Three scholars affiliated with Johns Hopkins have evaluated cryptocurrency investing, in a new paper available at SSRN, and have concluded that “near-term cryptocurrency markets are semi-strong form efficient.”

That bit of finance theorist jargon means that all publicly available information gets discounted quickly into an asset’s price, so that neither fundamental nor technical analysis can reliably produce alpha. In an ideal, or strong form efficient market, even private or “insider” information gets discounted into an asset’s price, so that insider trading (even aside from the question of discovery and punishment) won’t produce alpha either. The Johns Hopkins scholars,  Jim Kyung-Soo Liew et al, don’t make the strong claim, but they do make the semi-strong claim for cryptos, and that still represents a remarkable fact about those markets if their claim holds up in the face of continuing research.

The research that reached this result involved rolling prediction analysis run on 57 cryptocurrencies with 11 artificial-intelligence algorithms. Although some information sets, employed by some families of algorithms, did better than others, the general result was that “predictability may be difficult,” and day traders hoping to make a quick buck face grave challenges.

Literature Review

The paper includes a literature review. One 2014 study, published by Isaac Madan and associates at the University of Stanford, used three machine-learning algorithms to try to predict bitcoin prices. They confined themselves to predicting the sign of the coming day’s price move (that is, whether it would be up or down). With this binary focus, their algorithms were very accurate, at best 98.7% accurate. The following year another group of Stanford researchers sought to predict the hourly change, up or down. Their accuracy rate was considerably less impressive: only 55%. That’s a huge difference: near perfection on the day-to-day predictions, neat coin-flip results on the hour-to-hour predictions.