The LME nickel crisis made trade cancellation a market-control

Summary

Nickel prices rose with extraordinary speed on 7 and 8 March 2022. The FCA later found that the LME's real-time monitoring and volatility-control arrangements were inadequate for severe market stress, that price bands were disabled without effective escalation, and that those weaknesses allowed the price to rise much faster than it otherwise would have.
The LME suspended nickel trading at 08:15 London time on 8 March and cancelled trades entered from midnight. The FCA put the cancelled notional value at approximately US$13.373 billion. That figure is not a measure of realised loss; it is the notional value of transactions removed from the market record.
The High Court and Court of Appeal upheld the cancellation. Their judgments establish that the LME had legal authority to act and faced a credible threat to the clearing system. They do not erase the FCA's separate finding that preventive controls, staff training, escalation and regulatory notification were deficient.
Practical control was distributed. The exchange controlled monitoring, price bands, trading halts, cancellation powers and market notices. LME Clear controlled margin, default and clearing-house risk. Members controlled client information, funding and position reporting. Regulators set and enforced standards but did not operate the overnight market. Market users bore consequences without controlling those systems.
Repair evidence is real but incomplete. The LME introduced daily price limits, expanded over-the-counter position reporting, published clearer cancellation and volatility-control policies, revised risk governance and changed clearing controls. The FCA credited materially enhanced mitigants, and the Bank of England reported that LME Clear's remediation programme had been largely delivered. Public evidence still does not reproduce the March 2022 stress in a way that proves every control would work as intended under an equivalent event.
The durable accountability test is whether the operator can preserve a time-stamped record of control settings, alerts, overrides, decisions and communications; detect risk across venue and OTC positions; give trained staff explicit escalation duties; explain how value is redistributed when trades are cancelled; and publish enough independent assurance to show that repairs remain embedded.

The cancellation is the visible decision, not the whole failure

Trade cancellation attracts attention because it changes completed market outcomes. A buyer who thought it had acquired nickel and a seller who thought it had reduced risk were told that the transactions no longer stood. The intervention created winners and losers relative to the trading record that existed before cancellation. It also protected clearing members and the clearing house from obligations generated by prices that the exchange considered disorderly. Any assessment that looks only at the legal power to cancel, however, starts too late.

The public record shows two linked but distinct accountability questions. First, did the LME have lawful grounds to suspend trading and cancel the affected trades once the market and clearing system were in acute danger? The High Court judgment and the Court of Appeal judgment answer yes. Second, did the LME have adequate systems and controls to monitor, constrain and escalate the disorder before emergency cancellation became the chosen response? The FCA's 2025 Final Notice answers no for important parts of the relevant control framework.

Those conclusions are compatible. A lawful emergency response can follow preventable control deficiencies. A court can find that officials acted rationally on the information and choices available at 08:15 and 09:00 on 8 March while a regulator later finds that the institution should have had better information, better automated protections and better escalation before those times. Treating the court outcome as a complete vindication would collapse two different tests. Treating the FCA fine as proof that the cancellation itself was unlawful would make the opposite error.

The distinction matters beyond metals trading. Exchanges, payment systems, clearing houses and digital platforms all reserve exceptional powers: pause service, reject transactions, roll back records, impose limits or close a market. Such powers can be essential to continuity. They also concentrate control over other people's positions. Accountability therefore has to cover both the use of emergency power and the architecture that determines whether the emergency occurs, how quickly it is detected, what evidence is available and who bears the adjustment.

Who had practical control

The LME was the recognised investment exchange and operator of the nickel market. It controlled LMEselect, the electronic venue; its trading rules; real-time market monitoring; price-band parameters; staff access to override or suspend those bands; the decision to halt trading; the power to cancel, vary or correct trades; member notices; and the explanation given to market users and regulators. Those powers made the exchange the central controller of market order, even though it did not control every entity's position or funding.

Within the exchange, control was not held by a single person. The FCA found that Trading Operations was the only team with responsibility for real-time monitoring. Coverage followed the market across London and Hong Kong, and the relevant overnight period was handled by a Hong Kong team. Senior managers remained available by telephone, but the effectiveness of that arrangement depended on the operating team knowing what conditions required a call. According to the Final Notice, staff had not received adequate training to recognise that intentional, correctly entered trades could still evidence a disorderly market.

