Quant Moneymakers

Technological Advantage:

When James W. Marshall found gold in Sutter’s mill California on January 24, 1848, he kicked of an unparalleled gold rush in history. Over 300,000 individuals travelled to Coloma in hope of finding the American dream and the luckiest miners found large gold nuggets with some being the size of a chicken nugget. However, contrary to popular beliefs, the amount of gold is more abundant today than it was 170 years ago due to rapid technological advances. Large gold companies now use microscopes and advanced chemical processes to mine gold at infinitesimal levels. Technological improvements have not only impacted the gold industry, but the advancements have also had a dramatic impact on the investment industry. Back in the 1950s, knowing how to use a simple calculator was a technological advantage as market inefficiencies were ginormous. Taking advantage of a large market inefficiency using a calculator in the 1950s is analogous to finding large pieces of gold during the California gold rush. The evolving market has led to smaller inefficiencies and some experts even argue that the market is now perfectly efficient. However, Adaptive Alph would argue that there is more inefficiency in the market today than it was in the 1950s; the difference is just that these anomalies are infinitesimal. Quantitative Hedge funds have optimally adapted to this new market paradigm as they utilize computers, algorithms and statistical theory, which is the investment version of microscopes and advanced chemical process for turning market anomalies into profitable trades.

Computers us binary language, but can through abstraction generate complex relationships

What is a Quantitative Hedge fund

Quantitative Hedge funds are both glorified and vilified by the media across the world. Compared to other investment vehicles in the financial markets, such as the typical mutual fund or ETF, Hedge funds have a much broader investment mandate due to relaxed legal and regulatory constraints. A broader mandate increases the opportunity set for Hedge funds and therefore opens up the possibility to trade in illiquid risky markets, charge higher fees, apply a higher degree of leverage and use collateral creatively to maximize returns. The mandate’s flexibility also increases the risk associated with investing in Hedge funds and as a result, most countries only allow qualified investors to invest in a Hedge fund. A qualified investor typically meets a minimum asset threshold. For example, an accredited investor in the U.S. must have a net worth of at least 1.0 MUSD, excluding the value of the investor’s residential property, or have an annual salary above KUSD 200. What is then a quantitative Hedge funds secret sauce? The answer lay hidden inside their proprietary moneymaking model used to capture misspricing and patterns in the market.

Quantitative hedge funds rely on data

Quant Strategies

There is a vast array of quantitative Hedge fund strategies in the market and each strategy has a unique investment process to identify mispriced instruments. As a simplification, experts divide quantitative funds into those basing their strategy on fundamental analysis and technical analysis. The former is often convergent in nature and the latter is both divergent and convergent depending on what type of market anomaly the model is exploiting. An example of a convergent strategy is to use stock fundamentals, such as earnings, to estimate a stock’s intrinsic value and then compare its price to the market price. If the intrinsic value is over or under the market valuation there is a mispricing that the investor will arbitrage away and a convergent strategy is therefore mean reverting in nature. Unlike convergence, a divergent strategy profits from complex shocks to the system creating new expected means. A perfect example of a market shock is overreaction by investors to a particular news story, such as the Corona virus, which cause the market to trend in one direction. Technical quants then have models in place to profit from these market trends. No matter if the edge is fundamental or technical, all quant strategies combine proprietary moneymaking models into a portfolio to generate attractive returns for their investors. The fundamental moneymaking models are based on economic or company specific fundamentals. The input factors are either macro based tending to impact financial instruments on a systematic basis or fundamental idiosyncrasies. The technical moneymaking models are based on statistical analysis of patterns most frequently in prices of assets, but also on seasonal price, price derivatives and price volatility. The price patterns that can emerge in the markets are wide-ranging and in order to capture as many of them as possible the technical quant uses various statistical techniques to diversify across time horizon, instruments and other statistical factors.

