By Maria Lobanova
Cryptocurrency market whetted retail investors’ appetite for trading again, after they had been long disappointed by the sluggishness of traditional financial tools. It had all of it: no high barriers to entry, real cryptocurrency price roller coaster with steep daily price fluctuations, exciting prospects of profitability, and diverse investment portfolios. This is due to its spontaneity and unpredictability at glance, why the cryptocurrency market is best suited to become the basis for creating trade solutions based on artificial intelligence.
Volatility is both a friend and an foe to the participants in the cryptocurrency market. Its volatility attracts non-professional traders, who are fascinated by the prospects of high profitability, but the very same volatility reduces the flow of institutional investments in cryptocurrencies.
Forecasts of when the volatility of cryptocurrencies will decrease and turn them into a more stable means of payment range from “never” to “in the next 10 years”. According to some analysts, first steps in this direction were already made. Since early 2018, bitcoin volatility has dropped to its all-time lowest, as mentioned by Bill Baruch, broker and president of Blue Line Futures.
There is another technology that can possibly help the cryptomarket overcome excessive volatility and at the same time increase its liquidity – one that has already gained popularity in the traditional market – that is, artificial intelligence.
In order to understand how this can help cryptomarket overcome volatility, let us look into to the history of how AI has conquered traditional trading desks. The solutions for high-frequency trade (HFT) allowing for buying and selling securities in fractions of a second were first to become instrumental in the field.
Robots conquer Wall Street
Steve Swanson, a graduate of the Faculty of Applied Mathematics of the College of Charleston, was a pioneer in HFT algorithms. Together with the college teacher Jim Hawks and the teacher of finance at the University of Rutgers David Whitcomb he created first computer codes applicable to stock market. They were using prognostic formulas which could predict stock prices 30 to 60 seconds ahead. Everything has changed in 1989, the year marked with the birth of high-frequency trade in the US: the US Securities and Exchange Commission (SEC) allowed the usage of electronic solutions in trading. Swanson, Hawks and Whitcomb founded the Automated Trading Desk company, which was acquired by Citigroup in 2007.
As early as in 2010, algorithmic and high-frequency trade accounted for 60% to 70% of US exchange trading. By 2014, the share rose to 75%, and by 2017, according to JPMorgan, traditional traders accounted for mere 10% of trading volume.
Further progress in technology saw the emergence of AI-based artificial solutions capable of managing assets more effectively than human mind. According to Prequin research company statistics, by 2015, 9% of hedge funds (1360) with about $ 197 billion of assets under management were using computer models to perform trading operations.
In January 2016, the Hong Kong-based company Aidiya led by entrepreneur Ben Goirtsel launched a hedge fund that carried out all exchange transactions with American securities through the use of artificial intelligence only, without any human intervention.
“If we all die, it will still continue to trade,” – Goirtsel remarked ironically, describing his project. Aidiya Hedge Fund used several AI-based solutions, including one based on probabilistic logic. Every day, the fund’s system was analyzing a huge body of data – from market prices and trading volumes to macroeconomic data and corporate financial reports – and issuing its own market forecasts based on it. The results of this approach were staggering: according to Herzell, on the very first day the fund received a profit of 2%.
In October 2017, Wall Street saw the launch of the first fully AI-powered equity traded fund (ETF). During the first week of its operation, it went up by 1%, thus bettering the S & P 500 index, whereas by August 2018 its shares went up by 20%. ETF operates on the basis of IBM Watson, a supercomputer capable of analyzing the news backgrounds of 6,000 American companies (from financial statements to social network records) on a daily basis. At the same time, Watson is constantly learning – it analyzes every decision it makes. Should a transaction turn out to be unprofitable, the algorithm will act differently to get the desired result next time.
But the arrival of machines at Wall Street led to another result: markets became less volatile, as they were dominated by robots that left no room for emotional and irrational solutions. This is confirmed by research data provided by the scientists from the University of Erlangen-Nuremberg, who developed algorithms involving historical market data processing.
While reproducing the model of market development, they found a quantitative algorithm that allowed for 73% annula returns between 1992 and 2015. Though during the financial crises in 2000 and 2008 the algorithm showed an incredible profitability (545% and 681%, respectively), in general, scientists found that the profit from AI algorithms decreased after 2001. They came to the conclusion that this is due to the dominance of robots in the auction, with algorithms competing with each other for greater profitability.
It is this property of algorithms and AI solutions that can change the character of the cryptocurrency market.
Artificial Intelligence will change the crypto market
High emotionality of the participants of the cryptomarket has already become an object of study among developers who are trying to create AI-based solutions to maximize the profitability of trades. One of the first steps in this area was the creation of a bot that uses a neural network to predict cryptocurrency price rates.
The NeuroBot constantly analyzes prices through utilizing a set of crypto instruments, applies patterns from traditional technical analysis, including Fibonacci lines and Elliott waves, and considers signal indicators. According to the creators of the bot, it allows you to make forecasts for 24, 48 hours and a week with an accuracy of 70% to 90%.
NeuroBot offers a paid subscription service. The fee varies from $ 3 to $ 10 per month, depending on the trading pair and the cryptocurrency. Payment is accepted in the ETH. But according to users’ reviews, NeuroBot’s forecasts are too unstable, as they may change drastically in a matter of hours. This is because the bot is too susceptible to fluctuations in the crypto market, which leads to the issuance of contradictory forecasts.
Bitstox cryptocurrency exchange developed a more customized AI solution for traders. Based on NEM blockchain, the platform offers its users a unique service: the AI-assistant D.A.N.N.I. (Decentralized Artificial Neural Network Integration). D.A.N.N.I. analyzes the strategies of professional crypto traders, historical exchange rate fluctuations, and uses this data to provide forecasts and advice to non-professional players in the market. In addition, it analyzes the mood in the market through news feeds. This is a personal assistant-robot, which helps traders to make correct trading decisions and improve the quality of risk management. AI is not exposed to emotions and fears of losing profits. It is able to process and analyze huge data sets and help traders make the right decisions.
As long as more and more machine learning-based solutions keep on emerging in the field, the crypto market will grow and drift towards greater stability and away from high volatility, just as it was in case with traditional financial assets. However, if in the case with the traditional market machine learning and AI-based solutions are designed for large funds only, when it comes to cryptocurrencies, applications of such sort are available to all market participants.