 | |  | |
STONE MOUNTAIN CAPITAL RESEARCH PERSPECTIVE VOL. 17
The hedge fund industry is experiencing a dramatic change in the way it is operating and the current trend is shifting towards funds following systematic trading strategies, which constitute 10% of the overall hedge fund universe. Systematic strategies are off to a solid start in 2016, while other markets like equities and credit are still struggling in a low-yield and risky macro- and socio-economic environment. Black-box trading assets grow steadily every year since the financial crisis from $163.5 bn in 2008 to $314.9 bn in 2015, with the most recent activity in the industry indicating interest for algos and quant models, which will identify patterns in 'big data' sets with the help of 'artificial intelligence' and 'machine learning' techniques.
The hedge fund industry is experiencing a dramatic change in the way it is operating and the current trend is shifting towards funds following systematic trading strategies, which constitute 10% of the overall hedge fund universe. Systematic strategies are off to a solid start in 2016, while other markets like equities and credit are still struggling in a low-yield and risky macro- and socio-economic environment. Black-box trading assets grow steadily every year since the financial crisis from $163.5 bn in 2008 to $314.9 bn in 2015, with the most recent activity in the industry indicating interest for algos and quant models, which will identify patterns in 'big data' sets with the help of 'artificial intelligence' and 'machine learning' techniques.
Exhibit 1. Assets of Systematic Traders since 2008. BarclaysHedge.
The branch of artificial intelligence that hedge funds seem to adopt is machine learning, which combines complex statistical and mathematical models with numerous datasets ('big data') to identify profitable opportunities on the financial markets. Obvious challenges arise from the methodology applied and need to be clear when addressed to investors. First of all, the managers / developers need to assure the reliability of data sources they feed into the algorithm and avoid data dredging. The algorithm is aiming to uncover persistent return or price patterns by utilizing statistical factors which may give rise to spurious correlations in the dataset, which challenges the actual persistence of patterns. The developers need to integrate a synthetic metric using robust correlations and bumpiness coefficients to reduce the number of spurious correlations.
Black-box trading is relatively static, especially for trend-following and momentum strategies, which could not fit well in a dynamic financial environment, that can be viewed as a multi-factor non-linear equation. The major factor that quants fail to capture in a robust manner is the human behavior and this could lead to subsequent problems in the markets. An interesting case would be a stress signal from a model, which will stop trading high volumes and automatically reduce the liquidity leading prices to crash.
Systematic strategies use trading rules, which derive from backtesting the model over a specific period of time depending on the frequency of the trades the strategy implements. These models should account for slippage and transaction costs to depict robustly the backtesting performance of the strategy. Computer-driven processes do not guarantee success considering the bright examples of the 'Flash Crash' in 2010 or the 'Knight Capital Group case' two years later. The Flash Crash highlights an interesting aspect of systematic trading, which is spoofing (placing a bid or offer on a stock with the intent to cancel before the execution of the trade) and layering (placing multiple orders without the intention to execute the trade) and investors should be aware of these practices as those orders can signal false buying or selling pressure.
Systematic traders should be able to provide better downside protection by imposing strict trading rules on their system. This is evidenced from Exhibit 2 noticing the lower drawdown during 2005-2015 period and the platykurtic distribution of returns over that period. Algorithms are designed to reduce fluctuations and avoid “black swan” events by executing quickly trades within a pre-determined range of prices. Systematic CTAs exhibited increased volatility compared to discretionary strategies combined with lower returns, indicating the failure of the algorithm to exploit volatility for profit gaining.
Black-box trading is relatively static, especially for trend-following and momentum strategies, which could not fit well in a dynamic financial environment, that can be viewed as a multi-factor non-linear equation. The major factor that quants fail to capture in a robust manner is the human behavior and this could lead to subsequent problems in the markets. An interesting case would be a stress signal from a model, which will stop trading high volumes and automatically reduce the liquidity leading prices to crash.
Systematic strategies use trading rules, which derive from backtesting the model over a specific period of time depending on the frequency of the trades the strategy implements. These models should account for slippage and transaction costs to depict robustly the backtesting performance of the strategy. Computer-driven processes do not guarantee success considering the bright examples of the 'Flash Crash' in 2010 or the 'Knight Capital Group case' two years later. The Flash Crash highlights an interesting aspect of systematic trading, which is spoofing (placing a bid or offer on a stock with the intent to cancel before the execution of the trade) and layering (placing multiple orders without the intention to execute the trade) and investors should be aware of these practices as those orders can signal false buying or selling pressure.
Systematic traders should be able to provide better downside protection by imposing strict trading rules on their system. This is evidenced from Exhibit 2 noticing the lower drawdown during 2005-2015 period and the platykurtic distribution of returns over that period. Algorithms are designed to reduce fluctuations and avoid “black swan” events by executing quickly trades within a pre-determined range of prices. Systematic CTAs exhibited increased volatility compared to discretionary strategies combined with lower returns, indicating the failure of the algorithm to exploit volatility for profit gaining.
Exhibit 2. Macro Strategies Summary Statistics from 2005 to 2015. Stone Mountain Capital Research, 2016.
