on direction.

Gamma — how fast delta itself changes

Gamma measures how quickly delta moves as the underlying moves. High gamma means the hedge drifts fast and must be rebalanced often. A long-gamma book likes big moves — it makes money re-hedging into them — while a short-gamma book fears them, because it must chase the market, buying high and selling low to stay hedged. Gamma is why an options book is never "set and forget".

Vega — sensitivity to volatility

Vega measures how much the book's value changes when volatility changes. An options book is really a volatility position, and vega is the core exposure a volatility trader manages. When implied volatility rises, a long-vega book gains; when it falls, it loses — regardless of direction.

Theta — the passage of time

Theta measures how much value the book gains or loses each day simply from time passing. It is the rent paid or earned for holding options: a buyer of options pays theta (options decay toward expiry), a seller earns it. Theta and gamma are two sides of one coin — you pay theta for the privilege of long gamma, or earn theta for the danger of short gamma.

Rho and the higher-order Greeks

Rho measures sensitivity to interest rates; for exotic books, higher-order and cross Greeks (vanna, volga) capture how the main Greeks interact. These are smaller but not ignorable, especially in large or structured books.

The whole point is that the desk manages the *net* Greeks across the entire book, listed and OTC together. A single risk engine aggregates every position into one net delta, one net gamma, one net vega, and the desk hedges those net numbers. This netting is the economic advantage of a large book: offsetting client trades cancel internally, so only the residual must be hedged externally.

Hedging the book — neutralise what you are not paid to hold

A market maker does not want to bet on direction; it wants to earn the spread and hold only the risk it is paid to warehouse. Hedging is the continuous act of stripping out the unwanted exposures and keeping only the deliberate ones. It proceeds in a rough order, and it never stops.

First, delta-hedge the direction: trade the underlying or a future to bring net delta to about zero, so the book is not a directional bet. This is the constant first job. Second, manage gamma: because delta moves, re-hedge as the market moves — a short-gamma book must chase the market, which is costly in big moves, so gamma exposure is watched closely. Third, manage vega: offset volatility exposure with other options, and decide how much volatility risk to keep as a deliberate view versus hedge away. Fourth, net the residual: aggregate offsetting risks across all client trades so only the net exposure is hedged externally. Fifth, hold the paid-for risk: after neutralising the unwanted parts, keep the residual the desk is compensated for, and monitor it against limits.

A worked delta-and-gamma hedge

Numbers make the mechanics concrete. Suppose Meridian is short 1,000 EURO STOXX 50 call options, each with a delta of 0.40, on an index at 4,000, with each index point worth €10 per contract. The book's net delta is short 1,000 x 0.40 x 10 = 4,000 index-points of exposure — it loses if the index rises. To delta-hedge, Meridian buys index futures equivalent to +4,000 points of delta, bringing net delta to about zero. Now the book is directionally neutral. But it is short gamma: if the index jumps to 4,100, the calls' delta rises to, say, 0.55, so the short-option delta becomes 5,500 points while the futures hedge is still +4,000 — the book is suddenly short 1,500 points into a rising market and must buy more futures at the higher level to re-hedge. That is the short-gamma tax: re-hedging forces buying high and, on the way back down, selling low. The desk collected option premium (theta) for taking exactly this risk, and the trade is profitable only if realised volatility stays below what that premium implied. (Illustrative figures.)

The inter-dealer market — where dealers lay off risk to each other

When a desk cannot net a residual risk against its own client flow, it turns to the inter-dealer market — the wholesale market where banks and non-bank makers trade with each other to offload or acquire risk. This trades on the listed order book for standard products and through inter-dealer brokers and platforms for OTC risk, at thin spreads because both sides are professionals. It matters for two reasons. First, it is the release valve: a desk that has warehoused a large client position it cannot net internally lays part of it off with other dealers, spreading the risk across the market rather than holding it all. Second, inter-dealer prices are the reference the whole market marks against — the anonymous, professional level that anchors what a "fair" price is. A desk that reads the inter-dealer market well knows where it can actually hedge, and at what cost, before it quotes the client.

Cross-world hedging — using listed markets to manage OTC risk

The two worlds are not separate; the desk routinely uses liquid listed instruments to hedge the risk of bespoke OTC positions. A bespoke OTC trade rarely has an exact market offset — no one else wants precisely Garonne's odd-dated swap — but its risk factors do have liquid hedges. So the desk decomposes the OTC position into its Greeks and hedges each factor with the cheapest liquid instrument, which is usually listed. Garonne's OTC swap is delta-hedged with listed Bund or short-term-interest-rate futures. An OTC equity option is delta-hedged with the listed index future and vega-hedged with listed options. An OTC FX option is hedged with FX forwards and futures.

This connection means the desk runs one integrated book, and it also means basis risk never fully disappears. Because listed hedges are standardised and the OTC position is bespoke, the hedge is never perfect — a residual basis between the two must be monitored and managed. Liquidity flows both ways: OTC flow generates listed hedging volume, and listed prices anchor OTC marks. The desk always picks the most liquid, lowest-cost instrument to offset a given risk factor — cheapest-to-hedge wins — which is usually the listed market for vanilla risks.

Volatility trading — when volatility itself is the product

Once a desk has delta-hedged, direction is gone, and what remains is a bet on volatility — for an options desk, volatility *is* the product. The trade is between implied volatility — what the option's price says the market expects — and realised volatility — how much the underlying actually moves. A long-volatility position (buying options) profits if the market moves more than implied, and pays theta for the privilege; it loves chaos. A short-volatility position (selling options) earns the premium if calm persists, collecting theta, but fears jumps. The volatility trader's edge is judging implied against realised and managing gamma and vega through it.

Volatility is not one number but a whole surface, and its shape is itself tradeable. Implied volatility varies by strike and by maturity, forming a volatility surface rather than a single figure. Out-of-the-money puts often cost more than symmetry would imply — the skew, a reflection of crash fear — and the steepness of that skew is a view a desk can trade. Short-dated and long-dated volatility can diverge, giving rise to term-structure trades across maturities. And dedicated volatility products — variance and volatility swaps, VIX-style indices, listed volatility futures — let desks trade volatility directly rather than synthesising it from options. For Meridian's EURO STOXX 50 options book, pricing the skew and term structure correctly and staying vega-disciplined is the real skill; the delta-hedging is almost mechanical by comparison.

Collateral and the cost of trading

Margin is not free, and funding it reshapes what a trade is really worth. Every cleared or margined trade demands collateral. VM is usually cash, paid daily — cheap, but it must be funded intraday. IM is high-quality assets locked up for the life of the trade — capital that cannot be used elsewhere. Eligibility and haircuts determine what can be posted and at what discount, and collateral optimisation — posting the cheapest eligible asset — is a business in itself. A desk that funds margin cheaply and manages collateral well has a real, durable edge over one that does not.

This is where MVA and the other XVAs feed back into pricing. The lifetime cost of funding initial margin (MVA), the cost of funding uncollateralised positions (FVA), the credit charges (CVA/DVA), and the capital cost (KVA) are all consequences of collateral and balance-sheet use, and all of them are priced into trades so the desk recovers them. Margin removed most counterparty credit risk from the system — a real gain — but it replaced it with a funding cost that never goes away. Trading did not get free; it got collateralised.

Basis and arbitrage — trading the gaps that keep the worlds aligned

Small, persistent price differences between related markets are both a profit source and the mechanism that keeps the two worlds tethered. The cash-futures basis is the gap between a future and its underlying cash instrument, adjusted for carry; trading it links the derivative to what it derives from. The listed-versus-OTC basis is the small difference in pricing the same risk in the two worlds — desks arbitrage it, and that activity feeds cross-world hedging. Calendar spreads trade one expiry against another; cross-CCP and cross-currency basis capture the difference between the same swap cleared at different CCPs or across currencies. In every case the desk buys the cheap and sells the rich, expecting convergence.

