Artificial trader jumps candlesticks.

Sure, software is available to read candlestick charts. But here's how you can program your own system selecting the candlesticks and definitions that fit your method.

Candlestick chart analysis is subjective. You first identify the patterns and then judge their significance. For example, a white hammer pattern is bullish, but more so after a major downtrend.

Today, several vendors offer software that automatically identifies candlestick patterns. Some of these products also generate mechanical trading signals. They generally do well identifying the patterns but have mixed results in generating mechanical trading signals.

Let's now program our own system to identify and implement these signals. Using fuzzy logic we'll show you how to identify several candlestick patterns using TradeStation and how to integrate other forms of technical analysis with candlesticks to develop mechanical trading signals.

Clearly fuzzy Fuzzy logic is a powerful technology that allows you to solve problems that deal with vague concepts, such as tall or short. For example, if a person is six feet tall he might be considered tall compared to the general population, but short if he were on a professional basketball team. And how do you define a person who is 5'11" if 6 feet is considered tall? Fuzzy logic allows us to solve both these problems.

Fuzzy logic operators are made of three parts: Membership Function(s) shows how relevant data is to the premise of each rule; Fuzzy Rule Logic performs the reasoning within fuzzy rules; and Defuzzifier(s) maps the fuzzy information back into real world answers.

Here's how fuzzy logic works: Let's develop fuzzy rules that predict a one-year-old male child's height based on the height of his parents. The first step is to develop fuzzy membership functions for each variable's attributes. As such, we need to develop fuzzy membership functions for the height attributes of the mother, father and child. For our example, these attributes are tall, normal and short. We have defined generic membership functions for these height attributes as follows:

Tall=maximum(0,min(1,(X-Average Height)/(Stdev of height)))

Short=maximum(0,min(1,(Average Height-X)/(Stdev of height)))

Normal=Maximum(0,(1-(abs(X-Average Height)/(Stdev of height))))

When using these membership functions, substitute the following values for average height and standard deviation for the mother, father and child.