I just read an article in the most recent issue of Futures magazine (the May 2013 issue) called “New Approach in Analyzing Forex Markets“, by Abe Cofnas. The basic premise of the article is that traditional types of analysis aren’t serving us well; both types of analysis are weakening.
What are those two types of analysis? We have Technical Analysis (or “chartists”, as he calls us), and then Fundamental Analysis. Technical analysts use charts, indicators, support and resistance lines, and price movement to determine the movement of the markets; fundamental analysts look at company profit and earnings ratios, government decisions, major banking institutions and world economies to determine market movement.
As programmers, we are mostly into technical analysis; almost every strategy I have seen uses some sort of technical tool to determine entries and exits.
So, the new third type of market analysis is called “sentiment analysis”.
From the article: “Sentiment analysts believe that each week the market resets expectations and fears. Technical patterns of previous weeks become discounted as new concerns and upcoming event risks increase in their relevance. As each week begins, a new re-balance of fears holds the key to more accurate prediction”
What else do we know about sentiment analysis? Well, the traditional sentiment indicators are familiar ones: the VIX (volatility index), the Commitment of Traders reports, put/call ratios, and various other bullish and bearish indicators. All of these give us some indication where the markets are moving, where the big money is, where the “smart money” is.
What really caught my attention in the article were these kinds of sentiment analysis: text mining, word clouds and headline analysis. I did a bit of Googling to see what’s out there, and there isn’t much out there for traders. There are some services and software tools out there, but they seem to focus on customer service sentiment and reputation management. Not much for analyzing the markets. That means this idea is wide open for us as programmers.
What would a sentiment analysis tool look like for an automated trader? How about this: Your program first does a Google news search for the Euro and grabs all of the headlines from Google News. Then using a semantic text analysis engine, it determines if each headline is hopeful, fearful or neutral on the Euro. Maybe it could weight the headlines in terms of financial relevance. It then give you an up, down or flat indicator, which of course you use in your FOREX market entry to buy or sell.
In the semantic analysis tools and APIs that I found, keywords and phrase searching seemed to be the way things were going; searching for keywords and phrases in headlines wouldn’t be too hard: “Greek debt crisis”, “Euro failure” and “weakening of the yen” “sell-off” would be great terms to search for.
Or how about this for another idea: Your program searches twitter hash tags for tags related to the Yen. Again, using some semantic analysis, it would try to determine if the tweets were fearful or hopeful, and would then give you an up or down indicator.
This doesn’t sound like rocket science. I bet if you combine technical indicators with sentiment analysis, you’ll have the best of both worlds, especially if you’re on the daily charts or longer time periods. If this is the next big thing in trading, we’ve got some cool stuff ahead of us as programmers; this is seriously cutting edge stuff.
The article ends with this: “Predicting price direction is predicting where traders are going. Fading or following that crowd is the ultimate decision challenge for the traders…This will become a major third way of understanding markets.”
The Futures Magazine article is here. Abe Cofnas also has some books available: The Forex Trading Course: A Self-Study Guide To Becoming a Successful Currency Trader, Sentiment Indicators – Renko, Price Break, Kagi, Point and Figure: What They Are and How to Use Them to Trade and Trading Binary Options: Strategies and Tactics