Quant hedge fund primer: demystifying quantitative strategies

Statistical arbitrage funds generally utilise price data and its derivatives, such as correlation, volatility and other forms of market data, such as volume and order-book information to determine the existence of patterns. Research that identifies repeating patterns and linkages in the data can be used to help the manager forecast the future return of a stock, often over a relatively short timeframe. Relationships are identified through rigorous statistical analysis and back-testing of relationships. The strategy typically targets higher risk adjusted returns (Sharpe ratio) relative to more ‘traditional’ hedge funds, but the absolute level of returns is highly dependent on the leverage deployed and the volatility tolerance.

The most common signal types are mean reversion, momentum, and event driven.

• Mean reversion
  • Mean reversion looks to take advantage of the phenomenon of short-term price movements occurring due to supply/demand imbalances with a view that these will then revert back to an equilibrium level.
• Momentum
  • Momentum models look for patterns in price data that suggest that price movements will be more persistent (i.e., trend).
• Event driven
  • Other statistical arbitrage funds will look to incorporate more discrete information into their process from events. For example, the publishing of analyst earnings estimates, or systematically incorporating sentiment strength from news flow using techniques such as natural language processing (“NLP”). Announced mergers, share buy-backs, index rebalancing and corporate insider buying/selling of shares are other examples of event-driven data, all of which could be systematically incorporated into a trading system.

Whilst statistical arbitrage funds tend to focus more on ‘technical’ models driven by price and volume data, some may also incorporate some longer-term models that are driven by fundamental data, which can be both traditional and/or ‘alternative’ in nature (e.g., stock value models, balance sheet data, predictive sales data, etc.). If these more fundamental oriented models are the dominant driver of risk, then the fund is likely to be classified as ‘quantitative equity market neutral’ (see below).

Traditional QEMN strategies take fundamental and/or event-oriented data, such as , balance sheet information and cash flow statement statistics, and systematically rank/score stocks against these metrics in varying proportions. The weights of the scores of the different fundamental data sources may be fixed or dynamic.

Managers construct a portfolio that typically comprises hundreds to thousands of positions, long and short. Portfolio construction is done using an optimisation process, or by applying simple rules, combined with risk constraints, to create a portfolio that has netted out its long and short positions to be and/or beta neutral, and with minimal sector exposure.

QEMN fund price prediction models typically use a variety of signals as primary inputs, including:

• Fundamental data: this includes financial data such as earnings, revenue, profit margins, and cash flow, as well as non-financial data such as industry trends and macroeconomic indicators.
• Technical data: this includes information about past market trends and patterns, such as moving averages, relative strength, and trading volume.
• Sentiment data: this includes information about investor sentiment and market sentiment, such as news articles, social media posts, and analyst reports.
• Alternative data: this includes non-traditional data sources, such as satellite imagery, credit card data, and weather patterns, which can provide insight into market trends and consumer behaviour.

QEMN fund price prediction models typically use machine learning algorithms to analyse and process these various signals, looking for patterns and relationships that can help predict future price movements.

Commodity trading advisors (“CTAs”) take primarily directional positions in index level or macro instruments, such as futures or FX contracts, in a systematic fashion. Technically, a CTA is a trader of futures contracts as defined by the Commodities and Futures Trading Commission (“CFTC”) and historically, there were many CTAs who were not systematic; such traders are now more likely to be categorised as pursuing a ‘global macro’ strategy.

Trend following is the most employed CTA strategy. The rationale behind this is the view that human behavioural biases lead to the formation of trends which, as a result, can be exploited through a systematic strategy that aims to identify and trade such situations.

CTAs are extremely systematised with straight through processing from signal generation to execution. The addition of new and uncorrelated models, market instruments, trading time horizons, updated risk and portfolio construction models etc. can differentiate performance or add alpha over low fee more ‘simplistic/generic’ trend-following funds.

Today, trend-following quant strategies are very common; the basic concept is now heavily commoditised, and sometimes features within ‘alternative risk premia’ products. Fees tend to be quite low on more ‘generic’, trend-following CTAs.

• The vast majority of the managed futures/CTA universe incorporates some kind of trend following (using historical prices to determine predictable ‘trending patterns’) buying into markets where prices are rising and selling where markets are falling.
• When rising markets slow down/stop rising, trend-followers typically reduce their position and will eventually reverse its position into a short position, which it will hold until the market starts to rally again.
• The strategy is known for running with profits and cutting losses.
• There can also be a ‘grey area’ in some CTA funds, which employ a mix of trend and non-trend models, incorporating some fundamental data – some of these funds share similarities with quant macro.
• There can be some overlap with areas such as discretionary macro and quant macro funds. This is due to the relationships they exploit being synonymous with some of those also captured by macro funds.

• Price and volume data are the primary inputs to trend following price prediction models. Many trend followers focus their research on time-series analysis of returns of single instruments. This contrasts with an approach of looking at cross-sectional analysis of different instruments and contracts and looking for other predictive relationships based on correlations.
• Other models used in CTAs may incorporate other concepts such as: carry, seasonality, mean reverting or pattern recognition systems, models driven by fundamental data or non-traditional data sources (alt-data). As indicated above, cross-sectional analysis and correlations are integrated into the signal research process.
• Some CTAs can also trade very short-term signals driven by market microstructure anomalies and patterns.