They also lacked sufficiently clear instructions for escalating changes to price bands. An organisational chart could therefore show senior availability while the actual control path failed.

LME Clear had a different mandate. As the central counterparty, it calculated and called margin, managed member exposures, maintained default resources and had to protect its own solvency and continuity. Its information about unpaid margin and prospective member defaults became central to the decision to suspend and cancel. The Bank of England supervised LME Clear, while the FCA supervised the exchange. The UK authorities' joint statement of 4 April 2022 accordingly announced separate reviews of LME and LME Clear, with the Prudential Regulation Authority and FCA also engaging firms that held significant positions.

Members controlled another part of the risk. They knew their own exchange and client exposures, received margin calls, decided how to fund them and held information about client positions that was not necessarily visible to the venue. The Oliver Wyman independent review commissioned by the LME found that fragmentation between exchange-traded and OTC activity reduced market-wide visibility. It identified very large short positions, including OTC positions, and more than US$2 billion in missed OTC margin calls attributed to two clients. Members were not passive, but their control did not replace the exchange's duty to run an orderly venue.

Regulators controlled authorisation, standards, supervision, information demands and enforcement. They did not sit at the overnight operations desk or decide whether to suspend a particular band at 04:49. That boundary is important. Regulation cannot be used as a substitute for an operator's real-time judgment, while an operator cannot excuse inadequate controls on the ground that regulators later reviewed the event. The FCA's eventual enforcement action and the Bank's supervisory action show ex post authority. They do not imply ex ante operational control.

End users, including producers, consumers, traders and smaller firms using nickel prices for hedging or contracts, controlled least. They could choose orders, brokers and risk limits, but they could not see the exchange's internal band settings, force an escalation, preserve a trade after cancellation or inspect clearing-member liquidity. Their exposure to a control system they did not operate is why public evidence matters. When control and consequence are separated, the controller must supply the audit trail.

A timeline of pressure, lost safeguards and emergency action

The crisis developed before the cancelled trading window. Oliver Wyman described a combination of geopolitical shock, concentrated short positions, declining liquidity and a short squeeze. Its review recorded the three-month nickel price moving from US$27,080 per tonne on 4 March to US$101,365 during the morning of 8 March, an increase of roughly 270 percent over three trading days. The review was commissioned by the LME and should not be mistaken for a regulatory finding, but its market reconstruction is an important primary case document.

On Friday 4 March, LME Clear issued margin calls totalling about US$3.5 billion, then a record. Russia's invasion of Ukraine and the possibility of sanctions affecting Russian metal had intensified uncertainty. Nickel liquidity was already thinning. The risk was no longer just an unusual price move; it was a feedback loop in which rising prices increased short-side margin calls, funding pressure reduced entities' ability to trade, and lower liquidity made further price moves easier.

The pressure accelerated on Monday 7 March. The Bank of England's July 2022 Financial Stability Report described the three-month contract rising around 60 percent that day, from an opening level of US$29,770 to a close of US$48,000. The FCA used a different but compatible market marker: it recorded the last LMEselect price at US$50,300, more than 65 percent above the 01:00 opening price. These figures should not be forced into one number because they refer to different price points in the trading process.

LME Clear's margin demands also escalated. Oliver Wyman recorded calls of nearly US$7.5 billion by early afternoon on 7 March before the clearing house suspended additional intraday calls. The FCA recorded a US$5.1 billion morning call, also a record at the time, with several members paying late and one amount outstanding. Both figures can be true because they concern different call cycles and cut-off points. Together they show that liquidity pressure was visible before the final price spike.

Price bands were supposed to slow extreme moves. The LME had static and dynamic mechanisms, but the FCA found that their operational purpose had narrowed. Staff largely treated the bands as protection against erroneous orders or rogue algorithms, rather than as mechanisms that might also constrain correctly entered trades in an increasingly disorderly market. The exchange had not published enough information about static bands, and it had not embedded a clear policy for their use under severe but genuine trading pressure.

