Time Crunch Trading

Technical analysis and fundamental analysis can be viewed as separate and distinct ways at looking at the markets. They are. Quantitative analysis, oftentimes operates in conjunction with one or both of the other two. You can apply quantitative methods to fundamental data, and you can do the same to technical data.

Quantitative Analysis, for the sake of this discussion, is the use of mathematical and/or statistical methods applied to market data. This is primarily done for one of two reasons: 1) To compare two or more markets or securities; 2) To develop a probability-based view of market behavior.

In this part of the chapter, we will go through some of the methods employed by the quantitative analyst and see how you can apply them to trading.

Comparison

The first quantitative application we want to explore is market and/or instrument comparison. The easiest way to describe this approach is to use an example.

Investors Business Daily (IBD) publishes two figures in its stock tables (among other things). They are the rankings for Earnings Per Share (EPS) and Relative Strength (RS). Even if you have never seen an issue of IBD, or heard of the paper, you could still be familiar with EPS and RS as many stock broker screening systems have variations of them included.

 
Source: www.investors.com

In brief, the EPS rank is an assessment of all companies in terms of their rate of earnings per share growth over a given time frame (3-5 years normally). The companies are arranged in order of their growth rates and ranked. In the IBD version the ranking is done on a percentile basis such that the top 1% of all companies would get a 99 (99th percentile) while the worst 1% would be 1 (1st percentile). In this way, companies can be compared on an equal basis, without regard to size, industry, or anything else.

This is an example of using quantitative methods in conjunction with fundamental data.

The RS ranking takes more of a technical analysis view. It ranks, in the same manner as EPS, how well a stock has performed in comparison to all other stocks. The evaluation is based on price appreciation/depreciation over a given time period, so a stock which rose 10% would outrank one that rose 9%. Likewise, a stock which fell 5% would rate higher than one which fell 7%.

The EPS and RS rankings are quite obviously and intentionally comparison statistics. They are not very complicated in their calculation, but they quite handily serve the purpose of taking a given set of data and applying it in a useful fashion.

Market Behavior Constructs

The other primary form of quantitative analysis deals with the area of probabilistic behavior—defining or approximating the odds and likelihoods of given price movements taking place.

Here’s a quick example. Using the data in the S&P 500 futures from September 1997 through November 2006, we can make the observation that 47% of the time the market moves in the same direction two days in a row. That implies the market actually goes in the opposite direction 53% of the time.

This is a fairly basic example of observational quantitative analysis. It was a simple day-to-day comparison done in a spreadsheet with no heavy math. Even still, it provides us with worthwhile information. In this case we find out that the market does not generally have a tendency one way or another in regards to day-to-day directional continuation.

You might be saying “big deal” right about now. All we have come up with the not very surprising information that the market is about equally likely to go in the same direction two days in a row as it is to go the opposite. Granted, this is a simple observation.

If nothing else, though, it allows us to eliminate certain factors from our market understanding and/or to avoid certain paths of inquiry. Beyond that, the knowledge that there is no little or no bias in the figures, and the random behavior it implies, can become part of a larger model.

Here is another example. Below is the result of a study of forex price behavior from one day of the week to the next.

Currency Pair   Mon  Tue  Wed  Thu  Fri
AUD/USD          52%  44%  52%  44%  47%
EUR/USD          49%  45%  48%  42%  44%
GBP/USD          50%  48%  46%  47%  47%
USD/CAD          50%  38%  43%  50%  53%
USD/CHF          45%  46%  48%  45%  45%
USD/JPY           51%  47%  49%  46%  50%

The percentages indicate for each day of the week how often the market went in the same direction that day as it did the previous day. For example, on Monday AUD/USD trades the same direction as it did on Friday (up Friday, up Monday or down Friday, down Monday) 52% of the time.

Again, the results are not surprisingly neutral for the most part showing mostly no clear tendency. There is, however, at least one figure which points toward the potential for further research.

On Tuesdays USD/CAD tends to move in the opposite direction as it had done on Monday (since the 38% represents moving in the same direction, it would be 62% for going in the opposite direction). This would seem like tradable information. If we fade Monday’s price move (go against it) on Tuesday, we are going to be right 62% of the time.

On the surface, that seems like a workable system. The problem is while we know one thing—the tendency in absolute price behavior from one day to the next– we do not know any more than that. For example, we do not know how much price movement takes place.  That is an important piece of information. If one does not make sufficient profits on the winning trades to more than offset the losses suffered on losing trades, then it matters not one bit how often the winners happen.

The point is that statistics such as the ones we have just shown can be very useful, but you must understand the limits. Every statistical determination is done with certain constraints. In the above example, all that was considered was absolute direction, not amplitude of the moves. Constraints mean limitations. That is why such a study as we have just shown is generally just the first cut—a lead on to more comprehensive studies.

Types of Quantitative Analysis

The comparative and market price behavior analysis we have just discussed can be accomplished in a variety of ways. Some are very simple. Others are highly complex. They tend to fall in to one of the following categories:

Observation Counting
The table in the last section was generated through observation counting, which is nothing more than seeing how often something occurs. With a large enough data set you can use the results to get an idea of the tendencies of a market.

Examples of some of the types of things you can learn are:

• How often do 1% or greater moves occur?
• Does the market tend to move in one direction on a given day?
• Are high volatility periods clustered or randomly scattered?
• How long do trends and/or trading ranges persist?

Think of a question. Observation counting can probably answer it.

Statistical Evaluation
This category of market exploration includes things like regression analysis and other measures right out of most statistics text books. The most prevalent example of this kind of work is the well known Beta figure used in the stock market, which is based on the regression model. There are also commonly applied measures such as covariance which come in to play in portfolio composition.

Artificial Intelligence
The cutting edge of quantitative analysis is in the area of so-called artificial intelligence. This encompasses such things at Neural Nets and Genetic Algorithms. These are powerful tools for modeling and forecasting. Their use in the markets has been talked about for quite some time. Until recently, however, they were slow and unwieldy, making their application in actual trading difficult. Of late, however, performance improvements have begun to make them a more legitimate possibility for future use.

Words of Warning

Quantitative analysis can be a powerful tool, providing an array of avenues for research and market assessment. One thing must be kept in mind, however. The application of quantitative analysis to fundamental or technical data imparts the same limitations as seen in those methods. Using fundamental information means timing questions and lack of short-term applicability. Should technical studies be involved, lags (among other things) remain an issue. 

The bottom line, as always. is to know your tools, what exactly they are telling you, and how best to apply them.