None of this is a free lunch. "Arbitrage" here is rarely riskless: it needs funding, capital and patience, and the gap can widen before it converges. But it matters beyond its own profit, because it is the force that keeps listed and OTC prices consistent. Arbitrageurs and dealers close the gaps, so the two worlds stay aligned rather than drifting apart — which is why a Bund future and a bespoke bond hedge track each other at all.

Stage 6 — Settlement, expiry and the post-trade life

A derivative does not simply end; it lives through a lifecycle of events until it settles or expires, and that lifecycle is real operational work. Over a swap's life, coupons and floating-rate resets must be calculated and exchanged on schedule; over an option's life, exercise decisions and expiry must be handled; across all products, corporate actions on the underlying (splits, dividends, index changes) must be processed. Positions are rolled — closed in a near expiry and reopened in a further one — to maintain a hedge beyond a contract's maturity. Each event is a chance for an operational error, and each must flow cleanly through the same systems that captured the trade.

At the end, the trade settles or expires, and the mechanism depends on the contract. Cash settlement pays the net difference in cash — the norm for index products and most swaps. Physical settlement delivers the underlying — a bond, a commodity, shares — against payment, which for securities runs on a delivery-versus-payment (DVP) basis so that delivery and payment happen simultaneously, on a T+1 or T+2 timeline, with settlement instructions carried by SWIFT or ISO 20022 messages. Once the trade settles or expires, the position is closed, the mark goes to zero, and the collateral posted against it is released. For Garonne, at year five the final exchange settles, the swap matures, and Meridian's residual hedge is unwound.

Settlement risk — the danger in the moment of exchange itself

Settlement is not free of risk just because the trade is over; the act of exchanging value has its own failure mode, and it has a name. Settlement risk is the danger that you deliver your side of a trade and your counterparty fails before delivering theirs — so you are out the full amount, not merely a mark-to-market difference. The classic case is foreign exchange, and the classic name is Herstatt risk, after the German bank whose 1974 failure crystallised the problem: because the two currency legs of an FX trade settle in different countries and time zones, a bank could pay away one currency in the morning and, hours later, find its counterparty had failed before paying the other. The loss is the entire principal, not a spread.

Figure 11 Settlement risk in FX: without payment-versus-payment, a time-zone gap can cost the full principal (Herstatt); with CLS, both legs settle together or neither does.

The fix is the same principle as delivery-versus-payment, applied to two cash legs: payment-versus-payment. For securities, delivery-versus-payment (DVP) already ensures the asset and the cash change hands simultaneously, so neither side is ever exposed to the other's default at settlement. For FX, the equivalent is payment-versus-payment (PvP), delivered by CLS (Continuous Linked Settlement) — a specialised settlement system through which the world's major currencies settle their two legs simultaneously, so that either both legs complete or neither does. CLS did for FX settlement what the CCP did for counterparty credit risk: it engineered away a whole category of principal loss that used to be simply accepted.

Settlement is a plumbing and messaging problem

Settlement risk is reduced by systems and standards, not by good intentions. DVP and PvP require settlement instructions to be matched, netted and released atomically — carried by SWIFT and ISO 20022 messages, coordinated through central securities depositories for securities and through CLS for FX. The desk's operations systems must generate accurate settlement instructions, match them against the counterparty's, manage the netting cycle, and handle a failed settlement when an instruction does not complete. A break here is not a P&L wobble; it can be a full-principal loss, which is exactly why settlement infrastructure is treated as critical and why its resilience falls squarely under DORA.

How a derivatives desk makes money — four honest sources

The desk's profit comes from four sources, and none of them is free. The bid-offer spread is the core: buy at the bid, sell at the offer, capture the gap across many trades — the reward for providing liquidity and immediacy. Market-making flow lets the desk earn that spread repeatedly and net offsetting risk internally, so it earns more the more flow it sees. Basis and relative value harvests the small, persistent price differences that arbitrage keeps tight. Volatility and risk premium comes from selling options to harvest the premium the market pays for protection — the desk earns if realised volatility stays below implied, with real risk attached.

Figure 12 The four honest P&L sources, net of hedging cost, margin funding and capital. Gross spread is not profit — what survives the costs is.

Against every source sits a cost, so gross spread is not profit. Hedging costs money, margin and collateral must be funded (MVA and the XVAs), capital must earn a return (KVA), and there is always the risk that a hedge fails or the market gaps. Net of all this, the spread can be thin. The honest framing is that a derivatives desk is paid to warehouse risk and provide liquidity: its edge is pricing accurately, hedging cheaply and managing inventory, not gambling. The best desks make steady money on flow, not big directional bets — and the ones that forget this are the case studies in Part 5.

  1. Why does the effort-and-risk curve of a derivatives trade peak after execution?

  2. Contrast how a price is struck in the listed versus the OTC world. — Listed: anonymously on a central limit order book by price-time priority. OTC: by request-for-quote, where the client asks dealers for a price and deals bilaterally, with pricing reflecting the relationship.

  3. What does novation by a CCP achieve, and what stands behind the CCP?

  4. Distinguish initial margin from variation margin.

  5. Turn "model fair value" into the price a client pays.

  6. A book is short gamma when the index gaps overnight. What happens, and why is it feared?

  7. Why does a desk use listed instruments to hedge OTC positions, and what risk remains?

  8. Name the four honest sources of a desk's P&L and the one-line reason none is free.

2

  1. Why does the effort-and-risk curve of a derivatives trade peak after execution?

  2. What does CCP novation achieve?

  3. Initial versus variation margin: which is a buffer against future loss over a close-out period?

  4. How is a price struck in the listed world?

3Actors and Roles: Who Makes the Market Work

A derivatives trade touches a chain of specialised institutions, and inside the dealer it passes through a team of specialists front to back. This part maps both: the external ecosystem — clients, exchanges, CCPs, clearing brokers, venues, regulators — and the internal roles that make a desk function. The organising idea is the same one from Part 1: the dealer sits in the middle of every interaction, which is why understanding who surrounds it, and who staffs it, is the key to reading the business.

Four camps, and the dealer in the middle of all of them

Every actor in the derivatives market belongs to one of four camps, and the dealer is the one that touches the other three. The first camp is clients and end-users — corporates, asset managers, hedge funds, pensions and insurers — who generate the flow as hedgers, speculators and arbitrageurs. The second is the dealer itself, the bank that quotes two-way prices and warehouses risk. The third is market infrastructure — exchanges, central counterparties, clearing brokers, trading venues and trade repositories — the plumbing that executes, clears, guarantees and records trades. The fourth is regulators, who set and supervise the rules that shape all of it. The dealer is a party to almost every interaction: it faces clients on one side, infrastructure on another, and regulators overhead.

Figure 13 The dealer sits at the hub. Clients bring flow, exchanges and CCPs execute and guarantee, clearing brokers give access, and regulators supervise — the desk is one side of nearly every arrow.

Reading the ecosystem as a hub-and-spoke makes the flows legible. Clients send RFQs and orders and receive quotes and fills. The dealer executes on exchanges and, for cleared trades, gives up to a clearing broker who carries the position at the CCP. The CCP novates and margins. Trade repositories and reporting infrastructure record everything. Regulators receive filings and supervise conduct. The table below names the major information and instruction flows — who talks to whom, and at which stage.

From To At which stage What flows
Client Dealer Execution RFQ, order, hedging need
Dealer Client Execution Two-way quote, fill, confirmation
Dealer Exchange Execution Orders and quotes (listed)
Dealer Clearing broker (FCM) Clearing Give-up of the executed trade
Clearing broker CCP Clearing Novation, margin, default-fund contribution
CCP Dealer / client Margin Daily IM and VM calls
Dealer Trade repository Reporting EMIR trade reports
Dealer Regulator Reporting / conduct MiFIR transaction reports, best-ex, surveillance

The client taxonomy — who brings the flow, and why the dealer reads them

The buy side is not one thing, and a dealer that treats it as one prices badly and hedges worse. Different clients trade for different reasons, hold for different horizons, and behave differently under stress — and those differences drive how the dealer prices, how much flow it expects to net internally, and how sticky the risk it warehouses will be. Four broad types matter most.