During 7 March, the London operations team suspended and reapplied bands as prices moved. Senior managers knew that further nickel price rises could lead to multiple member defaults, yet the overnight Hong Kong team was not adequately briefed about the risk or given a specific escalation plan. Messages from market entities reached mailboxes that were not being monitored in real time. This is a classic control failure: relevant signals existed, senior concern existed and people were technically reachable, but the handover and trigger connecting those facts were weak.

After the electronic market reopened at 01:00 on 8 March, prices rose rapidly. By 04:49, according to the FCA, relevant price bands had been suspended and remained unavailable for more than three hours. The operating team did not alert senior management. The FCA concluded that disabling the bands allowed the price to rise much faster than it otherwise would have, increasing exposure for members and LME Clear. The regulator did not claim that intact bands would have solved the concentrated-position or liquidity problem.

It found that the mechanism could have constrained the speed and scale of the move, which would have preserved time for escalation and risk decisions.

The three-month price crossed US$60,000 and later US$100,000. At about 07:24, a clearing analysis using prices around US$80,000 projected approximately US$19.75 billion of additional margin due at the next call. The High Court record says at least five members were expected to default, with four more at risk. That was a forecast under extreme conditions, not a realised loss. The system calculated the prospective variation margin, but LME Clear did not instruct members to pay the US$19.75 billion.

Senior decision-makers met remotely around 07:30. The exchange confirmed suspension at 08:15. Its contemporaneous Notice 22/057 said the market had become disorderly, that prices during the early hours no longer reflected the underlying physical market and that the situation posed systemic risk. It suspended trading and announced that trades executed on or after 00:00 on 8 March in the inter-office and electronic markets would be cancelled.

Cancellation was decided after a further meeting beginning around 09:00 and announced at 12:05. The choice of midnight moved the market back to the last point the LME judged orderly, rather than preserving trades made while price formation was already breaking down. The FCA later calculated the notional value of cancelled trades at about US$13.373 billion. The High Court used an approximate US$12 billion figure. The difference is a matter of source scope and rounding, not evidence of two separate cancellations. Neither number should be described as the exchange's loss or entities' net damage.

Trading remained suspended for eight days. The resumption notice set reopening for 08:00 on 16 March, introduced a 15 percent daily price limit for nickel and imposed temporary position-accountability and reporting measures. The restart itself exposed implementation risk. The LME's 16 March trading update said a system issue had allowed a small number of trades below the lower daily price limit; it halted LMEselect again and cancelled those below-limit trades. Similar limit and execution issues followed over the next sessions. A control rushed into production can reduce one risk while generating another, so restart evidence belongs in the same accountability record.

The market reopened before the whole operating model returned. Asian-hours nickel trading did not resume until March 2023. That gap matters because it reduced the market's normal time coverage for approximately a year. It also gave the LME time to install reporting, monitoring and volatility-control changes. Continuity was therefore phased: first preserve the clearing system, then reopen under constraints, then rebuild a wider trading schedule.

Why cancellation was both protective and distributive

The LME's cancellation power existed in its rules. Rule 22.1, examined in the High Court judgment, allowed the exchange to halt or constrain trading following a significant short-period price movement and, where appropriate, to cancel, vary or correct an agreed trade or contract. The rule recognised that an exchange sometimes has to prefer market integrity and clearing continuity over finality for individual transactions.

The clearing rationale was substantial. If the 8 March trades had stood, the court record indicates that margin obligations tied to extreme prices could have produced a cascade of member defaults. LME Clear was contractually and operationally required to calculate margin by reference to standing trades. A suggestion that the trades could remain valid while margin was calculated at the previous day's close did not provide a lawful or operationally viable answer to that obligation. Cancellation removed the transactions that generated the most extreme new exposure and reset settlement obligations to a point the exchange treated as orderly.

Protection of the system did not make the distributional effect disappear. A entity that bought nickel below the later peak lost a contractual position it valued. A entity that sold at an extreme price escaped an obligation that could have crystallised a large gain for the buyer and a large loss for the seller. Hedgers had to reconstruct exposures in a closed or constrained market. Brokers and clearing members had to reconcile books, customer communications and collateral. Because cancellation changes private outcomes by administrative decision, the operator owes a more demanding explanation than it would for an ordinary trading halt.

