The Startup Magazine Dan Calugar on the Role of Sentiment Analysis in Algorithmic Trading: Harnessing Market Emotions for Profit

[ad_1]

Professional traders have always integrated sentiment analysis into their investment strategies. For years, traders have tried to predict how markets might react to certain breaking news information, for instance, based on whether those news items could result in particular assets increasing or decreasing in value, according to experienced investor Daniel Calugar.

investment strategies and advice

Today, though, sentiment analysis has simultaneously become not just more prevalent but more challenging to integrate into investment decisions and strategies.

There is so much more information available at our fingertips — and from so many more sources — today than there was a few decades ago that it’s become more mainstream to attempt to draw conclusions from this information. 

Algorithmic trading seeks to gauge market sentiment in various ways to better predict price movement that, in turn, will result in better returns.

Below, Dan Calugar will explain how essential tools such as sentiment analysis and natural language processing are integrated into algo trading systems to produce higher returns.

Sentiment Analysis Defined

Sentiment analysis is the process of systemically identifying, quantifying, studying, and extracting subjective information from various forms of data. In simple terms, it’s extracting meanings from things such as images and texts.

In the past, the way to do this was to simply read, read, and read some more. This included reading news headlines, texts, and opinion pieces, as well as going through company earnings reports to extract subjective information indicating whether the market would move in a particular direction or not.

Today, sentiment analysis uses high-powered computers to do all this manual work for us. It’s done by using computational linguistics, text analysis and natural language processing — all of which, as terms, are often used interchangeably.

Natural Language Processing and Computational Linguistics

Natural language processing, commonly referred to as NLP, is a discipline of computer science that trains computers to understand spoken words and text similarly to how humans do. In this way, Daniel Calugar says NLP is really a subset of AI or artificial intelligence.

Computational linguistics, meanwhile, is another computer science discipline that involves analyzing and comprehending spoken and written language. It also uses traditional computer science linguistics with AI to understand human language from a computational perspective. 

Together, the technologies of natural language processing, computational linguistics, deep learning models and machine learning all make it so that computers can process language from voice and text data to fully understand its meaning. 

The computers will arrive at not just the meaning of the text or spoken language but also the sentiment and intent behind it. 

NLP is used in many applications today. It helps to power programs that take text and translate it into multiple languages instantly. It powers programs that can respond to a human’s spoken commands. It even powers programs that can summarize a lot of text very quickly.

It’s present in many consumer products today, such as Google Translate, Amazon Alexa devices, GPS systems, dictation software, chatbots, digital assistants, and many other software products that are designed to bring convenience to consumers.

NLP in Algo Trading

NLP is extremely useful for algorithmic trading, as it can help uncover market sentiment in a number of different ways that simply would be too difficult for humans to do on their own. Even a large team of traders, for example, would have a tough time keeping up with the computers that process NLP — even if that team focused 100 percent of their working time on sentiment analysis.

That’s because computers can process information and data exponentially faster than humans can. Plus, they can work around the clock without resting and can instantly process information in multiple languages.

Dan Calugar says that bringing in all the different pieces of information and data from relevant sources is also relatively easy today, thanks to APIs. These tools allow computers to automatically pull in raw data and information from sources such as blogs, news media, social media platforms and more. This means humans don’t have to be involved at all in the process once the algorithms and APIs are set up.

Here are some practical ways that natural language processing can be used to derive market sentiment in algorithmic trading.

Breaking News

Sometimes, breaking news from around the world can have significant effects on the market. But maximizing returns from these breaking news events relies a lot on timing.

In other words, in order to get the best returns following breaking news, investors have to not only correctly predict the market sentiment from that item but then also act quickly to execute the trade.

On April 2, 2023, members of OPEC+ — the Organization of the Petroleum Exporting Countries — announced surprise cuts in oil production. That caused Brent Crude Oil Front Month futures to spike from $75.32 per barrel on March 21 to a high of $87.33 as of April 12 — an increase of 15.95 percent.

