AI in Trading

Have you seen how talented traders do calculations in their head? Or have you seen in films how they actively record recent deals while simultaneously monitoring current data? Such traditional trading methods are outdated and are gradually disappearing. Those who want to keep up with the times and continue making a profit have begun turning to technical specialists.

Programmers, together with mathematicians and analysts, have provided the market with a useful tool — trading bots that carry out operations using embedded algorithms and provided data. The popularity of algorithmic trading on exchanges has resulted in the emergence of high-frequency trading.

Traders, brokers, and investment funds can no longer go without bot developers because people are not able to trade with small spreads at high speeds and concentration. Some people create strategies, others write algorithms, and bots trade using them — this is how trading works in the 21st century. Or, more to the point, this is how it did work.

With increasing competition in the market and the development of the Big Data field, the capabilities of bots have become insufficient. In automated trading, they began to be replaced by machines that, with the same performance, can also think like a human — models of neural networks and artificial intelligence (AI).

The participation of artificial intelligence in trading on exchanges

Professional traders are periodically forced to update their practices as progress makes trading more difficult. In 2000–2015, they had to compete with trading bots and then learn how to tune them to equalise the forces. Since about 2015, traders and their bots have had to compete with artificial intelligence.

Over the past five years, the number of AI trading systems has grown significantly. As they spread and influence the market, traders using legacy automation are seeing a drop in revenue. Conversely, those who use artificial intelligence to trade on exchanges perform better than the market average.

AI abilities in trading on exchange

Compared to bots that need to be constantly reconfigured, artificial intelligence can operate independently, without human intervention. It can come up with trading strategies, test and refine them. It can take into account market trends in order to improve with newly acquired knowledge. That is, AI can imitate the thinking of analysts.

Here are a few more AI powers that are being used to benefit traders, brokers, and funds, as well as their clients:

• Collecting information from news, social networks, thematic websites for its application in fundamental analysis
• Processing market data and stock market indices for their application in technical analysis
• Advising on medium-term investments
• Managing the asset portfolio, its diversification
• Compiling a rating of analysts, checking their performance when choosing the best strategies for mirror trading
• Building behavioural models during market turmoil
• Identifying cases of collusion and manipulation in the market

What AI is not capable of: doesn’t show human emotions such as greed and fear, doesn’t make irrational guesses. Although, these are not disadvantages, but advantages. On the exchange, AI performs better than algorithmic bots, not to mention humans.

Creation of trading strategies by artificial intelligence

Artificial intelligence is becoming a major component in developing trading strategies for hard-to-predict markets. It not only trades according to a written algorithm but constantly collects and processes huge amounts of data, analyses events and trends, and makes decisions itself.

Data collection

Today, specialists apply deep learning to AI to make it intelligently collect unstructured data from various sources such as news and social media posts. Usually, incoming data looks like a chaotic structure. However, even in such a structure, it is characteristic that past events can influence present and future trends.

AI uses historical data to understand how the current market would react to past events so it can adjust its trading strategies. Although this does not immediately bring a beneficial result, in the long term it allows AI to learn productivity in future market conditions.

Data organisation

The principle of operation of AI software for trading on exchanges is not much different from the approach used by analysts. The next step after collecting the data is to organise it and divide it into specific groups. There are two datasets:

1. A training set that is intended for training and tuning the algorithm before testing.
2. A test set that brings a calibrated algorithm into action.

Building a trading algorithm

The idea of the algorithm is to predict the price dynamics of an asset being traded by a trader or investment manager. There are many ways to build a predictive algorithm. However, most of them try to simplify the problem and then follow a two-class model based on signal and predictability factors:

1. The signal indicates what is expected: an increase or decrease in the price.
2. Predictability shows confidence in the signal.

After the output data is generated, the trader receives the desired signal.

Practical application of artificial intelligence in trading

Financial and IT markets are dynamic, so machines need improvement as well. For the last 10 years, many companies have offered traders AI-based tools and services but half of them are usually shut down in two or three years. Let’s name some interesting projects that still operate in 2021.

Buzz AI-based follower

The Canadian company BUZZ Indexes collects big data in social networks and topic-related mass media and blogs about trading and investment. Then its AI system filters that data and interprets the opinions of the Internet community about often mentioned stocks: positive, negative, or neutral. Based on 15 million data points, it builds an analytical model according to investor sentiment.

