Machines have already conquered many aspects of finance, from prediction markets to high frequency trading. Could algorithms also do a better job of picking and trading volatile cryptocurrencies, the way some software already surpasses experienced stock traders ? So far, experts remain skeptical.
Nvidia deep learning consultant Michelle Gill told International Business Times current machine learning tools aren’t a great way to predict cryptocurrency markets because this new phenomena is full of infrequent events. AI is only as good as the data sets we feed it.
Comparisons to past economic bubbles, such as tulip mania in 17th century Holland, are insufficient because a flower is always flower. On the other hand, the bitcoin blockchain continues to evolve. Both the asset itself and the broader marketplace are dynamic. “We don’t have very many [comparable] examples of it,” Gill said at the Artificial Intelligence & Data Science conference in New York City. “Machine models are better at recognizing patterns.”
Regardless, several groups of researchers are now collecting and analyzing cryptocurrency data as tokens as proliferate. Among them are Jim Liew, assistant professor of finance at John Hopkins Carey Business School and amateur Ethereum miner. The DIY mining equipment he set up with some students cost $3,000 to build and requires $80 of electricity a month to mine around $300 worth of ether tokens. He sees it as a long-term investment.
“Most people look at the prices and what they are really missing out on is the blockchains,” Liew told IBT. “I think of bitcoin as AOL...Who is going to be the real Facebook, Google and Amazon? We don’t know that yet.” Liew’s research with graduate students focuses on rising alternative coins. They are training models to find patterns that might help predict which newcomers will be the next token to break into the top 10. Right now the leading cryptocurrencies, according to CoinMarketCap, include bitcoin, ether, monero, litecoin, dash, XRP and bitcoin gold, just to name a few.
“I think it will be a cryptocurrency that allows people to put their ideas inside, kind of like Ethereum,” Liew said. “It’s got to provide utility for the user.” One likeminded app available beyond academia, the Bitcoin Bubble Burst, tracks trading data and news coverage such as national bitcoin bans or new tax proposals. BBB’s software analyzes this data to automate price alerts. TechCrunch reported the app is relatively accurate so far. But many fintech experts are skeptical about any AI software's ability to predict the next bitcoin bubble burst or even reliable cryptocurrency trading strategies.
“There’s not enough data,” said Vasant Dhar, the New York University professor who founded one the first machine-learning hedge funds, SCT Capital Management ’s $350 million Adaptive Quant Trading program. “All we’ve seen so far as a strategy is to buy...once we start trading bitcoin futures that will change the market.” CME Group Inc. and Cboe Futures Exchange are among several mainstream companies set to start offering bitcoin futures over the next few weeks. Blockchain-based derivatives are expected to change the game entirely.
“At the moment there’s a buying frenzy because people expect it to go up,” Dhar told IBT. “When the futures are on track and professionals start trading it, several years after that maybe we’ll have enough data to come up with machine-learning based strategies...right now all we’ve seen is one little period where it’s moonshot. There’s no real basis for building a strategy.”
Editor's note: Newsweek Media Group and International Business Times partnered with Structure to host this week's Artificial Intelligence & Data Science event.