The effect of online herd mentality on cryptocurrency prices
$$$$$$$
The problem
•Unlike other asset classes, the price swings of cryptocurrency markets don’t always rely on their fundamentals.
•The characteristics of the coins lend themselves to let investor sentiment play a big role on price movements.
•This project aimed to use Reddit conversation as a proxy for investor sentiment in the cryptocurrency market.
The data
Three different datasets were used:
Reddit Data (Scraped)
Data examined ranged from 7/30/15-10/17/17.
Over 8k posts analyzed
Bitcoin (BTC) historical prices (Kaggle)
Data examined ranged from 7/30/15-10/17/17
Ethereum (ETH) historical prices (Kaggle)
Data examined ranged from 7/30/15-10/17/17
Reddit Data
•Initially only the Ethereum and Bitcoin subreddits were to be scraped, however Reddit only allows to scrape the last 1,000 posts. Therefore 11 of the largest relevant subreddits were scraped.
- After cleaning, the Reddit data was as follows:
- Posts
- Number of posts that day.
- Comments
- Number of comments on the posts that day.
- Score
- Average number of ‘upvotes’ per post that day.
- Posts
Correlations
•Unsurprisingly, Posts and Comments were highly correlated (0.88) as were the prices of both coins (0.94).
•The one correlation that was interesting was Score and BTC Price at 0.73. There was also a positive relationship with ETH Price but weaker at 0.62
The variables did not have a normal distribution so we couldn't use OLS. So we began exploring transformations
Conclusion
•Average Reddit Score was a statistically determinant factor in both BTC and ETH Prices. For each increase in average score there was around a $130 increase in BTC Price and a $12 increase in ETH Price.
•This suggests that ‘lurker’ activity (i.e. people who do not post or comment but merely ‘upvote’) was a good predictor for changes in price in ETH and BTC.
Caveats
- Subreddits/ Scraping
- Noisy
- Mainly a US platform
- Hard to time
- Day data vs. Hour data
DEMOANALITICA