Corporates — the hedgers with a real-economy need

Corporates like Garonne use derivatives to remove an unwanted operating risk: a rate on a loan, an FX exposure on foreign revenue, a commodity input cost. They trade infrequently, in size tied to real business, and hold to maturity — they are hedging, not trading. For a dealer, corporate flow is prized because it is directional, sticky and relationship-driven, and it often pays a fuller spread in exchange for advice and certainty.

Asset managers, pensions and insurers — the patient institutions

Large asset managers, pension funds and insurers use derivatives to manage the risk of vast portfolios: hedging duration, equity exposure or inflation, or gaining efficient exposure without trading cash markets. Their horizons are long and their behaviour is relatively predictable, which makes their flow valuable and their positions stable. Pension and insurance hedging — especially liability-driven investment — can be enormous and, as the 2022 UK gilt episode showed, can itself become a source of systemic stress when it must be unwound quickly.

Hedge funds — the liquidity and the flip risk

Hedge funds and proprietary traders take views for profit, and they are a major source of volume and liquidity. Their horizons are short, their price sensitivity is high, and they will move for a basis point — which makes their flow less sticky and their risk quicker to reverse. A dealer values hedge-fund flow for the liquidity and two-way business it brings, but prices it tightly and watches the concentration and leverage it can carry, because a levered fund unwinding fast (Part 5's Archegos) is a real danger.

The gap in the middle — and why type drives hedging

Between these sit banks, corporates' treasuries and smaller institutions, but the pattern is clear enough from the extremes. The table below is the key reference: it connects client type to how the dealer thinks about the risk each brings.

Client type Holding horizon Price sensitivity Why the dealer wants or wary of them
Corporate (hedger) To maturity Lower — values certainty Sticky, directional, pays fuller spread; prized
Asset manager / pension / insurer Long Moderate Large, predictable, stable risk; systemically big
Hedge fund / prop Short High Liquidity and two-way flow, but fast to reverse; watch leverage
Other bank / dealer Varies High Inter-dealer hedging and netting; thin spreads

The dealer is not always a bank

Some of the most important market makers in listed derivatives are not banks at all, and ignoring them misreads the modern market. Specialist electronic trading firms — the likes of Citadel Securities, Optiver, IMC and Jane Street — have become dominant liquidity providers in listed options and futures, precisely because listed market making has become an engineering and speed contest that plays to their strengths. They carry no deposit franchise and little bespoke OTC business; they win on latency, quote quality and inventory-aware pricing, streaming tighter markets in liquid contracts than many banks can. For a bank desk, this reshapes the competitive landscape: in the most liquid listed products the bank competes with non-bank makers on technology, while its durable edge lies in the OTC world, the client relationships, and the balance sheet that lets it warehouse bespoke risk a pure trading firm cannot. Reading who makes the market — bank or non-bank — tells you where the competition is really being fought.

Information barriers — one institution that must not tell itself things

A dealer bank knows things in one part of itself that another part must not use, and the wall between them is both a legal duty and a technology problem. A bank simultaneously advises clients, executes their flow, conducts research, and trades its own positions. If the desk executing a client's large order could tip the bank's own traders, or if material non-public information from the advisory side leaked to the trading side, the bank would be trading on knowledge it is not entitled to use. Information barriers — the "Chinese walls" — are the controlled separation that prevents this.

Four functions must be kept apart, and the separation is enforced, not merely promised. Client execution must not leak to proprietary or house trading. Advisory and origination, which handle material non-public information, must be walled from trading. Research must be independent of both. And personal account dealing by employees must be surveilled so staff cannot trade ahead of clients or on inside knowledge. The formal, controlled exception is wall-crossing: bringing a named individual across the wall for a specific purpose, logged and time-bound, with trading restrictions applied. Everything about this is systematised, because policy alone cannot enforce it at the speed and scale of modern trading.

Enforcing the wall

The barrier is enforced by a stack of systems: trade and communications surveillance that monitors messages and orders for abuse or leakage; a restricted list that blocks trading in names where the bank holds inside information; access controls that segregate systems and data between walled functions; a wall-crossing log that records every controlled exception; and personal-account-dealing systems that pre-clear and monitor employee trades. Market-abuse monitoring under MAR ties it together. The wall is only as strong as these systems and the completeness of the data feeding them — which is why surveillance is a first-class engineering discipline, not a compliance box.

The desk is a team of specialists, front to back

A derivatives trading operation is not a room of traders; it is a coordinated team of specialists, each owning a piece of the price-and-risk chain. The client sees a trader, but behind that trader stands a structure of roles that price, hedge, cost, clear, control and report every position. Understanding who does what — and where the information barrier runs through the team — is essential to reading how the business actually works.

Role Where it sits Core job Primary technology
Listed trader / DMM Front office Makes markets in futures and listed options; manages screen quotes, inventory, exchange obligations Auto-quoter, exchange gateways, FIX
OTC market maker Front office Prices and warehouses bespoke and vanilla OTC risk; runs the RFQ book; hedges the Greeks Pricing engine, RFQ platforms
Volatility trader Front office Runs the options/vol book — implied vs realised, the surface, skew — long or short vol Vol-surface tools, options analytics
Quant / strat Front office (aligned) Builds pricing models, calibration and risk analytics; supports auto-pricing and hedging logic Quant libraries, model dev tools
XVA desk Front office (central) Prices and manages CVA/FVA/MVA/KVA centrally; owns funding, credit and capital cost of trades XVA engine, funding/capital models
Clearing & margin ops Middle / back office Handles give-ups, novation, margin calls, collateral moves and reconciliation Collateral system, CCP connectivity
Risk & product control Independent control Sets and monitors limits and VaR; independent P&L and valuation control; challenges the marks VaR/stress engine, IPV tools
Compliance & surveillance Independent control Ensures EMIR/MiFID II conduct and reporting; monitors for market abuse Surveillance, reporting platforms
Technology / data Enabling Builds and runs pricing, risk, connectivity, clearing and reporting systems The whole stack

Two features of this structure are worth naming explicitly. First, the information barrier runs through the team: the front-office trader who owns the position and the independent risk-and-control function that sets its limits and challenges its marks must be genuinely separate, because the trader owns the risk but risk control owns the leash. A control function that reports to the desk it controls is not a control. Second, the XVA desk is a modern, central role: rather than every trader guessing at funding and capital costs, a single desk owns the XVAs across the whole institution and charges them back to trades — the organisational answer to the pricing shift described in Part 2.

Each role maps to a stage of the lifecycle and a system, which is how the team and the machinery line up. The traders and quants own execution, pricing and hedging (the pricing and risk engine); the XVA desk owns the cost layer; clearing and margin operations own novation and collateral (the collateral system and CCP links); risk and product control own the limits and independent valuation (the VaR and stress engine); compliance owns reporting and surveillance; and technology owns the interfaces between all of them. Part 4 is, in effect, the systems view of this same team.

  1. Name the four camps in the derivatives market and say why the dealer is central.

  2. Why does a dealer care whether flow comes from a corporate or a hedge fund?

  3. What is wall-crossing, and why is the information barrier a technology problem?

  4. What does the XVA desk do, and why is it centralised?

  5. Why must risk and product control be independent of the trading desk?

  6. Give one example of how a specific client type changes how the dealer manages the risk it just took on.

3

  1. Why does a dealer care whether flow comes from a corporate or a hedge fund?

  2. What does the XVA desk do?

  3. Why is the dealer central in the derivatives market?

  4. What is wall-crossing?

4Systems, Data and the Technology Stack

Derivatives trading is one of the most system-heavy activities in the entire bank, and most of its pain is not in any single system but in the interfaces between them and in keeping listed and OTC in one consistent view. This part builds the technology picture the way it should be built: first the clean functional question — what does any trade actually need? — then the vendors who fill those needs, then the data that flows through, the identifiers and messages that carry it, the resilience that protects it, and the regulations that specify it. Read each system as an answer to a requirement, not as a name to memorise.