Future belongs to big data and machine learning funds especially when they address the above mentioned challenges. Hedge funds are moving towards this direction by hiring data scientists and computer engineers from Tier 1 technological corporates such as Google and Microsoft. Artificial intelligence is coming to stay in the financial markets and is going to reshape the investment space.
From a practical perspective there are plenty different quant and algorythmic trading strategies in the hedge fund sector available: 'systematic broadly diversified CTA' across markets like equities, fixed income, currencies and commodities, metals, indices etc. trading in futures vs. 'sector specific systematic CTA' strategies trading only in equities, bonds, etc. Other strategies include 'trend followers' (based on rigorous research any trend in datasets is traded long or short) vs. 'contrarian' (the system holds until a trend has formed and after a specified period of time trades automatically in the other direction), 'systematic' vs. 'discretionary' (trades are executed by the traders not the system), 'discretionary global macro' trading based macro-economic trends in futures and exchange traded options broadly diversified in equities, fixed income, currencies, commodities, indices etc. vs. global macro sector specific like e.g. fixed income focused vs. global macro focused predominantly on a country like the U.S. or region like Europe and mixtures of discretionary and systematic strategies in which e.g. the trade selection is sourced from a system and the execution is done from traders or for certain asset classes the system trades automatically and for others trading is discretionary. Further differentiating factors can be long only vs. long/short vs. market neutral strategies. Recent developments in the drive for lower fees for investors differentiate quant strategies into actively managed hedge funds and lower fee 'smart beta' strategies. The fund of hedge fund (FoHF) market is mirroring the single manager funds. FoHFs with purely systematic CTAs, diversified FoHFs with broad allocations to equity hedge, credit hedge and CTAs in order to generate 'uncorrelated alpha' from the systematic part, and 'multi-strategy' funds allocating across the spectrum of single managers and / or focusing on specific asset classes, exist. In general, quant strategies of single managers should offer liquidity profiles for investors between daily and max. monthly, as the underlying traded assets are usually highly liquid and the performance should be uncorrelated to traditional equity and credit asset classes. Extensive investment and operational due diligence is required to assess the quality of individual strategies.
This perspective is neither an offer to sell nor a solicitation of an offer to buy an interest in any investment or advisory service by Stone Mountain Capital LTD. For queries please contact Oliver Fochler under email: oliver.fochler@stonemountain-capital.com and Tel.: +44 7922 436360 and Alexandros Kyparissis under email: alexandros.kyparissis@stonemountain-capital.com and Tel.: +44 7843 144007.
The views expressed in this article are those of the author and do not necessarily represent the views of, and should not be attributed to, Stone Mountain Capital LTD. Readers should refer to the Disclaimer.
From a practical perspective there are plenty different quant and algorythmic trading strategies in the hedge fund sector available: 'systematic broadly diversified CTA' across markets like equities, fixed income, currencies and commodities, metals, indices etc. trading in futures vs. 'sector specific systematic CTA' strategies trading only in equities, bonds, etc. Other strategies include 'trend followers' (based on rigorous research any trend in datasets is traded long or short) vs. 'contrarian' (the system holds until a trend has formed and after a specified period of time trades automatically in the other direction), 'systematic' vs. 'discretionary' (trades are executed by the traders not the system), 'discretionary global macro' trading based macro-economic trends in futures and exchange traded options broadly diversified in equities, fixed income, currencies, commodities, indices etc. vs. global macro sector specific like e.g. fixed income focused vs. global macro focused predominantly on a country like the U.S. or region like Europe and mixtures of discretionary and systematic strategies in which e.g. the trade selection is sourced from a system and the execution is done from traders or for certain asset classes the system trades automatically and for others trading is discretionary. Further differentiating factors can be long only vs. long/short vs. market neutral strategies. Recent developments in the drive for lower fees for investors differentiate quant strategies into actively managed hedge funds and lower fee 'smart beta' strategies. The fund of hedge fund (FoHF) market is mirroring the single manager funds. FoHFs with purely systematic CTAs, diversified FoHFs with broad allocations to equity hedge, credit hedge and CTAs in order to generate 'uncorrelated alpha' from the systematic part, and 'multi-strategy' funds allocating across the spectrum of single managers and / or focusing on specific asset classes, exist. In general, quant strategies of single managers should offer liquidity profiles for investors between daily and max. monthly, as the underlying traded assets are usually highly liquid and the performance should be uncorrelated to traditional equity and credit asset classes. Extensive investment and operational due diligence is required to assess the quality of individual strategies.
This perspective is neither an offer to sell nor a solicitation of an offer to buy an interest in any investment or advisory service by Stone Mountain Capital LTD. For queries please contact Oliver Fochler under email: oliver.fochler@stonemountain-capital.com and Tel.: +44 7922 436360 and Alexandros Kyparissis under email: alexandros.kyparissis@stonemountain-capital.com and Tel.: +44 7843 144007.
The views expressed in this article are those of the author and do not necessarily represent the views of, and should not be attributed to, Stone Mountain Capital LTD. Readers should refer to the Disclaimer.
RSS Feed