Marcus Storr, Head of Alternative Investments, FERI AG
Since 2007, offshore hedge fund assets have roughly doubled to more than USD 4 trillion (from USD 2 trillion), and they now manage eight times more than UCITS hedge funds, which have quadrupled to more than USD 500 billion (from USD 129 billion) over the same period.
In more recent years, multi-PM funds have been driving a disproportionate share of asset growth. Since 2018, overall hedge fund industry assets have grown by 5.6% per year, while multi-PM assets have grown by 16% per year.
This has come mainly from giant managers. The largest four multi-PM managers – Millennium Management, Citadel, Brevan Howard and Point72 – make up half of multi-PM assets. The top ten, adding Balyasny, ExodusPoint, Schonfeld, Verition, Walleye and Eisler, make up 70% of assets in the space.
However, over the past year, multi-PM assets have dropped by about 10% from USD 369 million to USD 366 billion, which implies somewhat larger gross outflows of assets despite their continuing positive investment performance.
The first half of 2024 saw net outflows of USD 31 billion, two thirds of which, FERI estimates, were investor-led redemptions. Some of these funds also periodically make compulsory redemptions and return capital to investors when they approach or hit capacity limits.
Some of this asset growth has come from the compounding effect of strong performance year after year, and not only from net inflows from investors. The ‘Magnificent Ten’ multi-PM firms have outperformed other multi-PM managers and hedge fund managers in general, almost as impressively as the Mag7 mega cap tech names have outpaced US equities. The top ten pod shops have achieved annualized returns of 13%, net of fees.
Risk-adjusted returns have been extraordinarily consistent. Since June 2019, an index of the top 10 multi-strategy funds has generated a Sharpe ratio approaching 3, more than triple that of the HFRX index, which has also had a worst drawdown four times as large. The Magnificent Ten have captured 70% of world equity market returns with less than 20% of the volatility.
Of course, any index of multi-PM managers has lower volatility than the individual managers due to diversification benefits amongst them. It nonetheless produces an almost suspiciously straight line of returns, which may remind some investors of Madoff, but this time the numbers are real!
The returns are magnified by substantial leverage. The top ten firms use an average leverage of 8.5x, significantly more than in smaller multi-manager firms, and consequently manage gross assets of USD 2.2 trillion. They make up just under 10% of hedge fund industry net assets but are now a much larger proportion of gross assets since most other managers are less leveraged. This scale of assets and share of the hedge fund market may raise concerns about potentially increased market impact and transaction costs when trading. There could also be concerns about the costs of leverage since interest rates have normalized and are expected to stay above zero for many years.
The Magnificent Ten all have a pass-through cost model, charging rents, salaries and other operating costs to investors. Median pass-through costs have reached 5% per year. In addition, pod team bonuses are passed through to investors without any netting amongst teams, which results in an effective performance fee of 15-30%, which is higher than the headline performance fee levels for most hedge funds.
These firms are some of the largest employers in the hedge fund industry: Millennium has over 5,000 staff, while Citadel and Point72 each have nearly 3,000. Balyasny is close to 2,000 and Brevan Howard around 1,000. Each firm could have 150-200 portfolio managers all trading independently of the others in pods, and the firms may be trading as many as 15 or 16 or more strategies.
The war for talent is not only between pod shops. They are also competing with tech giants such as Apple and Google and others in Silicon Valley. First year pay packages are reportedly exceeding USD 20 million for some individuals and even interns can earn as much as USD 25,000 per month.
Theoretically, if two firms deliver the same gross investment returns, the one with a pass-through fee structure would deliver lower net returns to investors since costs and fees consume a higher share of gross performance. In practice, as mentioned, firms with pass-through fee structures have delivered much higher net returns, which means that their gross returns have been even further ahead of the rest of the industry.
Most hedge funds offer monthly or lower levels of liquidity, but the multi-PM shops are locking up investors for increasingly longer periods.
Though the pod shops reportedly trade liquid strategies, only 24% of these funds could now be exited within a year, 48% would take over one year, 18% over two years and 10% now require over three years to fully redeem capital. Investor-level gates are one way to slow down redemptions. They typically restrict redemptions to 25% per investor per quarter and in contrast to the ad hoc freezes imposed in the GFC, this is advertised in advance.
One perception is that they may now need longer notice periods because strict risk limits make them the first sellers in a correction or crisis, and larger assets mean they are selling a higher proportion of daily or weekly or monthly liquidity. Therefore, they might need to slow down the pace of redemptions and portfolio liquidations because in a fire sale situation such as the GFC they would fetch lower prices. And any crowding into the same positions increases the potential market impact of a synchronized sale.