Functional architecture — what any derivatives trade needs

Before naming a single vendor, ask the clean question: what functions must exist for a derivative to be traded, risk-managed and reported? The answer is six functional blocks, in sequence, plus a set of cross-cutting capabilities that touch all of them. This functional view is stable even as vendors change, and it is the right mental model because it separates what must happen from who happens to do it.

Figure 14 The six functional blocks a trade flows through, plus the cross-cutting capabilities — data, surveillance, compute and resilience — that touch every block.
Functional block What it must do
Execution & connectivity Reach venues and counterparties — exchange gateways, algos, RFQ platforms, FIX, market data
Booking & trade capture Record the trade once, accurately, as the golden source for everything downstream
Pricing & risk engine Value trades and compute Greeks, XVA and scenarios across listed and OTC
Clearing & collateral Connect to CCPs and clearing brokers, calculate and optimise margin, move collateral
Risk, limits & P&L Compute VaR and stress, monitor limits, produce independent valuation and daily P&L
Reporting & surveillance Produce EMIR/MiFIR reports, best-execution evidence, and market-abuse monitoring

The integration challenge is the whole game: a trade must flow cleanly from execution to booking to risk to clearing to reporting with no re-keying. The value, and the pain, live in the interfaces between these blocks and in keeping listed and OTC positions in one consistent risk and valuation view. A trade captured once and flowing straight through is cheap and safe; a trade re-keyed between systems, or valued differently by the front office and by risk, is expensive and dangerous. This is where most transformation effort goes, and why the cross-cutting capabilities — reference and market data, surveillance, the compute grid, and resilience — matter as much as the blocks themselves.

The latency arms race — where microseconds are the product

In liquid listed markets, execution is an engineering contest measured in microseconds, and the architecture that wins it is a distinct discipline. When a market maker streams thousands of quotes a second and must re-price the instant the market moves, the time from a market-data tick arriving to an order leaving — the tick-to-trade latency — decides who captures the flow and who is picked off on a stale quote. This is why the most competitive listed making has migrated to specialist firms and to purpose-built low-latency stacks: it is as much a hardware problem as a trading one.

Figure 15 The tick-to-trade loop: market data in, normalise and book, price and decide, check risk, order out — the whole path runs in microseconds, co-located and often on FPGA.

The techniques are specific and they compound. Co-location places the trading servers in the same data centre as the exchange's matching engine, cutting the physical distance a signal travels. Kernel bypass lets the application read the network directly without the operating system's overhead. FPGAs — programmable hardware chips — run the pricing and risk-check logic in nanoseconds rather than software milliseconds. And pre-trade risk checks must sit inside this hot path, because even the mandatory limit check cannot be allowed to add meaningful delay. The engineering goal is a straight, minimal, deterministic path from tick to order, with every microsecond of jitter hunted down.

Vendor landscape — buy versus build as a map of advantage

Banks buy the commoditised functions and build the ones that are a source of edge — and where they draw that line is a map of competitive advantage. No bank builds its own exchange gateway or its own generic booking system; those are commoditised and bought. But the pricing models, the XVA engine, the auto-hedging logic and the risk analytics that decide how well the desk prices and hedges are often built in-house, because they are exactly where a desk can be better than its competitors. The buy-versus-build line is therefore not a procurement detail; it is a statement about where the bank thinks its advantage lies.

A handful of vendor names dominate the European landscape, and it is worth knowing them as the map, not as an endorsement. For cross-asset front-to-back platforms, Murex (MX.3) and Calypso (now Adenza) are dominant in EU banks, with FIS and Openlink strong in treasury and commodities. For analytics libraries, Numerix and FINCAD supply pricing and XVA components, though many banks run their own quant libraries. For OTC execution, Tradeweb and Bloomberg run the multi-dealer RFQ platforms, alongside regulated SEFs and OTFs. For clearing and margin, LCH SwapClear clears OTC rates, Eurex Clearing, ICE and CME clear their respective products, ISDA SIMM is the standard uncleared-IM model, and utilities such as AcadiaSoft coordinate collateral. For market data, Bloomberg, Refinitiv and ICE dominate.

Functional block Typically bought Often built in-house
Execution & connectivity Exchange gateways, RFQ platforms (Tradeweb, Bloomberg) Low-latency auto-quoting and algo logic
Booking & trade capture Cross-asset platform (Murex, Calypso) Golden-source data model and controls
Pricing & risk engine Analytics libraries (Numerix, FINCAD) Proprietary pricing, XVA and hedging models
Clearing & collateral CCP/broker connectivity, SIMM, collateral utilities Collateral optimisation logic
Risk, limits & P&L Vendor VaR/FRTB components Proprietary scenario and limit frameworks
Reporting & surveillance Reporting and surveillance platforms Reconciliation and data-quality controls

Data lineage — one trade, followed front to back

Follow a single trade from execution to regulatory reporting and you see the cascade principle: every downstream object depends on an upstream one, so an error at the top reaches the bottom. The fill or RFQ creates a price and size. That becomes a golden trade record at booking, stamped with a unique identifier. The pricing and risk engine derives the Greeks, XVA, P&L and VaR from that record. Clearing and collateral derive the margin call and collateral movement from the position. Regulatory reporting derives the EMIR and MiFIR reports and the audit trail from the same source. Nothing downstream is independent; each object is computed from the one before it.

Figure 16 One trade, captured once, cascades to risk, margin and reporting. An error or a duplicate at the top contaminates every number below it.
Stage System Data object created
Execution Venue / RFQ platform Fill or quote: price, size, terms
Booking Trade-capture platform Golden trade record with a unique trade identifier
Pricing & risk Pricing / risk engine Greeks, XVA, P&L, VaR
Clearing & collateral Collateral system Margin call, collateral movement
Reporting Reporting platform EMIR trade report, MiFIR transaction report, audit trail

The anatomy of a trade record — and why deduplication is hard

The golden trade record is the atom of the whole system, and its fields each drive something downstream. A derivative trade record carries, at minimum: the product type and payoff; the notional; the rate, strike or price; the start and maturity dates and any reset schedule; the counterparty and its legal-entity identifier; the clearing route and CCP; and the unique trade identifier. Each field drives a downstream calculation — the notional and rate drive the Greeks and margin, the counterparty drives CVA and credit-line checks, the clearing route drives which margin regime applies, the maturity drives the lifecycle events. A single wrong field silently corrupts every calculation that depends on it.

Deduplication is a mission-critical, non-trivial problem, because the same trade can appear more than once across systems. A trade executed, given up, cleared and reported may be represented in several systems, and if the reconciliation logic double-counts it — or fails to match the two sides of a novated trade — the risk, margin or regulatory numbers are wrong. Matching, deduplicating and reconciling trade records across execution, booking, clearing and reporting is one of the hardest and least glamorous problems in the stack, and it is where a large share of operational breaks originate.

Identifiers — the codes that make a trade unambiguous

A trade must be identifiable without ambiguity across firms, venues and regulators, and a small set of standard identifiers does that job. The ISIN identifies a security or standardised instrument; the LEI (Legal Entity Identifier) identifies the legal entity on each side of the trade; the UTI (Unique Trade Identifier) identifies the individual trade across both counterparties' reports so regulators can match the two sides; the UPI (Unique Product Identifier) classifies the product type for reporting; and FIGI provides a vendor-neutral instrument mapping. These are not bureaucratic overhead — they are what let a trade reported by two different banks be recognised as the same trade, which is the entire premise of trade-repository reporting.

Identifier What it names Where it bites
ISIN A security or standardised instrument Listed instruments; instrument reference data
LEI The legal entity on each side Counterparty identification in every report
UTI The individual trade, shared by both sides Matching the two sides of a trade at the repository
UPI The product type Classifying the trade for regulatory reporting
FIGI Vendor-neutral instrument mapping Cross-vendor instrument reconciliation

Integration patterns — how the systems actually talk

Systems connect in one of three ways, and the migration between them is a real modernisation, not just a speed upgrade. The legacy default is batch: systems exchange files on a schedule, typically overnight, which is simple but introduces latency and means risk and reconciliation lag reality. The modern on-demand pattern is the API: a system requests exactly the data it needs, when it needs it. The most modern is event streaming: systems publish events — a trade booked, a price moved, a margin call issued — onto a bus (Kafka-style) that others consume in real time.

