The Crypto-Mojo

In 1936 Keynes described how irrational emotions influence market pricing: “…our decisions to do something…can only be taken as a result of animal spirits…not as the outcome of a weighted average of quantitative benefits multiplied by quantitative probabilities”*. With hard to measure benefits and probabilities, animal spirits will therefore be a dominant component of crypto-asset pricing and, thereby, the solvency of its providers.

But how can we measure this market sentiment? And how have crypto-markets, in particular, performed over the past, tumultuous year? To address these questions, and to evaluate how different crypto-firms, like CEXs, Brokers and Hedge Funds, are performing, we used computational linguistics to measure the sentiment expressed in 100k crypto-news articles published on Cryptocompare.

Specifically, we used a machine learning technique, trained on the Wikipedia corpus by Google, called Bidirectional Encoder Representations from Transformers (BERT). BERT can both identify and classify objects as well as estimate the probability that a headline is either positive or negative. We then defined sentiment as the first minus the second. So a headline that is certainly positive scores +100% and vice versa.

‍

‍

The graphic shows the average headline sentiment each day along with the underlying trend. Intuitively, sentiment was high before falling in the New Year, recovering slightly and then falling heavily as Luna collapsed, taking down 3AC and then some of its lenders. Sentiment bottomed-out on 16th June, the day after Su Zhu’s Twitter post “We are in the process of communicating with relevant parties and fully committed to working this out”.

The red line shows Bitcoin’s value. For the first half of the year, sentiment led Bitcoin by about 1-2 days (i.e. the blue line is slightly to the left of the red). In other words, changes in market sentiment were influencing Bitcoin’s downstream valuation. This is in line with Keynes’ view that markets are driven by animal spirits and reminiscent of David Hirshleifer’s widely cited finding that daily stock returns are higher on sunny days**.

But June saw a phase change. Bitcoin’s price has remained flat while sentiment recovered. The markets have been relieved by the lack of further institutional failures. Moreover, post Merge, the crypto-ecosystem can again focus on future growth, all tempered by the on-going layoffs that have kept sentiment below its prior peaks. We’ll continue monitoring this metric. However, the divergence of sentiment from Bitcoin performance feels important.

_{*. Keynes, J. M. (1936). The General Theory of Employment, Interest and Money. London: Macmillan}

_{** Hirshleifer, D. & Shumway, T. (2003) . Good Day Sunshine: Stock Returns and the Weather. Journal of Finance, 58(3), 1009-1032.}

_‍