That explanation needs four parts. It should identify the rule power; the factual trigger; the alternatives considered; and the reason for selecting the affected time window. The court judgments provide much of that record after litigation. Good emergency governance would preserve it contemporaneously in a decision log, with market data, control status, margin scenarios, attendees, dissent, legal advice boundaries and notification times. Public users do not need privileged legal advice or entity identities, but they do need enough evidence to distinguish a principled market-integrity action from an unexplained redistribution.

The cancellation also created a moral-hazard concern. If entities believe extreme positions will always be rescued, risk discipline weakens. If they believe trades can be cancelled unpredictably, confidence in the venue weakens. The answer is not a promise never to cancel. It is a narrow, published power combined with credible preventive controls, position visibility, deterministic volatility mechanisms, explicit escalation and post-event review. Emergency discretion is most legitimate when the institution can show that it did not rely on discretion as a substitute for preparation.

What the courts decided, and what they did not

Elliott Associates and Jane Street challenged the decisions in judicial review proceedings. The High Court recorded claimed losses of approximately US$456 million for Elliott and US$15 million for Jane Street. Those were the claimants' asserted lost net profits, not damages adjudicated and awarded by the court. The court dismissed all grounds in November 2023. It found that the LME had authority under its rules, was entitled to act without consulting affected entities in an urgent situation and had not acted for an improper purpose.

The Court of Appeal dismissed Elliott's appeal in October 2024. It accepted that contractual rights represented by trades could qualify as possessions for human-rights analysis, but held that the interference was lawful, proportionate and justified by the need to prevent a market and clearing collapse. The court rejected the contention that the LME had improperly favoured one class of market entity. Jane Street did not pursue the same appeal.

The LME's litigation status page records that the UK Supreme Court refused Elliott permission to appeal on 29 January 2025 and that related claims were later discontinued. That is a company-maintained status source rather than a fresh merits judgment, so the durable legal propositions remain those in the published High Court and Court of Appeal decisions.

The courts reviewed the decisions actually made under public-law and human-rights standards. They did not conduct the FCA's later systems-and-controls enforcement assessment. They did not hold that overnight handovers were adequate, that band suspensions were properly escalated, that the exchange had sufficient OTC visibility or that the regulator was promptly informed. Nor did they quantify every entity's economic harm. The legal record resolves the validity of cancellation; it does not certify the whole control environment.

This boundary is central to institutional accountability. A defendant can win litigation because the emergency decision was within its lawful discretion and still face enforcement because its preventive arrangements breached regulatory requirements. Conversely, a regulator's systems finding does not establish that private claimants were entitled to the profits they expected from cancelled trades. Each forum asks a different question and applies a different evidentiary test.

The regulatory record moved upstream from cancellation

The FCA's action was the first enforcement action and fine against a recognised investment exchange. Its March 2025 announcement imposed a penalty of £9,245,900 after an early-settlement discount; without that discount the penalty would have been £13.2 million. The relevant period ran from January 2018 through 8 March 2022, showing that the concern was not confined to one frantic morning.

The Final Notice found breaches of recognition and technical requirements concerning orderly trading and automatic volatility mechanisms. The exchange had designed price controls, but policy, calibration, publication, staff training and escalation did not make them adequate for severe stress. Staff could suspend bands without senior consultation. They were not adequately prepared to identify disorder caused by intentional trades. The only real-time monitoring team did not have a robust overnight escalation route for the event that occurred.

The regulator also found a notification weakness. Although the FCA asked about price-band operation in April 2022, the LME did not disclose until 19 May that the bands had been suspended during the relevant morning. The FCA did not find that omission deliberate, but treated the delay as aggravating. Accountability after an incident depends on candid disclosure of control status, including facts that may make the operator look worse.

The FCA did not say that the LME engineered the price move, deliberately misled the market or acted recklessly. It said the failures were serious, undermined confidence and increased member and clearing-house exposure. It also noted that historic price-band settings were not retained in a form that allowed every detail to be reproduced exactly. That evidence-retention gap is more than an archival inconvenience. If a market operator cannot reconstruct which automated limits applied and when they changed, regulators and users cannot fully test the causal account.