Why streaming beats batch by more than speed

Moving from overnight batch to event streaming is not merely faster; it removes the need for reconciliation. Under batch, two systems each hold their own copy of the state and must be reconciled the next morning, and breaks are found after the fact. Under event streaming, systems consume the same event stream and stay in step continuously, so the divergence that reconciliation exists to catch does not accumulate in the first place. For a derivatives book that must value, margin and report in near-real time, that shift — from reconciling stale copies to consuming a shared truth — is one of the highest-value pieces of modernisation available.

Two capabilities sit above the integration layer and deserve naming: immutable audit and the proprietary pricing engine. Compliance-critical records — orders, quotes, decisions, communications — must be append-only: written once and never altered, so the audit trail is trustworthy. Governance platforms enforce this immutability. And the pricing and risk engine is the one component that is almost always proprietary or heavily customised, because it encodes the models and hedging logic that are the desk's actual edge. A bank will buy its booking system and build its pricing engine, and that choice tells you exactly where it believes its advantage lies.

Market data — the other torrent the desk must tame

A derivatives desk consumes a second flood of data as large as its trade flow: market data, and moving, storing and serving it is a major discipline in its own right. Prices, quotes, curves, volatility surfaces and reference data pour in continuously from exchanges and vendors — Bloomberg, Refinitiv, ICE — and every pricing, risk and surveillance calculation depends on them being timely, clean and consistent. A stale or wrong price does not just misprice one trade; because of the cascade, it corrupts the Greeks, the margin and the P&L derived from it. Market data is therefore not a feed to be plugged in; it is infrastructure to be engineered.

Three problems define the domain. Feed handling and normalisation: raw exchange feeds arrive in dozens of formats and must be normalised into one internal representation the systems can use. Tick storage: the desk must capture and query enormous time-series histories — every tick of every instrument — which is why specialist time-series databases such as kdb+ are standard in trading, built to store and query billions of records fast. Curve and surface construction: raw prices must be assembled into the yield curves and volatility surfaces that pricing actually uses, a calculation that must be correct and reproducible. Getting market data right is quiet, unglamorous, and the foundation everything else stands on — the "data first, models second" refrain made literal.

The compute problem — risk is an industrial-scale calculation

Valuing and risk-managing a derivatives book is one of the largest routine computations in the bank, and finishing it in time is a genuine engineering constraint. Every night, and increasingly through the day, the book must be fully revalued under many scenarios: VaR and stressed VaR, expected shortfall, the full suite of XVAs, and FRTB's capital calculations, each requiring the book to be priced thousands of times under shocked market states. For a large book this is billions of individual pricings, and the overnight batch that produces the morning risk report strains to finish before the desk needs it. A batch that overruns means the desk trades the new day partially blind.

This is why cloud elasticity and smarter approximation are among the highest-value pieces of modernisation. Fixed on-premise grids are sized for the average night and overwhelmed on the heavy one — precisely the volatile day when the numbers matter most. Bursting the calculation to the cloud lets the desk run heavier, more frequent scenarios without owning idle capacity year-round. In parallel, surrogate models — fast approximations, often machine-learned, of the heavy pricing functions — let the desk estimate XVA and full-revaluation risk in near-real time rather than waiting for the exact overnight number. The compute problem is not a back-office detail; it decides whether a desk can see its risk when it needs to.

Messaging protocols — two worlds, two message families

Derivatives run on two messaging layers — a fast front-office layer and a structured post-trade layer — and the split maps precisely onto the front-office/back-office divide. In the front office, the FIX protocol carries orders and executions: it is fast, standardised and machine-readable, and it is how orders reach venues and prices come back. For OTC confirmation, FpML — an XML standard — carries the full economics of a swap or structured trade, because a bespoke contract needs a richer description than a listed order does. In the post-trade layer, SWIFT MT messages carry settlement and collateral instructions, migrating to the richer ISO 20022 (MX) standard, and CCPs and repositories add their own messaging for margin and reporting.

Figure 17 Two messaging worlds: fast FIX at the front for execution, structured FpML and SWIFT/ISO 20022 at the back for confirmation, settlement and reporting.
Layer Standard What it carries Timescale
Front office (execution) FIX Orders, quotes, fills Milliseconds
OTC confirmation FpML Full swap / structured-trade economics Trade date
Post-trade (settlement) SWIFT MT to ISO 20022 Settlement and collateral instructions T to T+2
Clearing / reporting CCP proprietary + ISO Margin calls, trade and transaction reports Daily

Model risk — validating the models that price the book

Because the pricing and risk engine is proprietary and its outputs decide real money, the models themselves are a governed risk that must be independently validated. A derivatives desk marks bespoke positions on models, computes margin and capital on models, and hedges on model-derived Greeks — so a wrong model does not merely mislead, it loses money and hides the loss. Model risk management is the discipline that controls this: every material model is documented, independently validated by a team separate from the one that built it, tested against market prices and limiting cases, approved before use, and monitored for drift over its life. A model that has not passed validation cannot price a live book — the systems equivalent of the new-product gate in Part 2.

Supervisors treat model risk as first-order, which turns validation into a regulated, auditable process. European banks operate under explicit model-risk expectations — the ECB's targeted review of internal models (TRIM) and ongoing internal-model supervision — and US banks under long-standing model-risk guidance. The practical consequence is a lifecycle with teeth: inventory every model, validate it independently, re-validate periodically, and retire it when it no longer fits. This same governance is what any adoption of machine learning in trading (Part 5) must clear, which is why AI in a derivatives book is validated and explainable by design rather than dropped in as a black box.

Reconciliation — the daily proof that the numbers agree

Reconciliation is the unglamorous discipline that keeps the cascade honest: every day, the same trade and position must be shown to match across the systems that hold it. The desk's book, the clearing member's records, the CCP's records and the regulatory reports should all describe the same positions and the same margin — and where they diverge, the difference is a break that must be investigated and cleared before it compounds. Breaks arise from the ordinary friction of the machine: a trade booked slightly differently on two sides, a lifecycle event processed in one system but not another, a stale price, a duplicated record. Most are trivial; a few hide real errors or real losses, which is exactly why every break is chased.

The strategic point is that reconciliation is a symptom, not a cure — and modern architecture aims to need less of it. Under the legacy batch model, systems hold their own copies of state and reconcile them the next morning, so breaks are found after the fact and the reconciliation team is permanently busy. The event-streaming architecture of the previous section attacks the root cause: if systems consume a shared event stream rather than maintaining independent copies, the divergence that reconciliation exists to catch does not accumulate. A desk that has invested well in its data architecture spends less on chasing breaks and carries less risk of a break hiding a real loss — which is why "reduce reconciliation" is one of the clearest measures of whether a transformation actually worked.

Operational resilience — the moments with no second chance

A derivatives book has moments where failure is irreversible, and resilience is the discipline of guaranteeing those moments cannot fail. A margin call must be met on time or the desk is in default; a pricing engine that goes dark during a volatile session leaves the book unhedged and blind; a clearing connection that drops on a stress day means positions cannot be managed when it matters most. Unlike an issuance deal with a single listing morning, a derivatives book has these no-second-chance moments constantly — every margin cycle, every volatile open. Resilience is not a nice-to-have; it is the property that keeps "mission-critical" from becoming "mission-failed".

Concept What it guarantees Why it matters on the critical day
Redundancy No single point of failure; duplicate systems and links A dropped feed or server does not blind the book
Automatic failover Instant switch to a standby without human action A volatile open cannot wait for a manual restart
Disaster recovery The ability to restore from a separate site A site outage does not stop margin or hedging
Business continuity People and processes keep running through disruption The desk can still meet margin calls in a crisis

DORA — operational resilience becomes law

The resilience that Part 4 has treated as good engineering is now, in Europe, a legal obligation with teeth — and its name is DORA. The Digital Operational Resilience Act, applying across the EU from January 2025, turns the informal expectation that critical financial systems must not fail into a binding regime. It requires firms to manage their information-and-communication-technology risk explicitly: to map their critical systems and dependencies, to test their resilience (including threat-led penetration testing), to report major ICT incidents to supervisors within tight deadlines, and to manage the risk of their third-party technology providers — including, pointedly, the cloud platforms and vendor systems a derivatives desk depends on.