Quant macro aims to generate alpha through large data set analysis and making trading decisions that are based on economic, market and other fundamental indicators as well as statistical models.

Quant macro funds have the potential for overlap between CTAs and quant macro, as they often trade the same/similar instruments, i.e., typically ‘macro’ instruments such as futures (across all the main asset classes), but also ETFs, bonds, FX and sometimes swaps and options. It is common for terms like quant macro and GAA to be used interchangeably.

There is no single ‘correct’ definition for quant macro. Funds that are clearly ‘quant macro’ determine price predictions for macro instruments through long-term fundamental factors like economic data or disparities between countries. But there are also funds with short-term average holding periods which are also defined as quant macro. These funds consider some of the factors listed above, but are driven more by shorter-term technical factors, supplemented with economic data. The economic data can be traditional, and/or derived from alternative data and ‘nowcasting’.

‘Quant macro’ can cover funds trading over a wide degree of time horizons, incorporate a lot of different datasets and can vary massively in complexity. However, the long-term performance of the quant macro strategy suggests an average Sharpe ratio in the region of 0.8-0.9 over the long term (slightly better than CTA), also exhibiting positive skew, but not to the same extent as CTAs on average.

Given the information above, quant macro fund positioning tends to be more relative value based, however, funds may also take directional positions. The asset classes traded are unsurprisingly macro in nature, i.e., futures, FX, ETFs, swaps, and other instruments – which are generally highly liquid. Quant macro/GAA managers are looking to trade repeatable relationships between macro level markets and the signals traded tend to be medium to long-term (i.e., one to several months). However, some managers that can exhibit a much lower average holding period than this.

Quant macro is a highly data and technology intensive strategy. It typically relies upon complex mathematical models to identify key drivers of price formation/relationships and market trends, calibrate portfolio construction, and generate trading signals. The focus tends to be on macroeconomic factors such as GDP, inflation, interest rates, exchange rates, export/imports, growth, capital flows, market data, etc.

Signals are also often classified under a number of familiar factor headings such as value, carry, momentum (trend-following) etc. Some examples of ‘classic’ longer term quant macro signals could be value signals expressed in currencies, looking at purchasing power parity, equity flows and terms of trade.

The strategy of alternative risk premia in hedge funds involves leveraging the fundamental principles underlying specific hedge fund strategies. By employing a dynamic yet well-defined process, these funds aim to harness a substantial portion of the anticipated returns associated with those strategies. Specific risk factors, such as equity value, momentum, size, etc. are isolated and their risk premia are harvested in a systematic fashion, hence their inclusion in the quant hedge fund strategy grouping.

The strategies/phenomena or mispricing that alternative risk premia funds look to capture are typically very well understood, backed up by academic research and – importantly – can be implemented systematically using quantitative models and a rules-based process to identify and exploit these potential opportunities. Because of this, risk premia products tend to be very low cost in respect of fees when benchmarked to the rest of the hedge fund industry, with much lower management and performance fees.

Risk premia funds are well understood by the broader investment community and as such there is a very significant allocation of industry capital to these strategies. A key attraction is that they have a lot of capacity and due to their nature, do not command high fees. Often such funds are thought of as a cheap alternative to investing in traditional hedge funds.

The theory behind this strategy is that by offering investors access to a very diversified mix of different and non-correlated alternative risk-premia, you get a robust return stream that is accretive to a traditional portfolio.

Another key characteristic of alternative risk premia funds is the flexibility in implementation. While some investors access these strategies through dedicated funds, others may choose to implement them using customised portfolios or factor-based approaches. This flexibility allows investors to tailor their exposure to specific risk factors based on their preferences and investment objectives.

Funds tend to have exposure to a well-diversified portfolio of hedge fund premia. Premia can cover everything from equity premia (equity market neutral – trading across value, quality, growth and momentum factors, as well as EM premia), macro premia (e.g., trend following, or EM premia), to arbitrage strategies (e.g., risk arbitrage – holding a portfolio of merger targets diversified by sector and deal type; convertible arbitrage, etc.). There can be some crossover with some risk premia products and some QEMN funds as they are looking to benefit from some of the same phenomena.

Some of the most common examples of risk premia signals include:

• Value: this premium refers to the tendency of stocks that are trading at a low price relative to their fundamentals to outperform stocks that are trading at a high price. To harvest the premia, hedge funds look to buy undervalued stocks and short overvalued stocks.
• Momentum: the tendency of stocks that have performed well in the recent past to continue to perform. Risk premia hedge funds may use momentum strategies to invest in assets that have exhibited positive momentum and short assets that have negative momentum.
• Carry: this factor captures the tendency of high-yielding assets to outperform low yielding assets, again – a market neutral approach to capture these premia would be to be long the high yielding assets and short the low yielding assets.
• Volatility: this factor captures the tendency of assets that exhibit low volatility to outperform those with higher volatility, with hedge funds long low vol assets and short high vol assets.
• Quality: the tendency of assets with strong fundamentals (high quality) to outperform low quality assets. Hedge funds would go long strong companies/assets and short the weak companies/assets.
• Liquidity: the tendency of illiquid assets to outperform liquid assets. Some risk premia strategies may use liquidity strategies to invest in illiquid assets vs similar liquid assets to capture this premium.