The Bank of England's focus was LME Clear. Its March 2023 supervisory action said independent reviews had identified shortcomings in governance, management and risk-management capabilities. It required stronger governance and independence, clearer management accountability and wider improvements in risk management, with a skilled person appointed to monitor remediation. This was supervisory action, not a monetary penalty.

The Bank's 2025 Financial Market Infrastructures Annual Report said the disruption had undermined confidence and that LME Clear's remediation programme had been largely delivered. It highlighted governance, margin, default-fund, stress-testing and information-sharing improvements, while stating that supervisors would continue assessing whether changes were embedded. That phrasing is careful: delivery of a programme is evidence of repair, but continuing supervision is evidence that durability remains a live question.

The division of regulatory work also prevents a false single-cause story. Venue controls affected the speed of prices. Clearing controls translated prices into liquidity demands. Member and client positions determined who was short and how much collateral was needed. OTC opacity constrained the view across those layers. Governance determined whether warnings moved between them. No one control explains the crisis, but that does not mean responsibility is diffuse beyond use. Each institution can be assessed against the controls it actually possessed.

Harm and cost must be stated without turning exposure into loss

The most certain harm was market interruption. Nickel trading was unavailable for eight days, reopened under tight limits and implementation problems, and lacked Asian-hours coverage for about a year. Producers, consumers and financial users lost normal access to a global reference market. Contracts and hedges linked to LME prices faced uncertainty. Those effects are real even where the public record does not provide a complete monetary total.

The US$13.373 billion cancelled notional is a scale measure, not a damages figure. Notional value does not equal the profit a buyer would have made, the loss a seller avoided or the net market exposure after offsetting positions. The US$19.75 billion margin calculation was a projected call that was never issued, not cash paid or a clearing-house loss. The roughly US$471 million claimed by Elliott and Jane Street represented asserted lost profits in litigation, not judicially verified aggregate harm. Keeping these categories separate is essential to an honest cost account.

There were observable corporate costs. HKEX's 2022 annual report said operating expenses increased in part because of HK$56 million of legal and professional fees primarily related to the nickel incident. It also reported lower LME trading fees and lower chargeable average daily volume across metals, but the report does not support assigning every revenue or volume change to nickel. Market conditions and other products also mattered.

HKEX's 2025 annual report recorded a non-recurring HK$90 million expense for the FCA fine and described the conclusion of the principal litigation. It also reported record LME chargeable average daily volume of 717,000 lots in 2025 and exceptional growth in nickel activity. Recovery in volume is evidence that users continued to use the venue. It is not proof that every user recovered confidence or that controls have survived an equivalent short squeeze.

Members and clients also incurred funding, reconciliation, legal and operational costs that are not comprehensively public. Some received the benefit of cancelled obligations; others lost expected gains or had to replace hedges. The identity, net position and eventual settlement of every affected party are not available in the primary record. It would therefore be improper to produce a single net-winner or net-loss table from public data.

There is a wider confidence cost. A benchmark market depends on users believing that prices are discoverable, rules predictable and trades final except under clearly defined exceptional conditions. The crisis challenged all three beliefs. Confidence cannot be measured solely by subsequent volume. It is reflected in liquidity across conditions, willingness to hold positions, quality of bid-ask spreads, use of the benchmark in physical contracts and belief that governance will be consistent during stress. Public repair evidence should therefore include market-quality measures, not just headline activity.

Smaller firms face a particular asymmetry. A large member can maintain risk teams, multiple brokers, funding lines and direct regulatory engagement. A small manufacturer or trading business may depend on a broker and the LME price without having comparable alternatives. When trading is halted or trades are cancelled, the smaller user still has physical inventory, procurement and customer obligations. Market infrastructure accountability includes explaining how emergency procedures affect users that cannot cheaply duplicate access.

Repair evidence after the crisis

The first repairs were immediate. The LME's interim measures notice introduced daily price limits across physically delivered metals, additional reporting and an independent review. The notice acknowledged the limited visibility over OTC positions and the need to understand their interaction with exchange activity. Temporary position-accountability levels for nickel were tightened and reporting was expanded before reopening.