For a derivatives business, DORA lands squarely on the systems this part has described. The pricing and risk engines, the clearing connectivity, the market-data feeds and the surveillance stack are exactly the critical ICT services DORA governs, and the cloud providers that Part 5's transformation agenda relies on are exactly the third parties it scrutinises. The effect is to make resilience auditable: no longer "we believe our failover is fast enough" but a documented, tested, reported obligation. Read alongside the resilience discipline above, DORA is the point where the no-second-chance moments of a derivatives book stop being an engineering aspiration and become a supervised requirement.

Five regulatory regimes, each read as a system specification

Post-2008 regulation does not sit beside the systems; it is written into them, and the honest way to read each regime is as a specification the technology must implement. Five regimes shape every system in the derivatives chain, and each is better understood as an engineering requirement than as a legal summary.

EMIR mandates central clearing for standardised OTC classes, bilateral margin (UMR) for uncleared trades, risk-mitigation techniques, and reporting to trade repositories — a specification for clearing connectivity, margin calculation and a vast reporting pipeline. MiFID II / MiFIR imposes the Derivatives Trading Obligation that forces liquid standardised OTC derivatives onto venues, plus transparency, best-execution evidence and transaction reporting — a specification for venue connectivity and best-ex data capture. UMR with ISDA SIMM specifies exactly how uncleared initial margin is computed and exchanged — a specification the collateral system must implement precisely. Basel, including FRTB, sets the market-risk capital framework and makes cleared trades far cheaper in capital than uncleared — a specification that drives both the risk engine and the economics of clearing choice. And MAR (Market Abuse Regulation) requires surveillance of trading and communications — a specification for the surveillance stack.

Regime What it requires The system it specifies
EMIR Clearing obligation, UMR margin, reporting to repositories Clearing connectivity, margin, reporting pipeline
MiFID II / MiFIR Trading obligation, transparency, best-ex, transaction reporting Venue connectivity, best-ex capture, reporting
UMR / ISDA SIMM Standardised computation of uncleared initial margin Collateral and margin engine
Basel / FRTB Market-risk capital; cleared far cheaper than uncleared Risk engine, capital calculation, clearing economics
MAR Surveillance of trading and communications for abuse Surveillance and comms-capture stack

  1. Why start with functional architecture before naming vendors?

  2. State the cascade principle and one consequence.

  3. What is a UTI and why does it matter for reporting?

  4. Why is moving from batch to event streaming more than a speed upgrade?

  5. Explain the FIX-versus-SWIFT/FpML split.

  6. Why is redundancy alone insufficient for resilience?

  7. Read one regulatory regime as a system specification.

4

  1. What is a UTI for?

  2. A golden trade record is duplicated. What follows?

  3. Why is event streaming more than a speed upgrade?

  4. What does the functional view reveal about buy-versus-build?

5Risks, Failures and Transformation

A derivatives book is leveraged and non-linear, which means a small move can create a large loss — so this part is about how the risk is classified, how a well-run desk is distinguished from a lucky one, how books actually blow up, and where technology genuinely changes the economics. The failures here are real, chosen because each illustrates a distinct way a derivatives operation dies. Read them not as history but as the three shapes of the same underlying danger.

Six ways a derivatives book takes risk

The risks of a derivatives book are best organised into six mutually distinct types, because a taxonomy is a map of controllability, not just a list. Each risk has a different owner and a different main defence, and knowing which is which is how a desk decides where to spend its control effort. The six are market, liquidity and funding, counterparty and credit, model and valuation, operational and settlement, and conduct and regulatory.

Risk type What it is Who owns it Main defence
Market Loss from moves in price, volatility, rates The trading desk Delta/gamma/vega hedging; Greek and VaR limits
Liquidity & funding Cannot hedge, exit or fund margin when needed Desk + treasury Liquidity buffers; funding lines; limits on illiquid risk
Counterparty & credit A counterparty defaults on an uncleared trade XVA desk + credit risk Clearing; collateral (CSA); netting; CVA pricing
Model & valuation A wrong model or stale mark hides losses Quant + product control Independent price verification; model validation
Operational & settlement Mis-booking, failed give-up, bad data, fraud Operations + tech Straight-through processing; reconciliation; controls
Conduct & regulatory Market abuse, mis-selling, reporting failure Compliance Surveillance; best-ex; reporting controls
Figure 18 The six risks as a map of controllability: market risk is taken and hedged, counterparty risk is engineered away by clearing, and operational and conduct risk are controlled by process.

These six are not separate in a crisis; they cascade into each other, which is exactly what makes a blow-up a blow-up. A market gap triggers a margin call (market becomes funding); funding it forces selling into a falling market (funding becomes liquidity); the forced sale moves correlations against the book (liquidity becomes model, as the hedges that were supposed to hold no longer do). Understanding the taxonomy is understanding both the six clean categories and the ugly ways they chain together under stress.

Stress testing — imagining the move before it happens

Because value-at-risk describes normal markets and blow-ups happen in abnormal ones, the real defence is stress and scenario testing. VaR answers "how much could the book lose on a typical bad day?" — a useful number, but one calibrated to recent, orderly conditions, which is precisely when it is least relevant. Stress testing asks a different question: what if rates jump 100 basis points, volatility doubles, a currency peg breaks, or liquidity vanishes? These scenarios reveal exposures that normal-market VaR hides, especially in a non-linear book where a large move triggers effects a small move never would. Reverse stress testing inverts the question — what set of moves would actually break this book? — and works backward to the vulnerability, which is often more revealing than any forward scenario.

Supervisors run the same exercise on banks and CCPs, and for the same reason. European authorities stress-test both dealers and clearing houses precisely for margin procyclicality, liquidity in stress, and concentration — the mechanisms that turn a market move into a spiral. For the desk, the discipline is layered: Greek and position limits cap the risk directly, VaR monitors the normal envelope, stressed VaR and expected shortfall capture the tail that ordinary VaR misses, and scenario tests probe the specific breaks. No single measure is sufficient; the point of the layers is that each catches what the others miss.

How books actually blow up

Before the case studies, name the mechanisms, because every real failure is a combination of them. Prices gap and jump on news rather than moving smoothly, so a delta hedge calibrated for small moves is useless in a jump and short-gamma or short-volatility books lose fast. Liquidity evaporates in stress: the market you rely on to hedge widens or vanishes, and the hedge you assumed is not there. Margin spirals form when a big move triggers large variation and initial-margin calls, funding them forces selling into a falling market, and the selling causes more losses and more calls — a doom loop. Correlations and bases break: hedges rely on relationships holding, and in a crisis they snap, so a "hedged" book suddenly carries huge net risk. Model and mark risk lets a wrong assumption or stale valuation hide losses until reality arrives. And operational and rogue risk — mis-booking, failed give-ups, breached limits, concealed positions — is what controls and reconciliation exist to catch early. The three cases below are these mechanisms in real combination.

Case one — Barings, 1995: the control that was not there

Barings Bank collapsed because one trader controlled both the trading and its own settlement, so no independent check existed. Nick Leeson, running the bank's Singapore futures operation, was permitted to be both the front-office trader and the back-office settlement supervisor — a fatal breach of the segregation that Part 3's information barriers and Part 2's internal gates are supposed to guarantee. He took large unauthorised positions in Nikkei 225 futures and options and hid the mounting losses in a concealed error account. When the Kobe earthquake in January 1995 sent the Nikkei sharply down, his positions — including short options and long futures — blew up, and the concealed losses, ultimately around £800m (roughly $1.3bn), exceeded the bank's entire capital. Barings, over two centuries old, was sold to ING for a nominal £1.