Dribbling the basketball is fundamental for an NBA player

The Fundamental Story

Fundamental quant models are often based on intuitive factors such as yield, growth and quality, which are also concepts heavily used by traditional investors such as Warrant Buffet or Bill Gross. The comparative advantage of quantitative models is their unemotional and empirical nature. Through scanning and analyzing a large investment universe across fixed income, equity, commodities and currency markets, a quant locates those assets whose intrinsic value has deviated from the efficient market price by using predefined algorithms. The biggest difference between a quant and a discretionary trader is hidden in the conviction level. Due to infrequent trading, the discretionary trader must in general have a higher conviction level on the intrinsic value of a security than a quantitative trader relying on algorithms. The reasoning is that a human brain can at most analyze say about 30 financial instruments at one moment in time, while the algorithm can analyze many thousands of instruments and also the complex relationships between them. The conviction level must thus be higher for the human traders because if they misidentify securities as bargains they also miss out on other potential investment opportunities. Quants, however, have the capability to include all assets that meet the predefined constraints set by the algorithm as they can include an unlimited (constrained by $ in portfolio) amount of assets in the portfolio. A quantitative hedge fund can therefore hold any instrument that adds incremental value to the portfolio, which in turn involves the golden rule of diversification. Below is a description of the factors in fundamental investing.

Types of Fundamental Strategies

Bargain investing!

Value/Yield

Value investing is perhaps the most well known factor in finance. A security is considered value by an investor if it has a higher intrinsic value than the market is willing to pay. Value securities have generally fallen out of favor by the market for a particular reason. Perhaps the security is underpriced as a result of a poor business model or toxic management. The security might also belong to an industry or asset class considered unfavorable due to the cyclical nature of the market or possibly the industry or asset class as a whole is doing well, but this particular security’s price is lagging behind. The most classic case of value investing is analyzing the price to earnings ratio of a stock. The price is how much an investor pays for a stock and the earnings are the company revenue. More revenue should indicate a higher price of the stock, all else equal, as a greater amount capital allows a company to pay a dividend to investors or reinvest in new projects to increase future earnings. A classic relative value quant strategy is to buy companies within a sector that have a low PE and sell short companies in the same sector with a high PE. This type of relative value strategy also works in currencies or fixed income and is called a carry trade. For example, if the US banks typically pays an interest of 5% and the Japanese bank pay 1%, an investor could lend in Japanese Yen from a Japanese bank, then convert the Yen into Dollar and finally invest the Dollar in a US bank for a profit of 4%. The investor is then betting on a stable currency relationship between US and Japan. Below is a description of how to capture the value factor in fundamental investing.

Diversification: Value

· Equities

o Value indicators

- Book to market tiers (HML factor)

- P/B or P/E

- Market Cap tiers (Large vs Small Cap)

- Pair trading

- Other types of ratio analysis based on balance sheets

· Free cash flow, EBITDA multiples, Overhead costs

· Portfolio exposure

o Across various sectors, currencies and geographies

· Fixed Income

o Relative value

- Yield Curve anomalies

- Bond vs Bond mispricing

- Spread Volatility

- Credit ratings

- OAS and swap spreads

- Bond vs Yield curve (Libor) mispricing

· FX Currency

o Carry

o Relative Carry

· Model and Program Diversification

o Time horizon

- At some point the security might become overvalued.

o Investment universe

o Risk management

o Portfolio Construction

o Position sizing

Where are you planting seeds?

Growth

Before a security is considered a value asset it must first have experienced a growth period at some point during its lifetime. A growth quant applies algorithms to identify and screen instruments, markets and sectors that are most likely to experience rapid growth. A growth quant in equities, for example, generally utilizes forward-looking measures of revenue/sales/earnings-growth of companies instead of looking at current earnings as a typical value investor. A frequently used metric to identify growth companies is the P/S ratio. The P is the price of the stock and the S is the amount of sales. A lower P/S ratio indicates that revenues are relatively higher and therefore a potential buying opportunity. Growth companies may even have a negative profit margin, but the hope is that sales will continue to grow and costs will decrease as the company scales its operations. Below is a description of how to capture the growth factor in fundamental investing.