The independent review produced seven broad recommendations covering market-risk mandate, position controls and enforcement, OTC visibility, volatility controls, operational readiness, member and clearing-house resilience, and nickel liquidity. Its limitation is explicit: Oliver Wyman did not review the LME's decision-making or governance surrounding suspension and cancellation, which were reserved for regulatory processes. The report is therefore strong evidence about market dynamics and recommended controls, but not an independent validation of the disputed decisions.

In March 2023, the LME announced a two-year market-strengthening action plan. The more detailed action-plan document set out enhanced real-time monitoring, early-warning indicators, escalation playbooks, closer LME and LME Clear communication, weekly OTC position reporting, revised volatility controls, margin changes, stress testing, default-fund work and measures intended to rebuild nickel liquidity.

These actions changed the control architecture in useful ways. Weekly OTC reporting gives the operator a better view of total entity exposure. Daily limits create a predictable maximum move and provide time for margin and governance processes. Early-warning indicators and playbooks make escalation less dependent on an individual deciding that genuine trades look suspicious. Cross-entity communication reduces the risk that exchange staff see price disorder while clearing staff separately see liquidity stress without a shared decision.

The public action plan also admits limits. OTC data can be incomplete because positions sit across jurisdictions, affiliates and legal confidentiality regimes. A report received weekly can become stale in a fast market. Position visibility does not ensure that a member can fund margin. A daily limit can trap a market at the limit, reduce liquidity and delay price discovery rather than remove underlying imbalance. Controls must be tested for both their protective effect and their secondary consequences.

The current LME volatility-control policy is more explicit than the pre-crisis public record. It describes dynamic and static bands and daily price limits, says at least two mechanisms apply to relevant contracts, and specifies that the most restrictive mechanism governs. It limits suspension of dynamic and daily controls to technical circumstances and describes a multi-day framework for repeated daily-limit events. Publication makes the rules more predictable and gives auditors a baseline against which actual settings can be tested.

The LME also published a policy on order cancellations and controls. It explains circumstances for cancellation, testing, monitoring and records of control parameters and changes. A written policy is valuable because it reduces uncertainty about who can alter controls and what evidence must be retained. Its effectiveness still depends on live compliance: complete logs, independent review of overrides, trained coverage around the clock and evidence that alerts result in action.

Regulatory change widened the framework. The FCA's Policy Statement PS25/1 strengthened commodity-derivatives position management, trading-venue surveillance, OTC reporting and information sharing. The final rules came into force on 6 July 2026. As of 15 July 2026, they are no longer merely a future commitment. Their existence improves the authority and data framework, but only future supervision can show how consistently firms comply and how useful the information is during stress.

International standards point in the same direction. IOSCO's Principles for the Regulation and Supervision of Commodity Derivatives Markets, revised after the episode, emphasise market surveillance, large-trader information, OTC visibility, position management and powers to respond to disorder. These principles are not a case judgment and do not prove that any single reform would have prevented March 2022. They provide a benchmark for judging whether an exchange can see concentrated risk and intervene in a transparent, proportionate way.

Independent evidence of repair is stronger than self-report. The FCA said the LME had materially enhanced mitigants and gave credit for its review and action plan, even while imposing the fine. The Bank said LME Clear's remediation had been largely delivered, subject to continued assessment of embedding. Those are meaningful external confirmations. They stop short of certifying that no control could fail, and they do not publish every test result.

The strongest defensible conclusion is that the control environment was materially changed and partially independently verified, while performance under an equivalent event remains unproven in public.

Counterfactuals: what a better-controlled response could have looked like

The easiest counterfactual is also the weakest: let every trade stand and allow the market to clear. The court record shows why that was not a responsible option by the time the price approached US$100,000. Projected margin demands and expected member defaults threatened LME Clear. A clearing failure could have imposed wider losses and prolonged closure. Accountability should not require an operator to preserve transaction finality at the cost of probable infrastructure failure.

A more useful counterfactual begins earlier. If price bands had remained active after 04:49 and staff had escalated the rapid move, the FCA found that prices would have risen less quickly. That does not mean the bands would have restored equilibrium. It means senior managers and clearing officials could have gained time to assess margin, contact members, suspend trading at a lower price and preserve a cleaner boundary between orderly and disorderly transactions. The supported inference is that earlier action could have reduced the volume and controversy of cancellation.