This is the operational and rogue failure mode in its purest form. There was no market cleverness to it; there was a missing control. The lesson maps directly to the internal gauntlet of Part 2 and the independent-control principle of Part 3: the trader who owns the risk must never own the leash. Every reconciliation, every segregation of duties, every independent product-control function exists because Barings proved what happens when they are absent.

Case two — Long-Term Capital Management, 1998: the model that assumed the world would stay normal

LTCM was a hedge fund of Nobel laureates whose highly leveraged convergence trades unravelled when a crisis made everything move together. The fund ran relative-value and convergence strategies — buying the cheap instrument, selling the rich, expecting them to converge — at very high leverage, on the assumption that these small, diversified bets were low-risk because they were uncorrelated and mean-reverting. When Russia defaulted in August 1998, a global flight to liquidity made every risky spread widen at once: the diversification vanished, correlations went to one, and the leverage that amplified small gains now amplified catastrophic losses of around $4.6bn. Because the fund's positions were so large and so entangled with its dealer counterparties, the Federal Reserve organised a roughly $3.6bn recapitalisation by fourteen banks to unwind it in an orderly way.

This is the model and liquidity failure mode. The models were not wrong in normal conditions; they were wrong about the tail, where correlations break and liquidity disappears — exactly the regime that VaR underestimates and stress tests exist to catch. It also shows why a dealer must understand its clients' leverage: LTCM's counterparties were the dealer banks, and their exposure to a single overleveraged fund is the thread that connects this case to the next.

Case three — Archegos, 2021: concentration no single dealer could see

Archegos was a family office that built enormous, concentrated equity positions through total-return swaps with several prime brokers at once, and when the positions fell, the margin calls could not be met. Total-return swaps are derivatives that give economic exposure to a stock without owning it, with the dealer hedging by holding the shares. Archegos used them to take highly leveraged, concentrated bets on a handful of names, spread across multiple banks — so no single dealer saw the aggregate leverage. When stocks such as ViacomCBS fell in March 2021, the variation-margin calls exploded; Archegos could not meet them; and the banks, each discovering they held a slice of the same concentrated unwind, raced to sell the underlying shares into a falling market. The fire sale deepened the losses. Credit Suisse alone lost around $5.5bn, with billions more across Nomura, Morgan Stanley and UBS.

This is the counterparty concentration and margin spiral failure mode. Every mechanism from the blow-up list appears: a gap in the underlying, a margin call that could not be funded, a forced sale into vanishing liquidity, and a correlation break as every bank sold at once. The specific lesson is about seeing aggregate client leverage across the market, not just on your own book — a counterparty risk that clearing does not solve because these swaps were bilateral.

Case Failure mode Where it maps in the lifecycle The lesson
Barings (1995) Operational / rogue — no segregation Capture, control, the internal gates The trader must never own the leash
LTCM (1998) Model / liquidity — tail correlations Managing the book; VaR vs stress Diversification and leverage fail together in the tail
Archegos (2021) Counterparty concentration / margin spiral Clearing, margin, credit See aggregate client leverage, not just your own book

When the CCP itself becomes the risk

Clearing moved risk to the centre; it did not abolish it — and that makes the central counterparty itself systemically critical. By concentrating and mutualising credit risk, a CCP ensures that no single member default cascades through the market. But the same concentration means the CCP is now "too important to fail": if a CCP were ever overwhelmed — by a member default larger than its waterfall, or by a margin model that proved inadequate in a real stress — the failure would be systemic in a way no single bank's failure would be. This is why CCP margin models, default funds and resilience are so heavily regulated and stress-tested by supervisors.

Two second-order risks deserve naming, because they surfaced in real stress. The first is margin procyclicality: margin models that raise initial margin sharply when volatility spikes can force members to find collateral at exactly the worst moment, amplifying a spiral rather than dampening it — a live concern after the 2020 dash-for-cash and the 2022 commodity-margin spikes. The second is concentration at the CCP: because clearing is concentrated in a few systemically important houses, a problem at one has outsized reach. Clearing is a vast improvement on the pre-2008 bilateral web, but it relocated the systemic question from "which bank fails" to "can the clearing house withstand the stress it was built to absorb".

Telling a good desk from a lucky one

A single quarter's profit tells you almost nothing, because a desk can make money by taking hidden risk that has not yet cost anything. The discipline is to judge a desk by the quality and repeatability of its process, not the headline P&L, and a handful of key indicators do that. The point of these metrics is to separate skill — steady spread capture, disciplined risk use — from luck, which is a large directional bet that happened to pay.

Metric What it measures Why it separates skill from luck
Spread capture Realised bid-offer earned per trade The core, repeatable edge of market making
Hit / win ratio Share of quotes and trades that are profitable Consistent small wins beat rare big ones
Hedge slippage Cost of hedging versus theory Cheap, disciplined hedging is the real skill
Return on risk (P&L vs VaR) Profit earned per unit of risk taken High P&L on huge risk is luck, not edge
Inventory turnover How fast the book recycles risk Flow-driven desks turn risk; punters sit on it
Limit breaches Frequency and size of limit exceedances A clean limit record signals real control

The economics of a market-making franchise

A trading desk is a fixed-cost, scale business, and that shapes every strategic decision about it. The pricing engine, the risk and compute grid, the clearing connectivity, the surveillance stack and the quant team are largely fixed costs — they cost roughly the same whether the desk trades a little or a lot. The marginal cost of one more trade, once the machinery exists, is small. This is why scale matters so much in derivatives: a desk that spreads its fixed technology cost over more flow earns a better margin, sees more offsetting risk to net internally, and can invest more in the very technology that widens its edge. It is a compounding advantage, and it is why the business tends toward a handful of large franchises in each product rather than many small ones.

The same economics explain why cost pressure and transformation are permanent features, not passing phases. Because the fixed cost of running a modern derivatives book is enormous — and rising, as regulation adds reporting, margin and capital machinery — desks are under constant pressure to automate the commoditised flow, cut the cost of the post-trade machine, and push compute to cheaper elastic capacity. A desk that cannot lower its cost-to-serve on vanilla flow loses that flow to a cheaper competitor, bank or non-bank, and with it the netting and the relationships. The transformation agenda of this part is therefore not optional modernisation; it is the price of staying in the business.

Five ways technology is remaking derivatives trading

The transformation of derivatives trading is real but specific, and the honest test for any trend is whether it changes economics or risk rather than merely the appearance of the business. Five trends matter most, and each attaches to a specific step in the lifecycle rather than floating as a generality.

Automated market making puts auto-pricing and auto-hedging engines in charge of liquid listed and vanilla OTC flow, with no human in the loop for the standard business — the norm already, and spreading. OTC onto venues continues as SEFs, OTFs and RFQ platforms pull standardised swaps and options onto screens, narrowing the gap with the listed world — a change driven by regulation as much as technology. Straight-through processing automates the flow end to end, from execution to booking, clearing and reporting, removing the re-keying that is the source of so much operational risk. Clearing and collateral automation applies the same logic to the post-trade machine: automated margin calculation, collateral optimisation and reconciliation cut cost and error. And cloud and compute elasticity lets risk and XVA calculations burst to the cloud for scale, running heavier scenarios faster than fixed overnight batches ever allowed.

Trend Attaches to Why it matters
Automated market making Execution and hedging Liquid flow is priced and hedged with no human in the loop
OTC onto venues (SEF/OTF) Execution Standardised swaps move to screens; the two worlds converge
Straight-through processing Capture to reporting Removes re-keying and the operational risk it creates
Clearing & collateral automation Clearing and margin Cuts cost and error in the post-trade machine
Cloud & compute elasticity Risk and XVA Heavier scenarios, faster, than fixed batches allowed

One much-hyped technology deserves an honest, brief verdict: distributed ledger. Distributed-ledger and tokenisation projects promise to collapse the reconciliation, clearing and settlement layers into a single shared record, and there are genuine pilots in collateral mobility and tokenised assets. But in derivatives trading the reality remains early: the hard problems are still clean data, consistent valuation and resilient plumbing, and a shared ledger does not by itself solve any of them. Treat it as a technology to watch and pilot at the margins, not as a near-term reshaping of how derivatives trade — and apply the same filter as everything else: it earns attention only where it actually moves cost or risk.