The largest jump in price, though, occurred right after the announcement was made. On March 31, the price was $79.77. By the next trading day, that price had jumped to $84.93 — a 6.47 percent increase.

Computers using NLP can quickly and instantly gather and analyze that information and then act on it. While it may not be as vital for formal announcements from an organization such as OPEC+, the concept applies to any other breaking news information — some of which might not be as well publicized, and some of which might happen while much of the United States is sleeping.

Financial Reports

Publicly traded companies are required to release certain financial reports at regular intervals throughout the year. It’s never a surprise when this information is released, so you might not see the immediate value in NLP in this instance.

However, Dan Calugar says that when you consider that there are literally thousands of pages of financial reports released at these regular intervals, it’s easy to see how a computer can help identify the important nuggets of information that they all contain.

In just a fraction of the time that it takes a human to read only the introduction of one of these reports, a computer can literally analyze thousands of full reports.

While machine learning isn’t quite at the stage where it’s able to derive specific insights from a large amount of unstructured text data, it is good at identifying trends from multiple sources.

In other words, these computers might not be great at identifying market sentiment from one individual financial report, but they are great at making predictions for an entire portfolio.

The algorithms might produce significant variance from one report to the next, for instance, but that would likely balance itself out when combined with many other reports. In the end, investors can get figures for standard deviation and average predictions, input that back into an algorithm, and finally determine what and how much of each stock should be traded.

This helps to maximize profits and minimize risks in ways that humans simply couldn’t do on their own.

Online Crowd Sourcing

A very popular focus of NLP in algorithmic trading today is gathering insights from online crowdsourcing. This can be done on social media platforms, web forms, opinion pieces and reports from analysts.

Individually, any of these sources might be considered too ineffective to rely on for trading decisions. When taken together and combined with other tried-and-true analytical tools and information sources, however, online crowdsourcing can be extremely valuable. 

Since there’s so much information on these sources about so many different topics, it does take some finesse to glean usable information. One example is using online crowdsourcing to help make a final decision on whether to invest in a particular stock. 

Let’s say your analyses are telling you that Zoom would be a good company to invest in. However, you’re a little concerned that it could be overtaken by other similar software, such as Microsoft Teams.

Computers can be programmed to pull in data from social media platforms and then told to sort that data whenever people are talking about it. In this instance, you’d have the NLP system sort the data by keywords so that you could determine what the market sentiment is for Zoom and Microsoft Teams.

This wouldn’t include just text posts, either. It could include memes, images and videos that are posted to the platforms.

By doing this, you might be able to identify whether the general market sentiment is that Zoom is their preferred video communications platform or whether more companies are moving to Microsoft Teams.

Categorizing Market Sentiments

The above are explanations of how NLP and sentiment analysis can be used practically to make trading decisions. But Dan Calugar says it’s important to also understand the nitty-gritty of how market sentiment works from a technical standpoint.

Computers will take all the information that is inputted into them and then “spit out” results. These sentiments can be applied to the market at large or to an individual asset within a specific market. 

What do those results look like, though? 

Commonly, algorithmic trading systems will be set up to produce either a negative, positive, or neutral sentiment quantitative value for whatever it is they are analyzing. In order to produce the best results for traders, the analysis will usually express the degrees of the sentiment.

For instance, algorithmic traders might set up a 0-100 scale for their sentiment analysis. On this scale, 0 would represent extremely negative (bearish) sentiment, 50 would represent completely neutral sentiment, and 100 would represent extremely positive (bearish) sentiment.

There are many advantages to using a numerical system for sentiment analysis.

First, the 0-100 numerical scale described above allows investors to see just how bearish or bullish the sentiment is. A generic “negative, neutral or positive” sentiment isn’t really useful from a trader’s perspective — especially considering how advanced the technology is that they’re using to conduct the analyses.

Second, when the output of the sentiment analysis is displayed numerically, it can then easily be integrated back into an algorithm to make trading recommendations and/or execute trades automatically.