The public evidence cannot prove exactly how many trades or defaults would have changed.

A second counterfactual moves earlier still. Robust aggregation of exchange and OTC positions, lower or better-enforced accountability thresholds, and investigation of repeated excesses could have made concentrated short exposure visible before the crisis. Oliver Wyman found that the LME did not routinely investigate all accountability-level exceedances and that regulatory limits were too high to constrain the relevant positions. Better visibility might have triggered liquidity planning, member engagement, position reduction or tailored limits.

It cannot establish that a entity would have reduced the position or that geopolitical shock would not have overwhelmed the market.

A third counterfactual concerns the operating model. A formal handover from London to Hong Kong could have named the risk, specified price and margin triggers, required senior calls before any band suspension and assigned someone to monitor member communications. This is less speculative because it uses controls that the FCA identified as deficient. It would not require clairvoyance about the final price. It would require treating severe genuine trading as a possible disorder event rather than assuming bands existed only for mistaken orders.

A fourth concerns evidence. If every price-band setting, override, alert, email, call and decision had been retained in a single time-synchronised record, later investigators could have reconstructed the event without uncertainty. Better records would not themselves stop a squeeze, but they would improve real-time coordination and ex post legitimacy. They would also permit a more precise cancellation boundary if the exchange chose to remove trades only after a demonstrable control failure or disorder threshold.

The possibility of a narrower cancellation is therefore a conditional counterfactual, not an obvious answer. Selecting a later start time could have preserved more transactions, but it would also have left standing trades formed during a period the exchange considered disorderly. The High Court accepted that officials faced difficulty identifying a single precise point at which order was lost. A narrower remedy would be credible only if the venue had reliable control logs, stable price markers and a defensible threshold available at the time.

Current daily price limits provide a final counterfactual. They can stop a one-session price move before it reaches the scale seen on 8 March and spread adjustment over several days. They can also produce repeated limit sessions with little liquidity. The LME's multi-day framework attempts to handle that trade-off by creating a defined response after consecutive limit events. The accountability test is not whether a limit is always right. It is whether calibration, suspension conditions, collateral consequences and restart rules are known in advance and tested together.

Confirmed facts, supported inference and unknowns

The confirmed facts are extensive. Nickel prices rose extraordinarily on 7 and 8 March 2022. LME price bands were unavailable for more than three critical hours. The real-time operations team did not escalate effectively. LME Clear faced record and prospective margin demands. The LME suspended nickel trading at 08:15, cancelled trades from midnight and kept the market closed for eight days. The cancelled notional was approximately US$13.373 billion. Restart controls suffered implementation problems. The courts upheld the suspension and cancellation.

The FCA fined the LME for inadequate systems and controls, and the Bank required LME Clear remediation. The LME implemented new controls, and regulators later recognised material progress.

Several conclusions are supported inferences. Earlier use of bands and escalation probably would have slowed the move and may have reduced the required intervention, because the FCA explicitly linked disabled bands to faster price escalation. Better OTC visibility and position enforcement could have prompted earlier risk action, because official reviews identified those information and enforcement gaps. Clearer published rules and retained control logs should improve predictability and auditability. None of these inferences proves that the crisis would have been avoided.

Important facts remain unknown publicly. The complete entity-by-entity exposure, funding and hedge record is not available. The public record does not quantify net economic harm across all cancelled trades. It does not reveal every internal discussion, every legal view, every rejected alternative or every communication with affected members. It does not establish a private motive to favour a particular entity, and the courts rejected the improper-purpose case put before them. It does not show that the LME or LME Clear intentionally created the disorder.

The public record also does not provide all post-repair testing evidence. It does not include complete results from overnight simulations, parameter-change audits, member default exercises, extreme but plausible liquidity scenarios or independent validation of every OTC feed. Regulator statements support the conclusion that remediation occurred; they do not permit outsiders to reproduce every assurance test.

There are attribution limits around later performance. Higher 2025 volume demonstrates use, not causation. It cannot by itself show that the action plan restored confidence, because commodity conditions, fee changes, product demand and entity strategies also affect activity. The absence of another event of the same scale is not proof that controls would pass one. A durable claim must rest on control evidence, not only on a quiet period.

