Can Twitter Predict Crypto Prices?

During the volatile period of surging and subsequently collapsing cryptocurrencies in the late 2010s and early 2020s, researchers from Yale University closely observed the meteoric rise and rapid demise of numerous crypto coins. These digital currencies would often experience a brief stint of attention on social media, capturing fleeting public interest and generating wealth for a select few before abruptly vanishing without a trace.

The researchers turned their attention to Twitter, seeking to monitor the conversations surrounding emerging coins. By employing a meticulous approach to gauging these discussions, they discovered a method to identify coins with promising prospects over the following month.

“The fundamental principle with these coins is that long-term investment is not the objective,” the authors explain. “The aim is to seize the moment and then promptly exit. This is the guiding principle for trading cryptocurrencies at large.”

Value from the noise

Their findings were achieved through the development of an innovative technique for extracting meaningful information from the hype surrounding cryptocurrencies. Previous attempts had relied on quantifying the sheer volume of tweets pertaining to a particular coin, assuming that a high volume of tweets correlated with favorable future performance.

However, due to limitations imposed by Twitter’s data access restrictions, analyzing the raw volume became impractical, as it proved challenging to obtain a representative sample from the millions of tweets generated monthly.

An alternative approach explored the predictive potential of sentiment analysis, aiming to ascertain whether discussions surrounding a given topic were positive or negative. However, certain colloquial expressions commonly used within the crypto community, such as #buythedip or #hodl, which signify positive sentiment, proved elusive for machine learning algorithms to capture accurately. Similarly, memes conveying sentiment in either direction presented challenges for sentiment analysis.

Measuring engagement

Instead, the researchers devised an “engagement coefficient” that took into account the number of followers of the Twitter accounts posting tweets mentioning a specific cryptocurrency, as well as the frequency of likes and retweets garnered by each tweet.

These two metrics were combined to yield a single value ranging from zero to one, serving as an indicator of the level of discourse and attention surrounding a particular cryptocurrency over the course of a month. Employing this indicator, the authors monitored the mentions of 48 cryptocurrencies that entered the market between 2019 and 2021, making hypothetical month-long investments. These simulated investments yielded a return of nearly 200%.

It was unsurprising that the researchers discovered that if the engagement coefficient of a particular coin remains below a certain threshold, it signifies that purchasing that coin would not be worthwhile.

The right mix

However, the authors point out that excessive buzz is also an unfavorable indication. Such high coefficients appeared to be indicative of numerous automated accounts artificially inflating interest in the coin, hinting at a potential pump-and-dump scheme where individuals artificially boost buyer interest before the coin experiences a crash. There existed, he explains, a sweet spot where investment decisions became reasonable.

This valuable insight holds potential significance for regulatory authorities aiming to curb fraudulent activities. If, shortly after a cryptocurrency is listed on a crypto exchange, it generates an abnormally substantial amount of buzz, it could serve as a warning sign, suggesting possible manipulation of the coin.

“This gives you a way to take the overall temperature of a topic by sampling a relatively small amount of data, and it seems to predict success really well,” the researchers conclude. “With just a couple thousand tweets you can look at a crypto coin, or a movie, perhaps a new brand or product or politician.”

Facebooktwitterredditpinterestlinkedinmail