Three keys to Alpha Generation with text Analysis
by Tim Decker – CEO RelateTheNews
Alternative data is applied with increasing frequency to power the alpha generating capabilities at a large variety of firms – from quantitative hedge funds to global asset management firms. Alternative data refers to data that is distinct from traditional market data (ie financial statements, volume, volatility, historic market prices, historic financial data, etc.) often falling into the realm of unstructured, more organic datasets or data which is not created with the purpose of informing the investment lifecycle. (ie satellite images, weather data, traffic data, etc.). In the stock market knowledge is a key component to having an edge. Having an edge increases the ability to generate alpha. Knowledge is gained through experience and enhanced with information. So, how do investment firms, quants and even individuals evaluate, and apply new alternative data sources, specifically textual sources of data, to increase their alpha generating capability with a knowledge based edge? There are 3 keys to success with textual data to find alpha generating insights:
- The text matters: Over the last several years the industry has adopted the use of social sentiment data. This alternative data source is important and has proven to be valuable to some. Yet, at the core social sentiment is still merely a proxy for an individual’s feeling about a given stock or company. Moving beyond this early phase of sentiment analysis is accomplished by insuring that the measurement of more rigorous text is achieved. The text’s source and the text’s content matter. Analyzing in a repeatable quantitative manner regulatory filings, traditional media news, and correspondence is key to a robust and alpha enhancing alternative data program.
- The engine is everything: With cars it is simple and easy to experience the difference of the exact same car with two different engines. Take any make of car or truck, any model with two engine choices and pick the larger engine, the one with more horsepower, or the one with more torque and the difference will be instantaneously noticeable – to the driver. And yet to an outside observer they look the same. The engines of Sentiment analysis and natural language processing(NLP) are designed and built to achieve goals in hundreds of different ways. There are ‘bag of words’ approaches. There are sentence structure approaches. NLP has become a very hot topic lately providing the backbone of many of new chatbot applications as well as powering search engines. One example of a recent release of an NLP engine was by Google in May of 2016. Many of the generally available NLP tools are built to ‘guess’ the next word someone will type (in a search engine for example) or approximate a valid response to an inquiry(chatbots and search engines). The analysis applied in these situations does not make an engine that can be immediately nor necessarily successfully used to create an edge in the financial markets. To generate alpha with text based data it is crucial to use the right engine – an engine built to analyze financial market textual data.
- Integration: Using alternative data to power alpha generation depends on the data and the engine but most important is integration of that data into a firm’s systems. The necessity to quickly and simply plug data into investment selection, trading, and risk management or compliance systems is as critical to alternative data as it is to traditional market data sources. Forgoing the integration process can mean that opportunities are missed. Silos of alternative data diminish the composite benefit of all data sources being available and applied in every stage of the investing life-cycle. By in-lining alternative data analysis of textual data with existing systems the entire firm’s investment process is enhanced. Alpha generation as well as risk management within existing systems is enhanced when a complete yet simple integration is achieved.
With these 3 key factors in mind a quant, an institution, an advisor and asset management firms can begin to access the power of textual data analysis. The real power of this applied alternative data to generate alpha is the transformation of text into a quantifiable entity that can be compared to itself over time and to other quantified distinct and relevant financial data (ie earnings, prices, volume, volatility, financial statements, etc.).
Tammer Kamel the CEO of Quandl stated in April 2016 – “We’re in the middle of a data revolution.”
Now is the time to participate in that data revolution. By starting now – following these 3 keys, constantly seeking and applying new innovative data into the investment lifecycle; firms can gain an edge. An edge that is a critical component of driving alpha.