Can tweets help you trade smarter? Ting Li, professor of Digital Business, from RSM Rotterdam School of Management, Erasmus University, reckons social media can boost your trading success.
Her team – including colleague Dr Jan van Dalen and Pieter Jan van Rees, who pioneered the research as part of his thesis – scrutinised more than 1m tweets. Each contained stock mentions from the S&P 100. They then developed an algorithim that broke down buying sentiment, be it buy, sell or hold.
How they did it:
- Borrowing computational linguistics methods they went through an intense process of language detection and slang removal to boost their analysis accuracy
- They then simulated a set of trading strategies using microblog features. Data was analysed on a daily and a 15-minute basis
- Underpinning these methods was a belief that stock microblogs generate, share and spread information virally may have a contagion effect. “Disagreement in microblog messages positively influences stock features, both in interday and intraday analysis,” Li says
The ostensibly interesting bit is how this fintech sniffed out tweeted sentiment – a lot of ear-wigging on America’s biggest companies, from Apple to GE to Wells Fargo – with real-time follow-on price fluctuations (though trading volumes, which are also affected by ‘bullishness’, weren’t judged).
Noise = movement = edge
The conclusion drawn? When an S&P 100 stock is constantly tweeted – even if there’s a mass of opposing argument about its prospects – the brute amount of noise around it means it performs better. Essentially agreement levels are being measured here.
“In the simulation we showed that if you invested money in the S&P 100 and used the information gleaned from twitter using our algorithm, you would beat the market.”
Li says you could invest in one company, sell at the end of each day and reinvest the next. Or you could trade every other day or every three, four or five days. “Even when you take transaction costs into consideration, you’d still come out ahead.”
In other words, tweets around stocks don’t just build ignorance or insight (depending on the quality of the information). They supply a performance kicker.
“It can be concluded,” Li’s team report said, “that leveraging the information of expert users who have a larger follower-ship, have higher levels of mentions…can be used to amplify the relationship between bullishness and trading volume or volatility.”
Pick your venue
Perhaps. But this modelling has taken place during a very long momentum bull run. Traditionally, stock prices moved on real news. Not micro-blog clamour where the winner is the one who shouts loudest for longest, even if a good measure of consensus emerges.