Finally, the motives and commercial position of individual market entities should not substitute for institutional analysis. A concentrated short can contribute to disorder without relieving the exchange of its monitoring duty. A claimant can stand to gain from extreme prices without losing the right to challenge administrative action. An exchange can protect its clearing system without proving that every prior control was sound. Accountability becomes clearer when each actor is assessed only for decisions and information within its practical control.

A durable accountability test for market-control power

The first test is visibility. Can the venue aggregate exchange and relevant OTC positions quickly enough to identify concentrated directional risk, common ownership, client dependencies and funding pressure? Does it know what information is missing, how stale it is and which legal barriers prevent collection? Reporting that cannot be reconciled across members is not visibility.

The second test is control integrity. Are volatility mechanisms calibrated for genuine disorder as well as erroneous orders? Are at least two independent controls active where appropriate? Can an operator disable them, and if so, is that action time-limited, logged and subject to mandatory senior approval? A control that disappears silently at the moment of greatest stress is not a control.

The third test is escalation. Does each regional shift receive a documented risk handover? Are price, liquidity, margin, missed-payment and communication triggers explicit? Is there one accountable decision leader across the venue and clearing house, with named deputies? Being reachable by telephone is not the same as having a tested escalation system.

The fourth test is clearing resilience. Can the central counterparty model intraday exposure, collect margin without destabilising members, test default waterfalls and distinguish temporary liquidity pressure from insolvency? Do exchange and clearing staff share scenarios before the market reaches a cliff? Protecting a clearing house after its calls become impossible is more costly than managing the path to those calls.

The fifth test is emergency proportionality. Before cancelling trades, can decision-makers state the rule, the trigger, the alternatives and the proposed time boundary? Can they explain why a halt, adjusted limit, delayed settlement, additional collateral or partial cancellation would not manage the risk? Urgency can justify action without consultation, as the courts held, but it should not erase disciplined comparison.

The sixth test is evidence retention. Does the operator preserve market data, control parameters, overrides, alerts, emails, calls, model versions, margin scenarios, meeting attendance and decisions on a common clock? Can an independent reviewer reconstruct what each responsible person knew at each stage? If historic settings have to be approximated, causal accountability remains incomplete.

The seventh test is public communication. Are notices prompt, factual and consistent about market status, affected trades, collateral treatment, restart conditions and known uncertainties? Does the operator tell regulators about control failures without waiting for the perfect reconstruction? Users can tolerate uncertainty better than unexplained reversals, but only if updates separate facts from provisional judgment.

The eighth test is user impact. Does the recovery plan account for physical-market hedgers, clients of clearing members and smaller firms that cannot maintain redundant access? Are brokers given controlled reconciliation and communication procedures? Does the venue measure replacement-hedge difficulty, spreads, depth and benchmark use, rather than treating aggregate volume as the only confidence metric?

The ninth test is repair assurance. Are action-plan milestones backed by independent testing, regulator review and evidence that controls remain embedded across staff turnover and regional shifts? Are adverse simulation results disclosed at least in aggregate? A policy document proves design intent. It does not prove that the control will operate at 04:49 during a fast market.

The tenth test is institutional learning. Do later rules change the incentives that produced the failure, or merely add more reporting? The FCA fine, Bank supervision, OTC reporting, daily limits, revised policies and governance work represent material change. Durable learning would be shown by integrated drills, reliable data, disciplined overrides, clear decisions and publication of performance evidence over time.

The LME nickel crisis should therefore be remembered neither as a simple rescue of short sellers nor as proof that exchanges must never reverse trades. It is a case in which an emergency power was legally upheld while the systems preceding its use were regulatorily deficient. The cancellation protected the clearing system and redistributed market outcomes. Both facts can stand.

The lasting standard is demanding but practical. A market operator does not have to prevent every geopolitical shock, squeeze or liquidity withdrawal. It does have to see the risks its rules require it to see, keep its controls active or account for every override, escalate across time zones, coordinate with its clearing house, preserve the decision record, disclose failures promptly and prove that repairs survive realistic stress. When an institution can cancel billions of dollars of trades, accountability begins long before the cancellation notice and continues long after the market reopens.