Diversification: Growth

· Macro analysis

o Monetary and fiscal policy

o Inventory reports

o Investor sentiment

· Balance sheet

o Sales figures

o Costs

· Growth specific for company/industry/sector

· Relative growth value

o Long fast growth companies and short assets with slow growth

- Peg ratio -> forward looking concept of value

- GARP -> growth at a reasonable price

· Commodities

o Demand increases for commodities used in production

o Analyze inventory reports for supply shocks

· FX

o Relative strength of economies

o Interest rate analysis across the curve

· Model and Program Diversification

o Time horizon

- Growth does not last forever

o Investment universe

o Risk management

o Portfolio Construction

o Position sizing

If its not high quality then short it!

Quality

One of the most recent factors that have gained popularity among quantitative researchers is quality. When investing in either a company or asset class high quality is preferable for the portfolio’s long exposure and poor quality for the short exposure. In equity investing, the quality of a company is relatively greater if the leverage used by the firm is low, the revenue is generated from a diverse set of sources and the quality of management is high. By applying algorithms to company balance sheets, a quant is able to capture all of the above metrics and compare a specific company to other companies in the industry or other sectors. For example, fraud is an indicator of poor management and it is possible to detect fraud by looking at the balance sheet. If there are inconsistencies in certain financial ratios it suggests that the management is hiding something. It is also possible to capture the amount of leverage applied by the company by analyzing debt to equity ratios. Below is a description of how to capture the quality factor in fundamental investing.

Diversification: Quality

· Quality indictors for a company/industry/sector

1. Leverage

· Lower levels of leverage equals less risk

· Debt/equity ratio

· Fixed income specific

o Lower prob of default outperforms high prob of default during credit downturns

o EBITDA relative to interest expenses (12 month)

o Leverage factor

2. Diversity of revenue sources

· Revenue is generated through product differentiation

· Volatility of revenue

· Stable cash flows is preferable all else equal

3. Management quality

· Analyze accruals of the company FS

4. Fraud

· Less fraud decrease the risk of an investment trap

5. Sentiment investors

· Prospective changes in leverage, revenue diversity, management quality and fraud risk

· Model and program diversification

o Time horizon

o Investment universe

o Risk management

o Portfolio Construction

o Position sizing

Maths

The Technical Story

Technical models utilize advanced statistics to forecast prices of financial instruments in the markets instead of relying on intuitive fundamental factors such as interest rates or GDP growth. To generate trading signals, the technical models scan historical data for patterns. The hope is then that these patterns will persist in the future. Patterns that the models are searching for are either based on theory or data driven statistical assessments. In theory driven models a hypothesis is formed about what drives the underlying price of an asset. Currently there are three types of widely accepted theory based quant strategies and they are trend, mean reversion and technical sentiment. Data driven models, however, are machine learning models built purely on mathematical construct, but may include a sprinkle of theory as the quant analyst is the model builder and generally must have an idea of how to apply the data driven algorithm on market data. Below is a description of the factors in technical quant investing.

Types of Technical Strategies

The trend is your friend and your guide - follow it!

Trend

Is based on the concept of herding behavior that drives asset prices away from efficient levels. The simplest technique to measure a directional trend is the price break out strategy. For example, if the price of the S&P 500 is higher today than the price of S&P 500 3 months back, then the S&P 500 is in a positive trend. Another straightforward method to measure trend is to compare moving averages (MA) of S&P 500 price with different time horizons or lookback periods. When the short 50 day MA is greater than the long 200 day MA of the S&P 500, then the S&P 500 is in a positive trend. Advanced trend followers create complex trend indicators based on advanced statistical methods such as spectral analysis, principal component analysis and basket analysis. These advanced methods have recently outperformed the simple methods described above. Below is a description of how to capture the trend factor in technical investing.

