The StatsTrade Insights are the product of our proprietary services, relying on our core research framework. They provide daily insights into the stock markets, but also into human behavior in general.
Some of the Insights are related to a specific stock or company. The aggregated analysis results per stock or company are processed by our beloved Robot-Analysts to finally generate some “predictions” about the behavior of that stock. As of 2012, Nate‘s results are being distributed freely (why?) via an experimental Free Newsletter, while the other Insights are only shared with a selected group of Partners.
Insights goal and data
Our core research framework was — and still is — designed with purely research-oriented agenda in mind. We study the way people behave in the markets, to deduce insights on the markets and people behavior in general. To do so, the framework consolidates data of various types:
Technical data: By far the most commonly used data type by advanced traders or automated trading systems, due to its accessibility and ease of use, even without understanding the associated hidden factors (or, for that matter, mathematics). Data types that fall into this category include: Stocks closing prices (either intraday or on daily basis), volumes, interest rates, etc.
There is also a variety of “indicators” that crunch these technical numbers in a very specific way. A crunched technical still falls into the technical data category. The well-known MAs or MACD are the simplest, but even an Artificial Neural Network or other statistical learner is not that different, and therefore we may also consider it as a naive approach (yes, we have just referred to some uses of Artificial Neural Networks as naive approaches. And yes, we know that a lot of organizations use this buzzword (!) to impress their clients).
Fundamental data: The Fundamental Analysis approach attempts to evaluate the intrinsic value of a company, regardless of the technical data. The fundamental analyst tries to determine which stock is overvalued or undervalued and act accordingly (sell or buy, respectively). Data that falls into this category includes, for example: The company’s revenues in the last quarter, future estimates for the general demand in the company’s associated industry etc.
We at StatsTrade view Fundamental Analysis as one particular case of a more general approach — the study of hidden factors that influence the markets. Namely, the Fundamental Analysis “merely” focuses on evaluating a few very specific hidden factors related to the company’s chances to succeed in the future. Some example factors may be: The management skills, the efficiency of the sales cycle or administration, etc.
External data: There’s much more data available outside of those boxes. Outside of the traditional financial world. Naturally, we will not list here the specific data types we use. Instead, we will illustrate our point of view on the subject via the two examples:
- There’s a belief that bad weather affects an investor’s mood, and hence her buy/sell actions or at least the volume of the trade. There are countless studies on the matter, mainly because it’s so refreshing and “cool” read, but (sadly) very few, if at all, are rigorous in understanding the exact motives and the relations between the hidden factors in discussion. In addition, the research methodology used is usually very weak (e.g. no control group).
- Most “stocks analysis systems” are focused on analyzing existing data. It seems like it never occurs to those analysts that they are allowed to generate new and valuable data of their own. It did occur to us. Even the simplest survey to a pilot group of users and other controlled audiences sometimes hold priceless information, as long as it is carefully designed, with strong research foundations.
A large part of our framework deals with organizing the data, modeling the exact semantic relations, validating its integrity and other important pre-processing tasks. The more important part of the framework analyzes millions of entities and relations, evaluates a lot of strategies (many of which are computer-generated, based on previous insights and existing strategies) and after a serious crunching of the data comes up with a lot of analysis results, or as we sometimes call them, Insights:
Some insights are simple, e.g. “Given the assumptions, the volume of the trade in GOOG in the next trading day is more likely to be higher than average”, or “Given the assumptions, the price of GOOG is more likely to increase than decrease 10 days from now” (actually, our system never produces such vague sentences, and instead generates probability distributions for the events in discussion. However, for the sake of clarity we omit many technical details)
Other insights are less straightforward to exploit on their own, e.g. “Given the assumptions, in the past month there seemed to be a stronger correlation between the general sentiment in the leading news sites and the trading volumes in the Technology industry”. Such insights are primarily concerned with with hidden factors that affect the market, and with the relationships between such factors. Only by combining many such partial insights over time one may generate concrete trading decisions.
Some insights are purely abstract and are concerned with relations between mathematical entities, that have no straightforward interpretation in real world concepts. Such insights are of use only to the research framework who produced them.
Noone can predict the future. What one may do is assume the existence of certain factors that persist today and in the future, learn or guess the relations between such factors and predict the likelihood of specific result. Most of the systems that generate such “predictions” are usually very naive. Most of them rely on Technical Analysis, and some incorporate other “cool” data types such as text analysis of news articles or twitter feeds, to get a feeling of people’s sentiment in general or in relation to some entity. In contrast, our technological core was primarily designed with a different goal in mind — to facilitate studies on people’s behavior in the Financial markets, a pure research-oriented goal, with a focus on the hidden factors and the relations between them. In some sense, our proprietary Insights services are “merely” a byproduct of the original technological core.
Sharing our Insights
As explained above, the original design goal of our core technology was and still is purely research-oriented, aimed to study people and their behavior and reveal the many hidden factors that control their actions, and the relations between such factors.
The “bottom line”, as we see it, is not some transient trading decisions, but the ongoing contribution to the research framework itself.
However, an important byproduct of the hard work constantly put into the research framework gives its immediate fruits in the form of a lot of analysis results. These can be shared with other researchers, or with advanced investors who are eager to put their hands on this valuable data stream.
As of August 2011, the Insights service is still not open to the general public, but only to a very limited number of partners.