Risk Management: Avoiding the Next “LME Nickel” Market Incident
In early March, a Chinese investor with a short position on nickel through derivatives suffered losses when the cost of nickel began to increase substantially. Along with other commodities, nickel prices have soared since the beginning of Russia’s invasion into Ukraine.
According to reports, the Chinese nickel investor suffered $8 billion of losses and went into default, and other investors trading on the same platform – including those with long positions on nickel – also suffered losses or were denied gains when the exchange operating the nickel market canceled $3.9 billion in nickel trades.
The reader might wonder why those investors on the winning side of the nickel trade still suffered losses or were denied gains – those investors with winning positions are wondering the same. Apart from the specifics of any particular risk-management failure, the answer relates to how derivatives exchanges and clearinghouses typically operate and manage risk. Another part of the explanation is that traditional infrastructure for derivatives trading relies extensively on credit extended to market participants.
Occasional market shocks are expected by participants in any market, but when they cause substantial losses that can reverberate throughout the system, all stakeholders return to the question of how those shocks can be more effectively contained. Those involved in the nickel market are asking these questions right now.
What do digital-asset exchanges have to do with this? It turns out, potentially a great deal, because exchanges like FTX use a risk-management system that could stem the type of losses and denied gains seen in the LME debacle.
Like many other asset classes, digital-asset markets do indeed experience shocks – in March 2020 the price of Bitcoin, for example, declined in value by roughly 50 percent over two days, and again declined by roughly 53 percent in May 2021.
But digital-asset platforms like those operated by FTX have been designed and built with risk-reducing features to better address these market shocks and volatility in markets more generally. To be sure, during those price declines for Bitcoin in 2020 and 2021, some market participants suffered losses, but the FTX platform avoided any retroactive cancellation or busting of positions of those with market gains, and trading in the market continued in an orderly manner.
The risk-reducing features of digital-asset platforms like FTX’s represent a more conservative, but more efficient approach to risk management.
For example, FTX has an application before the U.S. CFTC that would introduce a real-time risk model that would better contain market shocks and resulting losses and prevent broader contagion in a market. While the risk model is designed for digital-asset derivatives trading, the principles of the risk model have broader applicability and could be considered for other markets as well.
Key Features of FTX Risk Management System and Default Waterfall
What are the features that make the FTX risk-management system safer? First, the FTX risk-management system assesses risk on a nearly real-time basis, assessing customers’ trading positions every few seconds to determine whether a customer has adequate resources or collateral in their account. This risk-exposure time period is substantially shorter than what is typically seen on other derivatives exchanges in traditional finance, ensuring on a more frequent basis that adequate collateral is on hand, rather than waiting longer for risk in the portfolio to potentially increase. This contrasts with what happened in LME nickel markets, where risk was only monitored on a daily basis.
Second, the system also requires that customers transfer the required collateral to support their trading to the FTX platform before they can begin trading. The amount of collateral required is based on a proven risk methodology that would cover at least 99 percent of the one-day portfolio returns using appropriate weightings for base VaR and stress VaR. To account for stress scenarios for a particular asset, the model looks at both historical as well as hypothetical scenarios to appropriately calibrate necessary resources. Notwithstanding the shorter risk-exposure time period the FTX system relies on, for its CFTC risk model FTX relies on a time period of 24 hours to calculate collateral requirements based on regulatory requirements, building in an added buffer to the margin calculation. On traditional derivatives exchanges–including LME–collateral is instead generally based on credit, exposing all market participants if that credit decision turns out to be unwarranted.
Third, the risk system has an auto-derisking feature to prevent a build up of risk in a customer portfolio. If a customer begins to suffer trading losses and their collateral balance declines toward minimum margin requirements, an automatic derisking process uses rate-limited, marketable limit orders to reduce risk as the customer account value falls below the maintenance margin level. As a result, customers are incentivized to manage their account collateral and proactively add collateral or reduce risk positions prior to partial auto-derisking. Notably, unlike traditional platforms, the FTX risk system does not extend calls for additional margin or extend credit to the customer until such a call can be met – the system is based on a presumption that FTX will not have recourse against any customer for credit losses. On LME it took days to begin attempting to derisk the large position, by which time it was substantially more underwater than it was initially.
Fourth, the FTX risk-management system relies on backstop liquidity providers (BLPs) to take on the portfolio of a participant in default. To wit, if a customer’s account value continues to decline further to a determined margin threshold, then the system declares a default and the risk position is moved automatically to the contractually bound BLPs. Depending on the speed and magnitude of price changes, the defaulting risk position could have positive or negative account value when it is passed to the BLPs. Firms volunteer to be BLPs–no one is forced to–but when a firm does become a BLP, they are automatically passed liquidating positions in real time and are unable to reject it, promising to provide liquidity when it is most important.