In this example, algo traders would set up their algorithms with parameters, just like they do for all other signals. Then, the algorithm might execute a buy automatically if the sentiment score for a particular asset rises above, say, 85 or execute a sell automatically if the score drops below 25.

Conversely, the algorithm could just send signals to the trader if it reports any scores within a certain range. This would enable you to use the sentiment score as a supplementary piece of information for trading decisions rather than the only one.

Algo Trading Strategies Based on Sentiment Analysis

Daniel Calugar says that investors can create various algorithmic trading strategies based on sentiment analysis. Tools such as NLP and computational linguistics can help traders harness market emotions for profit. 

Below are some common types of algo trading strategies that are built around sentiment analysis.

Contrarian

While many of the examples outlined above showed how you can use sentiment analysis to follow market sentiment — what would be called a trend-following strategy — the contrarian strategy would see you do the opposite. It involves taking a position that’s opposite to that of the prevailing market sentiment. 

If market sentiment reaches a point where it would be considered excessive — either negative or positive — traders who believe in the contrarian strategy argue that the market is due for a reversal. 

When market sentiment is very negative, the contrarian strategy will say to buy in anticipation of that asset or market bouncing back. When market sentiment is very positive, the contrarian strategy will say to sell, as a downward correction is anticipated.

Momentum

A close cousin to the trend-following strategy is the momentum strategy. This strategy combines momentum indicators with traditional market sentiment analysis.

The goal of this strategy is to identify situations in which market sentiment begins to rapidly shift one way or the other. When strong market sentiment momentum appears, the strategy will align with the shifting momentum.

The theory is that the strong shifting momentum is an indicator that there is likely to be further movement with the asset’s price in that same direction.

The key to being successful using this strategy is getting in on the position while there’s still opportunity for positive returns. This is where a solid and tested algorithm can be extremely beneficial.

It could be difficult for humans to employ this strategy manually on their own because they might miss out on the window and end up buying or selling too late.

Event Driven

Dan Calugar points out that a lot of valuable trading information can be gleaned by the reaction to specific events. This includes the release of economic data, long-awaited regulatory decisions, earnings announcements and even product launches.

Since many of these events are planned well in advance, savvy algo traders will use market sentiment before the event in question to set a baseline. Then, once the event concludes, they’ll run a market sentiment again to gauge how the market has reacted.

Oftentimes, the immediate reaction people have to these events is very telling of where the markets will go. If a product launch announcement event is considered successful, then it could lead to huge sales of that product.

If the reaction to the event is positive, then algo traders might take a long position on that asset or market. Conversely, if the reaction is negative, then a short position is what might be in store.

Intraday Sentiment

One of the big advantages of algorithmic trading is the speed with which opportunities can be identified, and trades can be executed. This makes intraday trading — when multiple trades are executed on the same day — viable. 

The intraday sentiment strategy involves executing trades on a short-term basis. Traders will purchase an asset at one point in the day and then sell that asset later in the same day.

This strategy seeks to capitalize on shifting market sentiment and the price movements that occur based on that. Intraday trading like this can be considered somewhat risky for a few reasons.

On the one hand, if you are looking to capitalize on very small price movements, you might have to trade a large amount of the stock to make it worth your while. On the other hand, there’s obviously a very tight window in which to execute both the buy and sell trades. And if you miss out on the second part of the strategy, the consequences could be disastrous. 

There are many other ways that sentiment analysis can be integrated into algorithmic trading, including plenty of other sentiment-specific trading strategies that can help traders harness the power of market emotions for profit.

About Daniel Calugar

Daniel Calugar is a versatile and experienced investor with a background in computer science, business, and law. He developed a passion for investing while working as a pension lawyer and leveraged his technical capabilities to write computer programs that helped him identify more profitable investment strategies. When Dan Calugar is not working, he enjoys spending time working out, being with friends and family, and volunteering with Angel Flight.

[ad_2]

Source link

Leave a Comment

Your email address will not be published. Required fields are marked *