Though the multi-PM shops seem riskier in terms of leverage, they may be less risky on other risk measures. Pod shops are subject to very strict risk limits on equity market beta, style and factor exposures, as well as position sizes. These risk controls are all strictly enforced: portfolio managers and entire teams at the multi-PM firms can often be immediately fired for breaching risk or loss limits. Traditional discretionary hedge fund managers run much more concentrated portfolios with larger individual positions and wider risk tolerance.
The returns are not only independent of equity and bond beta but are also independent of other well-known styles and factors. The return profile appears to be genuine, pure and idiosyncratic alpha. Even FERI’s sophisticated non-linear factor decomposition cannot explain the returns.
It is undeniable that the net returns to investors are excellent even after the high fees, and the idiosyncratic nature of the return pattern provides a strong diversification benefit for portfolios. Time will tell if this level of alpha generation proves to be sustainable with much larger amounts of assets, higher costs of leverage, and higher staff costs.
FERI scours the globe for hedge fund talent and holds no bias to large US multi-strategy managers. Managers presenting at previous FERI Hedge Fund Days have come from ten cities in Europe, nine in North America, three in South America, four in Asia, and one in each of the Middle East and South Africa.
Given the macroeconomic backdrop of interest rates and inflation, FERI expects that long/short equity will be a valuable part of portfolios. Equity market valuations are historically high, particularly in technology, and hedge fund managers have often been short of more expensive parts of the equity market. Market neutral and often sector specialist stock-picking strategies form the core of many multi-strategy, multi-PM funds, but these alpha streams can also be accessed in other hedge funds.
Given the macroeconomic backdrop of interest rates and inflation, FERI expects that long/short equity will be a valuable part of portfolios.
Marcus Storr, Head of Alternative Investments, FERI AG
Tom Morris, Managing Director, Co-Head of Research, Systematic Active Equities, BlackRock Inc., London
BlackRock Systematic has 230 staff and manages around USD 292 billion in London, New York and San Francisco (as at end September 2024). Within equity hedge fund strategies, Tom Morris and Kevin Franklin co-manage the Global Alpha Opportunities fund which has a strategy AUM of USD 1.3bn. More broadly the BlackRock Systematic Active Equity team manages around USD 8 billion in hedge fund strategies. They also benefit from the resources of the world’s largest asset manager running around USD 12 trillion.
AI and machine learning are increasingly used to optimise selection and timing decisions for longs and shorts using vast amounts of data in a very scalable way, for circa 15,000 global equities, and circa 3,000 credits, in 45 developed and emerging markets.
Daily data volumes on the internet have grown from one to 180 Zettabytes (a Zettabyte is one sextillion bytes, or one trillion gigabytes). Computers can digest this: computer power used to double every two years but now does so every six months, as distributed computing and high performance TPU (Tensor Processing Unit) and GPU (Graphics Processing Unit) computer architecture can analyse vastly more data than older CPU (Central Processing Unit) systems. Meanwhile the run time for BlackRock’s daily five-year simulation has improved from three hours (10,800 seconds) 20 years ago to 1.8 seconds – and continues to get faster. This enables faster innovation and the development of more powerful alpha models with increasing amounts of data as input.
BlackRock’s systematic equity strategies have roughly quintupled their number of alpha signals from 200 to 1,000 over the past decade. Typically, at least 50 signals are added per year.
BlackRock’s mantra is that everything that can be measured, will be measured. This includes broker reports, company filings, blogs and news and conference call sentiment, all of which can be read by NLP (natural language programming) machines in a scalable way. The physical field trips and channel checks of former times can now be done much faster and more extensively through satellites, GPS, geolocation and streaming data feeds measuring transactions, internet search traffic, retail footfall and EV charging density. “Nowcasting” also estimates economic statistics in real time before they are announced.
Large, messy, big data is unstructured and was once unwieldy but is now being turned into structured data using modern AI techniques to extract actionable insights. A proprietary knowledge graph for a single company could hold 1 billion facts in 300 million nodes mapping over 400 sorts of relationships and can be centralised for access by all researchers. Data dimensionality is growing with measures of consumers, product development, company trends, regional variations and macro informing forecasting. Market regimes can also be identified with better accuracy.
BlackRock provided three case studies to illustrate the growing power of their analytics.
