Sentiment trading has been around a long, long time – and will continue to be. In this post, we argue that, although extremely important and integral to trading, the market’s focus on polarity sentiment for short-term trading is a bit exaggerated and that other uses and new forms of sentiment will begin to augment it.
Sentiment directly impacts asset prices through somewhat predictable shifts and overshoots – anyone who has spent a considerable amount of time on a trading desk can attest. Though sentiment has been around, its quantification is rather ‘new’, and its quantification in real-time even more so.
Measurement began with polls and surveys. These were and are still great. The problem of course is that they lack a real-time element. Additionally, issues with respondent selection are apparent.
Social media and the new world of full transparency have opened up financial market analysis to more modern types of sentiment. Looking at tweets, cashtags, other micro-blog posts, news, and/or user generated content can greatly improve insights into sentiment.
Why wait for an end-of-day or even end-of-week survey to be released? Sentiment can be yours right now. It is quicker and can be measured with an impressive amount of scientific certitude. All of this makes perfect sense and is a large benefit of the advent of Big Data.
The issue that has appeared, however, is that positive / negative polarity natural language processing (“polarity NLP”) sentiment has become too central when discussing Big Data and finance. In fact, when you mention Big Data and investing to most people, they automatically think of two things:
Yes, these are important, but if your horizon ends there, you are frankly missing so much that the next generation of financial data and analysis has to offer. Furthermore, the rather clear impression is that going forward you will miss even more as Big Data continues to produce revolutionary insights.
Let’s state that sentiment derived from NLP polarity analysis is extremely important. It offers rapid interpretation of events and of market moods. It can also provide very interesting insights into market trends that would not be apparent otherwise. This post is not arguing to not use this type of sentiment, just that the market should better understand its limitations and be open to other new forms of data and even other measurements of sentiment.
To help portray these points, we will publish over the coming days two posts.
The first highlights a somewhat counter-intuitive use case for negative NLP polarity sentiment. To this point, the generally accepted interpretation is that an unusually large negative sentiment spike offers a great selling or short selling opportunity with potential follow through. In this case, the reverse is shown to be true in that after a few days of declines the stock bolted higher – thereby reinforcing the 360 degree approach that is required when interpreting any single variable. In other words, do not rely or focus on just one variable.
The second highlights a new form of sentiment and how it can be used. Specifically, it highlights (what we believe) is a first in directional sentiment analysis, which has proven both intuitive to understand and highly useful in its application.