Generative AI in derivatives trading — where it earns its place

Generative and machine-learning tools have real, emerging uses in derivatives trading, but every one of them is governed as tightly as it is useful. The genuine uses cluster in a few places. Smarter auto-quoting: models optimise spread, skew and hedge decisions from flow and win-rate data. Risk at scale: machine learning and cloud accelerate XVA and full-revaluation risk, with surrogate models approximating heavy calculations in near-real time. Hedge optimisation: algorithms choose the cheapest, most effective hedges and net across the book. Surveillance and operations: anomaly detection flags market abuse and mis-bookings, and large language models summarise, document and speed reporting. Collateral and margin analytics: models forecast margin calls and optimise what to post.

Against each use stands a governing risk, and the accounting is honest only when both are stated together. Auto-quoting and auto-hedging can build risk fast, so they need hard limits, kill-switches and human oversight. Models can be wrong — a mispriced surface or a bad hedge model loses real money — so independent validation is essential. Large language models can hallucinate, so they cannot be trusted with unverified outputs in a pricing or reporting path. And any model touching client flow risks handling material non-public information, so confidentiality and access controls bind it. The governing principle is that AI is an assistant, not an autopilot: it augments pricing, risk and operations, but it does not remove the trader's judgement or accountability.

Use of GenAI / ML The gain The governing risk
Smarter auto-quoting Better spread, skew and hedge decisions Builds risk fast; needs limits and kill-switches
Risk at scale (surrogate models) XVA and full-reval risk in near-real time Model risk; needs independent validation
Hedge optimisation Cheaper, better-netted hedges A bad hedge model loses real money
Surveillance & anomaly detection Catches abuse and mis-bookings earlier False positives; explainability of alerts
LLM drafting & reporting Faster documentation and reporting Hallucination; MNPI and confidentiality

  1. Give the six risk types and say which two are the usual detonators of a blow-up.

  2. Match each case to its failure mode.

  3. Why is one quarter's profit a poor measure of a desk?

  4. State the filter for a transformation claim, with one pass and one fail.

  5. Why is a human in the loop a control rather than a limitation for AI in trading?

5

  1. Which two risks are the usual detonators of a blow-up?

  2. Match: Archegos 2021.

  3. Why is one quarter's profit a poor measure of a desk?

  4. Which transformation passes the filter?

Module Summary — The Six Ideas That Matter Most

If everything else fades, these six ideas are the load-bearing structure of derivatives trading. Each was built up through the module; here they stand together.

Idea 1 — Trading is price plus risk, not product design

A derivatives desk owns exactly two things: the price it shows the market and the risk it holds afterward. It quotes two-way prices, executes flow, and warehouses and hedges the resulting risk — it makes markets, it does not design products, which is structuring's job. Everything else in the module is a consequence of these two verbs: quote the price, manage the risk. A desk that forgets it is paid to warehouse risk and provide liquidity, and starts gambling on direction, has stopped being a trading desk and become a punter.

Idea 2 — Data cascades, so errors are never contained

One trade, captured once as a golden record, cascades to risk, margin and reporting — so an error upstream contaminates every number downstream. The trade record is the atom, with each field driving a downstream calculation, and deduplication and data quality are mission-critical precisely because the cascade carries a corrupted or duplicated record into the Greeks, the margin call and the regulatory report alike. This is why the recurring refrain of derivatives technology is data first, models second: the cleverest pricing engine is worthless running on bad data.

Idea 3 — The risk lives on the book, and the Greeks are how you manage it

The money and the danger are not on the trade ticket; they are on the warehoused book, managed through the net Greeks across both worlds. A desk nets delta, gamma, vega, theta and rho across every listed and OTC position, delta-hedges the direction, manages gamma and vega, and routinely uses liquid listed instruments to hedge bespoke OTC risk. Once delta-hedged, an options book is really a volatility position, trading the surface, skew and term structure. Hedging enough to stay safe without hedging away the spread — netted across a large book — is the actual skill.

Idea 4 — Clearing, margin and compliance are core architecture, not an afterthought

Post-2008, credit exposure must be collateralised, and the plumbing that enforces it is as much of the business as the trading. CCP novation, initial and variation margin, and UMR removed most counterparty risk but added a funding cost that never goes away — the XVAs turn model fair value into a real, chargeable price. Messaging standards, resilience and reporting are not background; they are the specification the systems must satisfy, and the moments with no second chance — every margin call, every volatile open — are exactly where resilience is measured.

Idea 5 — The two worlds are one trade-off and one book

Listed and OTC are the same trade-off on every dimension — safety and liquidity versus flexibility and precision — and the desk runs them as a single book. The same three products trade in both worlds; what differs is only how they are executed, cleared and margined. The desk nets risk across both, hedges OTC exposure with listed instruments, and relies on basis and arbitrage to keep the two worlds tethered. Reading listed and OTC separately is the beginner's error; reading them as one book with two front doors is the professional's habit.

Idea 6 — Real transformation moves cost or risk — including GenAI, governed

The transformations worth attending to are the ones that change economics or risk, not the ones that merely change the surface. Automation, straight-through processing, clearing and collateral automation, and cloud compute pass that filter; speed for its own sake does not. Generative AI has genuine uses in quoting, risk-at-scale, hedge optimisation and surveillance, but it rests on clean data and a human in the loop, and it is governed as tightly as it is useful — an assistant, not an autopilot.

Process Cheat Sheet — The Whole Trade on One Page

A single reference for the entire derivatives trading flow — the lifecycle, the numbers, and the one-sentence version. All figures are illustrative rules of thumb, not market levels.

The lifecycle at a glance

Stage Timing Core activity Key output Lead & systems
1. Execute Seconds Strike a price — order book (listed) or RFQ (OTC) Fill / agreed trade Trader, auto-quoter, FIX, RFQ platform
2. Capture & confirm Minutes Book once, accurately; confirm terms Golden trade record (UTI) Middle office, trade-capture platform
3. Clear / give-up T to T+1 Novate to CCP via clearing broker; or stay bilateral Cleared position Clearing ops, CCP connectivity
4. Margin Daily, for life Post IM; exchange VM against mark-to-market Collateral movements Collateral ops, SIMM, collateral system
5. Manage & hedge Continuous Net Greeks; delta/gamma/vega-hedge; run within limits Hedged book within limits Trader, risk, pricing/risk engine
6. Settle / expire At maturity Cash- or physically-settle; unwind or roll; release collateral Closed position Operations, SWIFT / ISO 20022

Key numbers and rules of thumb

Metric Rule of thumb (illustrative)
Global notional outstanding Hundreds of trillions — but notional is not exposure
Initial margin on a cleared position A small single-digit percentage of notional (a few days of moves)
Variation margin cycle Daily, sometimes intraday in stress
The first hedge to put on Delta — neutralise direction before anything else
What an options book really trades Volatility (vega), once delta is hedged
Default-waterfall order Defaulter margin, guaranty fund, CCP capital, mutualised fund
VaR confidence / horizon Commonly 99%, 1-day (necessary but underestimates the tail)
The four P&L sources Bid-offer spread, flow, basis, volatility premium
The four cost drags Hedging cost, margin funding (MVA), capital (KVA), gap risk
The three blow-up modes Control failure, model/liquidity failure, counterparty/margin spiral
Uncleared initial-margin model ISDA SIMM, posted to a segregated account under UMR
Two messaging worlds FIX (front, milliseconds); FpML / SWIFT / ISO 20022 (back)

The entire deal in one sentence

A derivatives desk quotes a two-way price the market will deal on, executes the flow in the listed or OTC world, warehouses the resulting risk on one netted book, hedges the unwanted Greeks — using liquid listed instruments even for bespoke OTC positions — clears and margins the trade so credit risk is collateralised, and manages the position within limits until it settles, earning spread and flow net of hedging, funding and capital, and surviving because it manages risk rather than gambling.