Using RavenPack News Analytics’ version 4.0 data, which tracks the entity Bitcoin, we were able to test if media attention and sentiment would help explain bitcoin price movements over the last three years.
We created a Bitcoin media attention index, weighted by the relevance score of each mention, and a sentiment index focusing on when the media reported large price movements up or down to see if that could impact future prices. We tested our data against the 24-hour average daily Bitcoin prices from Quandl.com (a great resource, by the way). Their Bitcoin price index looks at an aggregate across multiple exchanges and is updated daily at 6:00 pm EST.
We ended up creating a dynamic model using the previous 6 months’ prices as a training set, and the previous 30-day lagged values of Bitcoin returns, media attention and sentiment as predictor variables.
We found that media attention and sentiment did in fact improve on a model simply considering past returns. Not only where we able to improve the risk-return trade-off (Information Ratio) of a simple buy-&-hold strategy from 0.99 to about 1.85, we were also able to reduce the max. drawdown by more than 25%.
As you can see from the chart, had we been tracking media attention and sentiment in the news, we would have participated in the amazing price ride of Bitcoin and would have been able to do so at much lower risk than in a simple buy-&-hold. (see Figure 1 below).