Predicting stock prices using Google and Wikipedia

Tobias Preuss and colleagues from the University of Warwick Business School undertook to check, is it possible to track processes in the real world through the virtual world. Для этого они проанализировали открытые материалы Google Trends, as well as Google Hot Trends - public apps, showing, how often a specific search query is entered in relation to the total volume of queries in different regions of the world and in different languages.

DETAIL_PICTURE__13076451

It has already been noted before, that due to the sharply growing frequency of requests such as "flu treatment", such a tool can detect an influenza epidemic a little earlier, how the health departments of a particular country or region will "guess" about it. What about business?

Researchers tracked 98 search queries on financial and economic topics - "debt", "a crisis", "Derivatives"… — with 2004 on 2011 year. To detect the connection between such requests and events in the stock markets, we used an investment game simulator created by the authors.. When googling financial terms went down, the computer model “bought” stocks and opened long positions, and when all these "crises" of yours were "googled" often - the model, against, закрывала позиции, playing short in the hope of buying the same shares at a lower price soon. You know the logic: until the thunder breaks out, man does not cross, that is, curiosity about certain issues captures the masses just then, when there is a massive failure in their relationship.

Simulation worked: as soon as the search for words of this kind came to life, stocks were falling all over the market. The most reliable word for the United States was the word "debt". Tracking markets only on it, scientists have increased their hypothetical portfolio of securities by 326% in just seven years. While the usual strategy of this kind, simulated in parallel, gave no more 16% growth, that is, it only compensated for dollar inflation.

  Main portfolio

Important to consider, that markets are adaptable. И если каждый broker будет руководствоваться Google Trends, using the described strategy, it will bring less profit.

But, as noted by the authors of the study, there is also Wikipedia, also providing information on, how many people view certain articles hour by hour. This - so unexpectedly - makes the online encyclopedia another potential predictor of the behavior of stock markets and other phenomena of the so-called real world..

srep01684-f3

The success of the strategy depending on the choice of keywords, entered into Google Trends.

Scroll to Top