Finally, after BLPs assume and manage the risk positions of participants in default, and if there remain accounts with negative value, the FTX guaranty fund will absorb those remaining negative values. For a full explanation of how FTX sized its guaranty fund for purposes of its CFTC application, please see https://www.ftxpolicy.com/ftx-guaranty-fund.
All other things remaining equal, this type of system is a more conservative approach to managing risk. So long as the collateral required by the system’s risk model is adequate, and so long as the platform deploying the risk system is otherwise operated in a resilient manner, this type of system will better prevent massive losses by a customer that could have implications for the broader market by requiring collateral to be posted to the clearinghouse, and by acting promptly in the case of large market moves.
Addressing Pro-Cyclicality in Markets During Times of Stress
A common question that arises in discussions about the FTX risk-management system is whether the auto-derisking feature could promote “pro-cyclicality” in the market, particularly during a time of market stress. Procyclicality refers to the tendency of financial variables to fluctuate around a trend during the economic cycle, or a state where the behavior of an asset price moves in tandem with the cyclical condition of the market. To put it into context, when asset prices are declining, the question becomes whether the FTX auto-derisking feature increases downward price pressure on those assets as the risk engine places market sell orders into the FTX order book.
To be sure, any increase in market sell orders will have an impact on prices, but the FTX risk-management system has a number of other features to address procyclicality on the platform by preventing unnecessary additional impact. FTX believes that combined with these features, the FTX risk system offers a superior program for risk management compared to other offerings from traditional infrastructure.
First, the FTX risk program includes regulatory requirements for the margin model that address risk during periods of stress. These include a liquidity charge for all positions in a customer portfolio, reflected as a transaction cost, during periods of low liquidity; and a concentration charge when a participant’s position in a particular contract is large enough relative to trading volumes for that contract on the FTX and other global platforms. These features dis-incentivizes the accumulation of positions that could lead to excessive asset-price trends; and if larger positions are put on, FTX requires that large amounts of collateral are deposited to the clearinghouse beforehand to help protect the system. Second, the FTX risk program imposes a stress loss limit that limits the aggregate risk that a participant can assume across positions in their portfolio, another control that protects the guaranty fund from extreme residual loss.
Additionally, the FTX trading platform sets slowly moving price bands for certain contracts, where the exchange will not accept orders that are set outside the minimum and the maximum of the price range for that particular contract. These price bands have the effect of mitigating the impact of erroneous orders, momentary illiquidity, or large concentrated buying or selling of contracts that could momentarily exhaust the orderbook. They also act as a temporary circuit breaker, preventing markets from being able to quickly decline or increase more than a certain amount while creating time for algorithms to be inspected and liquidity to refresh.
Finally, FTX’s risk engine limits the rate at which it closes customer positions to be within a small fraction of global volume. While this will not entirely eliminate the price impact of liquidations, it will ensure that the liquidations are much slower than the rate at which liquidity can be transported to the orderbook by sophisticated market participants, mitigating the risk of inefficient short-term price impact.
Together these market and risk controls work to stem procyclical trends in the FTX order book, including trends influenced by the auto-derisking feature of the FTX risk engine. With appropriate calibration of each of these controls, the FTX risk-management system promotes risk-reducing platform operations that also limit systemic risks throughout the market ecosystem.
It is worth noting here that the absence of the auto-derisking feature would have a pro-cyclical impact on markets that would only manifest in a different manner. Without auto-derisking, there would be a call for additional collateral from a customer whose position suffered enough losses to require it. During a period of market stress and declining asset prices, market participants operating under this model would be under pressure to find liquid resources to make a margin call at a time when liquidity becomes more scarce*. To conclude, in times of market stress pro-cyclicality always will be a risk to address and manage, but FTX believes its risk system does so most effectively and appropriately.
Empirical Results
The FTX risk model has been running internationally for three years, including during days in which markets moved 40%, using a model substantially less conservative than what FTX US Derivatives has proposed.
FTX International’s insurance fund has drawn less than $10m lifetime, with a maximum daily draw of less than $5m–on the day that markets were down 40%. Had FTX been using the anticipated parameters for FTX US Derivatives, the largest daily insurance fund draw would have been less than $1m. FTX has also processed as much as $1.6b of liquidations in a single day, while markets remained orderly and liquid.
For more information about the historical performance of FTX’s risk engine and insurance fund, see here.
Conclusion
The recent LME nickel incident happened because excessive risk was allowed to build up in the portfolio of an investor on the LME platform. The specifics of how risk controls were calibrated on the platform presumably had an impact on that outcome. The design of the risk-management program, however, is also important. Unlike the typical risk-management program, the FTX platform is a real-time risk system that relies on pre-funded collateral and the automatic de-leveraging of customer positions to prevent excessive risk from accumulating. FTX believes that this type of innovative system can effectively prevent the build up of excessive risk.
*See Basel Committee on Banking Supervision Committee on Payments and Market Infrastructures, Board of the International Organization of Securities Commissions, Consultative report – Review of margin practices, October 2021, for a further discussion of the concerns