Diversification: Trend

· Alpha indicators

o Price breakout

o Moving average

o Exponential moving average

o Double exponential moving average

o Volatility

o Price acceleration

· Conditioning variables

· Varying inputs used above

· Heuristic risk limits such as max VaR levels at instrument, market and sector level

· Volatility targeting

· Stops

· Model and program diversification

o Time horizon

o Instrument universe

o Risk management

o Portfolio construction

o Position sizing

If you can find an average in the market and it is not priced at that average = anomaly

Mean Reversion

Markets that have trended strongly tend to snap back when the price is overextended according to investors. If herding type behaviors drive trend then mean reversion is driven by rationalization or the tendency of more normal outcomes typically following extreme events. The concept underpinning mean reversion is “regression to the mean” and was first observed by the English polymath, Francis Galton. The simplest form of mean reversion is to pair trade two securities that tend to move in unison. If a statistical relationship is observed then this relationship tend to oscillate around some type of mean. For example, if the historical exchange rate between USD/EUR is around 1.10 and all of a sudden the value of the dollar depreciates to 1.50 for a euro it is likely that the currency relationship will snap back to 1.10 again. Sometimes statistical relationships breakdown (LTCM crash) so as always, risk management, conservative position sizing and careful monitoring will be critical to maintaining survival and profiting over the long term. Another method to take advantage of a security that has dramatically diverged from its efficient price is to apply a relative strength indicator (RSI) on that security. If the RSI model concludes that a 20-day lookback period is optimal on the S&P 500 and the S&P 500 had more positive then negative moves over the 20-day time period then the RSI indicator is low indicating that S&P 500 is overbought. The model would then forecast that S&P 500 will fall back to its normal level and a short position on the S&P 500 is entered. Another method to capture mean reversion is to look at standard deviation or so call Bollinger bands. Below is a description of how to capture the mean reversion factor in technical investing.

Diversification: Mean reversion

· Alpha indicators

o Bollinger bands

o RSI indicators

o Moving averages

o Exponential moving averages

o Double exponential moving averages

o Volatility

o Price acceleration

· Conditioning variables

· Varying inputs used above

· Heuristic risk limits such as max VaR levels at instrument, market and sector level

· Volatility targeting

· Stops

· Model and Program Diversification

o Time horizon

o Investment universe

o Risk management

o Portfolio construction

o Position sizing

Multiple dimensions

Machine Learning

Quantitative machine learning models (ML) have the potential to extract a differentiated source of return across all asset classes adding diversification benefits to the typical convergent investment portfolio. For a ML based model to successfully deliver high returns, a prerequisite is that the model is sophisticated and building groundbreaking ML models require a strong background in mathematics. As a result of the complexity of developing ML models, prominent investors consider ML an alpha investment strategy. Important to note is that ML itself is not an alpha strategy, but rather a quantitative technique deployed to squeeze alpha from known and unknown market anomalies. Adding ML to a quantitative program therefore enables extraction of alpha from factors impacting the market that are not flying under behavioral banners such as herding or loss aversion. The common feature of all ML models is that the forecasting parameters constantly evolve with patterns emerging from the market. If the model parameters are carefully selected and applied correctly, the constant adaption prevents the model from overfitting to historical data compared to more traditional quantitative and econometric models when making likelihood predictions. Each ML technique also posses a unique edge and the quant should therefore diversify across various techniques to increase the risk reward tradeoff. As a visual ML example, imagine that there is a magnet, a table and more than a million pennies in front of a quant analyst. The magnet represents the ML algorithm and its objective is to attract a bunch of pennies on top, under and to the sides of this table. The pennies are equivalent to historical price data points generated by the market and used by the model for future prediction of a time series. The table surface in turn represents a two-dimensional hyperplane. The quant analyst’s job is to create an optimal ML magnet and then throw the magnet on the table to let the magnet figure out the optimal location for future maximum penny attraction. If the quant analyst’s forecast is made in two-dimensional space based on a univariate time series, the pennies and the magnet are bounded by the table surface. More complex relationships takes place in multi-dimensional space undetectable by the human brain, but lucky for us an ML algorithm is based on mathematics and is able to forecast none-linear complex patterns. The ML algorithm therefore most likely shifts the tossed magnet away from the table surface to maximize penny attraction from a multi-dimensional perspective. When a new penny is added to the table or somewhere else in space, the magnet stays put or moves in any direction that maximizes its magnetic pull. The patterns that are considered valuable by the ML algorithm are then used for prediction purposes. Below is a description of how to capture the ML factor in technical investing.

