Speed and accuracy in processing real-time data are key requisites for interpreting the volatile cryptocurrency market, said Won Hyuk, chief technical officer at BlockSquare Seoul.
The crypto trading assistance company has created a charting tool for investors, which helps with trading analysis. Its AI-powered Goya algorithm suggests the best time to buy and sell, reading into the market price fluctuation of cryptocurrencies.
“We convert figures from orderbook, on-chain data and liquidation attempts into scores through calculation, having the algorithm figure out when would be the right time to place orders,” Won said in a recent interview held with The Korea Herald.
“Plus points are given when asset holders maintain a ‘long’ position, minus points are given when liquidations happen on a ‘short’ position,” Won explained.
According to Won, the Goya charting tool has shown a high accuracy rate in its suggestion of trading signals.
“We recently compiled the data which shows the tool’s accuracy rate stands at between 80 to 90 percent for top-traded coins such as Bitcoin, Ethereum and Ripple. The figure for other cryptocurrencies is expected to be somewhere in the 80 to 90 percent range, too,” Won said.
The most difficult part of designing the algorithm for the charting tool was pulling the data from cryptocurrency exchanges. Though the data was accessible as on-chain transactions are carried out on a blockchain network, the challenge lay in figuring out the right speed with which to attain it.
"For instance, when we take data from Binance at a speed that is too fast, we are considered as a cyberattack and the exchange blocks out our Internet Protocol address. But if we bring the data at a much slower pace, it would not be real-time data anymore because of the time gap,” Won said.
“Through many failures, we have figured out an optimal pace to bring in the data,” he explained.
Adding to the charting tool service, the company plans to launch an automated trading bot next year. Though it currently offers automated trading for buying cryptocurrency, it also plans to expand the service to selling.
“There are many trading autobot services out there. But from my understanding, most of the services place buying or selling orders for customers one by one. Then, there would be a gap between the first and last transaction, leading to a price gap,” he said.
“But we have worked out a way to put in the signals all at once. That was not easy to figure out, but we managed.”
Though the algorithm is powered by artificial intelligence, a certain amount of human interference is inevitable as AI can only be trained to perform specific tasks for the time being.
"We made the AI algorithm think like a human being. Because in the end, investments are powered by human decisions. But as the algorithm is run mechanically, it can perform ideal decision-making, unlike humans who are often swayed by emotions,” Won explained.
One of the ways that the company’s autobot can halve risks is by liquidating assets when volatility in a cryptocurrency increases.
“By putting in a code for liquidation on the cut-line, the autobot can automatically minimize losses,” Won said, adding the service in its test trial period has been seeing an early 10 percent profit on average.
Won believes tech can only excel when data processing is continuously improved. It is not just about building the system but keeping things updated.
“Though there have been many companies offering a similar solution to ours using scores and charts, most have disappeared. This shows maintaining the service through updating data and improving accuracy matters more than creating the technology itself,” he said.
“Tech will only further develop as time passes. I hope to launch more services that can provide convenience and improve profitability in crypto trading for investors,” he said.