Each month, the company forms the BUZZ NextGen AI US Sentiment Leaders Index licensed by Van Eck Associates Corporation. It is a sentiment index of 75 popular stocks of big-cap companies.

AI employed for quantitative trading at Walnut

The French company Walnut Investments operates in the field of quantitative investment management using AI in their strategies with absolute return. It combines financial expertise with machine learning to create profitable self-learning trading systems. Walnut builds trading strategies the following way:

• Machine learning — for precise, quick, and continuous analysis of large data volumes. It provides thoroughness when assessing and controlling statistical reliability without strict rules.
• Quantitative investing — for trading based on balanced decisions without regard to intuition and chances. Thus, statistical significance is increased, due to the consistent application of mathematics and market underpinning.
• Diversification — for reducing risks, especially when there is a correlation during a crisis. It helps to secure positions and control the level of risk exposure.

Walnut also has Singularity, which is a program for short-term futures contracts with systematic management. It uses machine learning to detect long and short impulses as well as mean reversion signals. Futures trading is an intra-day activity: positions are not secured overnight.

For those who plan on developing or getting a similar project: from June 2016 till July 2017, Walnut Investments SASU raised venture capital investment and received a grant for a total of €1.5 million.

AI for managing the hedge fund Numerai

The Californian hedge fund Numerai uses AI to manage long- and short-term investments in stocks. As conceived by developers, it’s set to become a global crowdsourced hedge fund capable of making predictions for capital markets. Currently, it is supported by a network of anonymous data scientists. They have already created and trained 2774 models based on data refined and organised by the hedge fund.

Numerai was the world’s first hedge fund with AI that issued its own cryptocurrency — Numeraire (NMR) token. The company conducts a weekly tournament for data scientists who submit their predictions and then rewards them with native tokens or ethers (ETH). Based on the best predictions, the hedge fund builds trading strategies to profit itself and make a profit for its investors.

For those who plan on developing or getting a similar project: from April 2016 till June 2020, Numerai GP, LLC raised venture capital investment for a total of $21.5 million; in August 2021, the NMR price was $45.

Artificial intelligence on the exchange in comparison with a human trader

Today, with the spread of AI, the operations of traditional traders account for 10% of the total trading volume, while back in 2012, their operations in the United States accounted for 55%. At the same time, more than 2000 hedge funds (and there are about 11,000 hedge funds in total) use AI in the development of most trading strategies. This is a high indicator of using artificial intelligence in trading and investment activities.

Multitasking

Trained machines are capable of processing countless amounts of data in a matter of minutes. In the same manner, they can find and process historical data and repetitive patterns for smart trading that are often hidden, inaccessible, or not obvious to humans.

Traders are unable to process that much data or even detect it. For example, when it comes to high-frequency trading, some people use AI to decode over 250 million different data points from the New York Stock Exchange in the first hour after the opening of trades.

Trade speed

Although AI is not a revolutionary technology, it significantly speeds up trading operations. Today, every millisecond counts. If you are a broker or an investment fund manager, then AI will make your job easier: clients will not need to call and place orders because trading will be automated.

Sentiment analysis and prediction

By studying the headlines of articles, news, posts on social networks, blogs, and other thematic sources, AI can predict the movement of stock prices and the possible actions of other traders. It conducts sentiment analysis — the process of categorising the opinions (or sentiments) that people actively share on the Internet.

Self-training

AI is not perfect at the outset, but it is capable of improving its skills. It will learn from its own mistakes and constantly improve. For these aims, there are automated trading advisers, with the help of which AI works to improve performance, not only by fine-tuning the existing data but also by adding and analysing new.

Modernising trading with artificial intelligence

Many traders and investors are confident that in the near future, the use of artificial intelligence on exchanges by companies will spread everywhere. AI-based systems are easy to use, they operate transparently and trade accurately.

Artificial intelligence has already begun to change the rules of trading. Take advantage of its benefits and modernise your brokerage firm or investment fund. It will help your clients to rationally invest and manage assets wisely.

If you want to use AI for trading or managing investments, then Polygant is ready to develop the necessary solutions for you. To find out the cost of work and development time frames, as well as to receive detailed information about the services, fill out the short form, and we will immediately contact you.