Diversification: Machine Learning

· Machine Learning Techniques

o Recurrent Neural Networks

o Convolutional Neural Networks

o Support Vector Regression

o Random Forrest

o PCA

o Naïve Bayes

· Online Learning

o Strengthening forecasts as data becomes available in sequential order

· Regularization techniques

o Adding a penalty function in the objective function

o Early Stopping

· Ensemble Techniques

o Varying layers, nodes, activation functions and inputs in a ANN

o Combine view of different algorithms to strengthen the final prediction

· Alpha indicators

o Fundamental Factors

- Growth rates, P/E ratios, interest rates and many more

o Bollinger bands

o RSI indicators

o Moving averages

o Exponential moving averages

o Double exponential moving averages

o Volatility

o Price acceleration

· Conditioning variables

· Optimization techniques

· Heuristic risk limits such as max VaR levels at instrument, market and sector level

· Volatility targeting (Or not)

· Model and Program Diversification

o Time horizon

o Investment universe

o Risk management

o Portfolio construction

o Position sizing

Cheap sentiment might be an opportunity

Sentiment Strategies

Trend, mean reversion and machine learning are indirect sentiment based strategies, but there are also strategies that are trying to directly capture value from sentiment. Overall market sentiment is often a topic in the news as media companies measure and report to the public if the temperature of the market or asset class is hot, cold or neutral. Many companies also provide opinion polls where they ask question such as “do you think the market is stronger this year than last year”. There is a great deal of noise in opinion polls as people have a weak incentive to answer truthfully and often answers are based on how one feel at the moment when the question is answered. That is why most investors turn to other sentiment indicators such as the put to call ratio (P/C) or the VIX index. When an investor buys a long put option on a market index, the investor expects the market to fall and a P/C ratio greater than 1 therefore signifies that the marginal investor expects a market down turn. A quant investor could have models that take positions in the market based on the P/C ratio. The VIX measures real time 30 day forward looking volatility derived from option price inputs of the S&P500. Volatility measures, over time, the size of price movements that a financial instrument experiences.A higher volatility is equal to greater uncertainty of future price movements and is therefore considered a fear indicator. The quant can use the VIX as a predictor of future market movements as well. Below is a description of how to capture the sentiment factor in technical investing.

Diversification: Sentiment

· Scan newspapers and social for positive and negative sentiment

· Prospective on changes in leverage, revenue diversity, management quality and fraud risk

· Put/Call Ratio

· VIX

· Margin Debt

· Mutual funds cash position

· Short interest ratio

· Model and Program Diversification

o Time horizon

o Investment universe

o Risk management

o Portfolio Construction

o Position sizing

Either way you must invest

Conclusion

A quant systematically applies an alpha‐seeking investment strategy that is specified based on exhaustive research. What makes a quant a quant, in other words, almost always lies in how an investment strategy is conceived and implemented. It is rarely the case that quants are different from discretionary traders in what their strategies are actually doing ~ Rishni Narang

Hedge funds are different from the typical mutual fund and only allow qualified investors to invest in their alpha capturing strategy. Quantitative firms in turn are a subset in the Hedge fund space applying algorithms to capture micro anomalies in the market. From a macro perspective, the fundamental and technical approach is the two types of quant strategies that currently exist in the quantitative world. The former uses value, growth and quality indicators to forecast prices in the market. The fundamental approach is generally more intuitive. The latter, technical strategies, is spearheaded by trend, mean reversion, ML and sentiment indictors to profit from the markets. The technical approach tend to rely more on statistics and a background in mathematics is therefore of great importance. Both types of strategies, if applied correctly, are better suited at capturing small market anomalies as compared to their discretionary counterparts. By combining these seven branches of quantitative investment strategies (counting ML as a separate strategy), one conclusion is that there exist an infinite amount of unique quantitative investment strategies. Due to the rapid improvement in technology, the future is bright for quantitative investing and Adaptive Alph look towards the future with great interest.

The end.

//Stay adaptive

Well done!