AI can extract company sentiment from various market participants, sell side broker reports, retail investors, institutional investors, hedge funds and news reports. Focusing purely on one of these – broker sentiment – BlackRock has an archive of 27 million reports and ingests 6,500 new reports per day from over 2,000 brokers in multiple languages. This could not be easily read, distilled and processed by humans.
Over the years since 2012 the definition and measurement of sentiment has evolved and become much more sophisticated. It started with a binary split between positive and negative sentiment based simply on counting numbers of positive and negative words. The second iteration applied higher and lower weightings to certain words based on their predictive power for equity returns. The third and fourth versions considered context and timing around, before and after earnings. The fifth version superseded human-curated dictionaries with machine learning for the entire vocabulary.
Now, large language models (LLMs) as used in ChatGPT transformer architecture are being applied to process the information more holistically, considering each word and sentence in the context of the whole report and further improving the signal quality. The models can also identify conditionality, negation and nuance: ‘slowdown’ may sound negative but could be positive if it is referring to a market environment in which the business actually outperformed. BlackRock Systematic’s proprietary LLMs are smaller than large expensive commercial LLMs but have performed better because they home in on the relevant parameters and information. And the accuracy of their return prediction continues to improve.
BlackRock are also cognizant of potential for game theoretic elements and gaming behaviour where companies contrive their earnings releases language to trigger positive algorithms.
AI can also be used to navigate broader market themes. A thematic robot thinks through economic implications and business models and picks baskets that are winners and losers from a human curated theme or scenario. For instance, GLP1 weight loss drugs that suppress appetite and induce weight loss were approved by the US FDA in 2023. This is clearly positive for their manufacturers (such as Novo Nordisk and Eli Lilly) and a second order beneficiary group could be makers and sellers of healthy food. The drugs may however also be negative for firms involved in bariatric surgery as well as some making weight loss supplements, fast food, obesity treatments and certain medical devices, which could be short candidates. There can also be more subtle and less obvious second, third and subsequent order effects.
AI can also be used for so-called scenario driven recommendations (SDR) to generate top-down stock selection. LLMs technology can replicate the thought process of top-down fundamental investors, in an automated streaming fashion. The starting point is sell side macro and strategy reports, which can be fed into the LLMs and used to distil actionable summaries of potential future market scenarios. The LLMs work out which sorts of business models and companies could benefit and make actionable recommendations for specific stocks.
Lower container freight rates in late 2023 could result in empty ships, lower capacity utilisation and use of alternative routes, which might lead to negative impact on some global shipping companies. A European tourism growth scenario could most obviously lead to a long in some low-cost airline.
In one day, BlackRock ingested 364 sell side strategy reports, cleansing superfluous text such as disclaimers and focusing on the important market scenarios. Each SDR output identifies 5-6 scenarios and multiple stocks for each scenario. Over 10 years the SDR easily generated 300,000 distinct scenarios and 400,000 distinct buy and sell stock recommendations.
An interrogator tool drills down by scenario, company or keyword. Plugging in ‘nuclear’, unexpected delays in the restart of Japan’s nuclear fleet could drive up LNG prices, and lead to a short in a utility using LNG and a long in a US LNG exporter. Thus, the model has been trained to think laterally across sectors and geographies to generate long and short ideas.
Back tests for these models have demonstrated high Sharpe ratios approaching two and recently the out-of-sample, real live performance has shown an extraordinary Sharpe ratio of five. BlackRock do expect some normalisation of the alpha.
Nonetheless, 2024 is proving to be the best ever year for a quantitative equity market neutral fund that has seen an upward inflexion of performance since it pivoted to AI in 2019.
Importantly, only 6% of the alpha can be explained by factors, meaning that 94% of it is genuinely differentiated.
Large, messy, big data is unstructured and was once unwieldy but is now being turned into structured data using modern AI techniques to extract actionable insights.
Tom Morris, Managing Director, Co-Head of Research, Systematic Active Equities, BlackRock Inc.
Cecile Herzl-Greuter, Portfolio Manager, and Dr. Robert Wiesner, Head of Asset Management, Bank für Tirol und Vorarlberg, Innsbruck, Austria
Innsbruck-based BTV is a conservative asset manager running EUR 2.8 billion in equities, bonds, gold, commodities and multi-asset class strategies. The client base includes high net worth individuals, entrepreneurs, institutions, foundations and churches.
