Their goal is to create an algorithm that can find price "signals" from the vast amounts of information on sites such as Reddit, WeChat, Twitter, and Telegram.
- Bitcoin broke through $5,100, and the decline in trading volume is worthy of vigilance
- Coindesk Senior Analyst: The dollar is coming to an end, bitcoin or become one of the global reserve currencies?
- Who is the information of the user who sells the coin? What have the leaked information been taken?
- Denial of service (DoS) attack: Chengye miner, defeated miner
- John McAfee: I know who is Nakamoto.
- 180 degree turn! The godfather of emerging market investment: If bitcoin goes up again, I will go buy it.
Many investors have begun to use computer models to identify the spreads of hundreds of cryptocurrency exchanges.
Through interviews with six hedge funds and asset management companies and three software developers, Reuters found that as this emerging industry develops, arbitrage opportunities are diminishing, and large companies are increasingly seeking to manufacture or buy more complex Robots to discover the clues that affect the market online.
However, while more and more people are beginning to use algorithms that analyze social media, some interviewees say that from cost to complexity, broader deployments still face significant challenges and risks.
Bin Ren, CEO of digital asset management firm Elwood, said:
"For fund managers, this is an arms race."
“Few people can create such an algorithm, but I believe it is very profitable.”
In traditional markets such as stocks and foreign exchange, this “sentiment analysis” (the computer's interpretation of social media sentiment) is a tool that uses consumer perceptions of a company or asset.
But in the cryptocurrency market, this may have more significance. In the cryptocurrency market, there is almost no authoritative source of information like the central bank, and there is hardly any reliable data to measure the value of assets such as economic indicators and financial statements, and the proportion of individual investors is high.
Encryption technology is still in its early stages in the field of cryptography, lacking performance data from industry dimensions. There are also many problems with the validity of such data. The agencies interviewed by Reuters are reluctant to disclose the details of their algorithm performance, saying it is due to trade secrets.
Extract clues from massive information
The trend of digital currencies does have some common drivers with traditional markets, such as those of policy makers. Bitcoin is particularly sensitive to regulators' comments: Last week, bitcoin fell sharply after the Fed chairman called for a halt to the Libra cryptocurrency project in Facebook.
But considering that cryptocurrencies have been intertwined with the Internet since they were born 10 years ago, when they first spread in forums and chat rooms, it seems meaningful to search for price triggers on the Internet.
Still, designing an algorithm that finds signals that influence market trends in the noisy social media world, analyzing a large number of posts in dozens of languages, and filtering out unreliable information is far from easy.
Andrea Leccese, president of New York-based investment firm Bluesky Capital, said that an upfront cost of a robot that can only read English tweets ranges from $500,000 to $1 million, most of which is spent on developers. He said that this is why Bluesky does not use this technology.
Another challenge is the number of social media channels. In addition to Twitter, websites often used by cryptocurrency enthusiasts include Telegram and Reddit, the former is a messaging app with a public channel, and the latter is a forum.
In Asia, many retail investors are using apps such as Line and Kakao, which are popular in Japan and South Korea.
Tens of thousands of comments on cryptocurrencies are coming out at all times.
Only Reddit's bitcoin sector has 1.1 million subscribers. According to the BitInfoCharts website, Twitter has tens of thousands of bitcoin content every day, ranging from 14,000 to 32,000 in the past three months.
To extract useful parts from this confusing piece of information, the algorithm uses natural language processing—identifying keywords and emotions—to show how social media users' perceptions of certain digital currencies have changed.
Investors who use algorithms say they can also learn what information to send to get attention online.
Elwood's Ren said:
"The information is not spread randomly, but spread through a very clear structure, like a tree. This is very similar to the spread of simulated viruses."
Fear dominated by false news
Other investors have emphasized the challenge of church machines identifying biased or inaccurate information.
The project party can even exchange positive feedback on social media.
BitSpread CEO Cedric Jeanson said BitSpread used its own funds to trade using algorithms that were developed about a year ago. BitSpread is a cryptocurrency asset management company based in London and Singapore.
However, the software collects limited information, collects only tweet feedback, and looks for information about clearing or closing of exchange positions.
“The problem is collecting all the information, trying to understand who is trading, and what kind of liquidation will happen. This is a meaningful strategy.”
However, he admits that this strategy is flawed.
“The emotion itself, which we see on Twitter, may be really related to fake news. We are always very cautious about what we read in the news, because most of the time we have seen prejudice ""
Many algorithms use machine learning and should be improved through experience and better understand how social media posts translate into market trends.
According to Bijan Farsijani, a Berlin-based startup, Augmento, developers often find key people who speak frequently and have a lot of fans to get more weight in their algorithms.
He said that since the launch of the software, a number of hedge funds have purchased the software from his company.
Programmer, we need you
Bitcoin is the largest cryptocurrency market capitalization and a leader in the industry. This year, the price of Bitcoin has soared by more than 180%, driving interest from large investors such as trading companies to hedge funds.
Bitcoin's last rise was in the last month, which analysts believe is because the market expects Facebook Libra to be used more widely.
This move also reflects a surge in interest on the web. On June 18, when Facebook announced the news, Google’s cryptocurrency search reached its highest level in three months.
However, this is very much like the question of chicken and egg: Is the online discussion leading to price fluctuations or price fluctuations leading to a lively discussion on the Internet?
Leccese of Bluesky Capital said:
“There may be some value in sentiment analysis, but most of the time, the information that people post on Twitter may be a lagging indicator of price movements.”
“But there is still potential. In the next 5 to 10 years, people will start to pay more attention to this issue, because the traditional strategic competition will intensify and the rate of return will continue to decline.”
According to a report by PricewaterhouseCoopers, in the first quarter of this year, cryptocurrency funds using a variety of methods, from arbitrage to sentiment analysis, performed significantly better than long-term investment funds, despite the lack of data specific to this technology.
The programmers said they saw an increase in demand.
Taiwan-based Marc Howard works with more than 500 machine learning experts to integrate data from Google Trends, Reddit and the development platform GitHub into sentiment analysis algorithms.
Howard said that as of June 24, he used algorithms to make bitcoin investments, which is 54% higher than funds that only track prices. He added that funds in New York and Taipei have sought help from him to help them develop their own analytical tools.
“Now this form is very hot. Any competent fund is investing some resources and configuration for sentiment analysis.” (Babbit)