BTV has partnered with FERI to create an award-winning fund of UCITS funds. Since FERI manages data analysis, manager selection, due diligence and reporting, BTV can now spend more time on product development, customer support and communication. The fund of funds was tailor-made for BTV clients after BTV first allocated to FERI’s Flex strategies and used FERI’s economic research 10 years ago. FERI has given BTV access to some of the best hedge fund managers globally. Fund selection considers managers’ capacity and clients’ tax efficiency.
US equity markets are at record highs, driven heavily by the MAG 5, and since these mega cap tech stocks are making billions of profits and cashflows this rally is very different from the speculative bubbles seen in the German Neuer Markt. Investors are nonetheless looking for diversification. BTV are not market timers: they will not try to predict when the rally ends and are neutrally positioned on equities. It is very difficult to predict the timing of downturns, and some cautious investors have already exited too early. BTV would rather maintain a diversified portfolio than try to continually trade the market. BTV are aware of economic, recession, inflation, political and geopolitical risks for equities but would not take a specific view on these risk factors and nor were they positioned for Trump or Harris trades. The guiding star remains well diversified.
Cash can be seen as defensive and increased interest rates mean that some investors might feel happy sitting in fixed term cash deposits or overnight money market funds earning 3.5% in the short term. However, BTV points out that interest rates have already started coming down, and historically cash has not beaten inflation. Therefore, BTV do not normally keep more than 10% in cash.
Most BTV clients want up to 20% in alternatives, which could include gold, commodities and liquid hedge fund strategies. As with overall asset allocation, within the alternatives bucket BTV takes a well-diversified approach and would not concentrate into any one of these.
Within alternatives, the liquid alternative fund of UCITS hedge funds is one useful portfolio building block, providing stable returns with low volatility, low correlation and defensive diversification, while maintaining portfolio liquidity.
The BTV fund of funds has five strategy categories: equity hedge, event driven, relative value, tactical trading, and some defensive protection. Each one is expected to perform better in different market phases and the defensive part is designed to provide protection in a crisis. As of September 2024, the product was underweight of the protection part. This could be increased in the future but is unlikely to be. Incidentally, FERI and BTV are closely monitoring some AI driven strategies but have not yet allocated to anything that is 100% AI driven.
The product targets returns of 4-6% per year with low volatility but is not currently exposed to cryptocurrencies. Bitcoin is an example of a liquid alternative asset that may become a store of value and a standard part of portfolios. BTV are not opposed to it and recognize that Bitcoin and gold both have limited supply: gold output increases are running at about 2% a year, which is below GDP growth. The bank is conscious that investors can be procyclical: demand for crypto was high before the last crypto winter. There is also a danger of investors having too much confidence in the diversification benefits of digital currencies. Most importantly, cryptocurrencies are very volatile, which is the opposite of BTV’s objective with the liquid fund of funds.
Liquid alternatives may seem fundamentally complex, but their basic idea is simple: to provide diversifying returns in a liquid format. On the less liquid side, BTV does research private markets, including infrastructure and green infrastructure, but prefers to stay more liquid now so that investors can withdraw at any time. BTV do not want to face longer notice periods and withdrawal periods as seen in the multi-strategy pod shops and are generally looking for weekly dealing on funds.
BTV is democratizing access to liquid alternatives and reducing barriers to entry. BTV also insists on full transparency and close contact with portfolio managers. BTV carefully explain strategies such as arbitrage and insurance linked securities to clients and make them accessible. Of course, these strategies do not appeal to all customers. BTV quote a German aphorism: ‘The worm has to taste good to the fish not the fisherman’, and the portfolio must suit the customer.
BTV recognizes that any time series can be too short to reach conclusions about correlation because patterns of correlation can break down or increase close to one in a crisis situation. BTV are prepared for the risk of a resurgence of inflation and want to stay uncorrelated, even in times of crisis.
Amongst other strategies, BTV expects that some discretionary market neutral strategies can work very well in difficult financial markets.
The bank is conscious that investors can be procyclical: demand for crypto was high before the last crypto winter.
Cecile Herzl-Greuter, Portfolio Manager
A profile of FERI and BTV’s fund of funds product, which received The Hedge Fund Journal’s 2024 UCITS Hedge award can be accessed below:
Tom Tang, Chief Executive Officer, WaterValley Capital Management (HK) Limited, Hong Kong
More than thirty years after China’s commodity markets first opened, a geographic arbitrage strategy is seeing one of its best market environments and delivering double digit return with single digit volatility in 2023.
The team synthesizes 18 years of experience in high frequency trading, market making and short-term arbitrage strategies on Asian exchanges – in proprietary trading and at Millennium Management. The CIO and CTO have worked together since 2005: at Societe Generale, they arbitraged one tenth of the South Korean KOSPI 200 options market when this was the world’s most active equity derivative market, and at Millennium Management they traded both option and equity arbitrage strategies.
The cross-border commodity arbitrage which was first implemented by Credence funds at Splendor Capital Management in 2008 and continued at WaterValley, was launched in 2013. Mr. Tang, the co-founder of Splendor Capital Management, joined the WaterValley team in 2017.
The techniques, technology and systems used for arbitraging equity derivatives have also been applied to commodity markets. The in-house technical systems, programming languages C++ or Python, were originally designed for trading and risk controls in options markets and can now easily manage commodities. They were designed to capture the smallest spreads of 0.10% in low latency cross border arbitrage with tight risk controls.
Since trading began in 1991, China’s commodity markets have grown to a size matching the US and some of them are even deeper. Liquidity in Chinese commodities has tripled in recent years, helped by retail investors, even as volumes traded in Chinese equities went down. There are more than 70 commodity contracts but the uniquely Chinese ones such as apples, eggs and PVC are not of interest to geographic arbitrageurs, who instead focus on crude oil, gold, silver, copper, aluminium, iron ore and soybean, where there is plenty of liquidity in both the Chinese and the Western markets.
For instance, the Shanghai gold future could be traded against CME gold and the Golden Week proved true to its name in 2023. Around China’s October 1st, 2023, seven-day Golden Week, including the Mid-Autumn Festival, the arbitrage between gold inside and outside China spiked to record levels of 5-6%, as the RMB currency was dropping fast. This was almost double the previous record spread, and well above the typical 2% range. Amid China’s Central Bank restriction on gold import, individuals in China noticed the record spread, travelled to Hong Kong to buy physical gold and profited from the arbitrage. Once central banks ended the restriction, the spread came back to a normal range and WaterValley made about 5% in September and October 2023.
China makes up to 60% of demand for copper against only 8% for the US and in 2024 prices in the two markets diverged by even more than the record gold arbitrage, though on this occasion the Chinese market was at a discount and the Western one at a premium. On 15 May 2024, CME Group listed copper in New York for July delivery was suddenly priced 10% higher than the contracts in Shanghai and London – nearly double the previous peak price discrepancy of 5-6%, which had occurred around the Russian invasion of Ukraine in 2022. The spread widening was caused by some logistical and contract specification complications: there were some delays moving copper into warehouses, CME Group did not accept China deliveries, and it took time to transfer the position to one acceptable by CME Group, as August CME inventory tripled. WaterValley was able to avoid a big drawdown from this unusual dislocation and recovered fully after rolling the contracts.
The strategy is subject to a range of risk limits. Gross leverage is capped at 800% but maximum delta or net directional exposure is 10%. Maximum exposure per commodity is set at 100% for agriculture, 200% for energy and 300% for metals. There are also stop losses at individual commodity level and at fund level. Moreover, there is a discretionary override for situations that quantitative caps cannot manage, such as the GFC, wars, conflicts, famines etc.
The strategy has an average annualized return of 11% at a Sharpe ratio around 2. The past two years have posted much higher or close to double the average return. 2023 was the best year and 2020 the worst, though it still made 1.6%. The strategy has also shown very low correlations, having recently not been impacted by volatility in the US equity market nor the Libyan oil panic nor the August 2024 mini-correction.
The strategy has limited scalability. Capacity is estimated at USD 500 million of net assets in the fund, though given the leverage, this works out at much more in terms of gross assets. This is partly because China has strict position limits in commodity markets. Larger players may prefer to use their quotas for directional strategies.
Additionally, margin requirements are high as cross-margining is not possible between the two legs of the trades. Therefore, margin needs to be managed carefully, since profits on one leg of the trade might need to be swept back to cover any variation margin on the other side of the trade.
There can also be ‘Black Swan’ risks blowing out geographic spreads, such as the nickel short squeeze in March 2022 when the largest player tried to squeeze the nickel market, and LME had to cancel trades. This could have been a very dangerous situation for cross-border arbitrage. WaterValley have seen such episodes many times before inside China in small markets and use their knowledge to avoid pitfalls and navigate the markets.
China makes up to 60% of demand for copper against only 8% for the US and in 2024 prices in the two markets diverged by even more than the record gold arbitrage.
Tom Tang, Chief Executive Officer, WaterValley Capital Management (HK) Limited