1
/
of
1
store-traders-com
V.7:3 (73-75): Analyzing indicators with the cells method by David R. Aronson
V.7:3 (73-75): Analyzing indicators with the cells method by David R. Aronson
Regular price
$3.95 USD
Regular price
Sale price
$3.95 USD
Quantity
Couldn't load pickup availability
Analyzing indicators with the cells method
by David R. Aronson
Does a given technical indicator have predictive value or not? This fundamental question must be answered before using it to forecast market trends or as input to trading decisions. An intuitively appealing way to answer this question is what I call the signal event method. The signal event method evaluates the net profit or loss that would have resulted over some period of past data had a trader acted on the buy and sell signals generated by the indicator. Applying the signal event method, the analyst defines rules (i.e., one or more conditions) that generate buy and sell signals from the indicator. For example, assume the indicator is a price momentum oscillator calculated by the difference between a 3-day and a 10-day moving average. The indicator takes on positive values when the value of the 3-day moving average is greater than the 10-day moving average and negative values when the reverse is true. One possible signal rule could be: buy when the oscillator becomes positive; sell when it becomes negative.
The signal event method has limitations. First, it requires the analyst to define a signal ruleand there are literally an infinite number of rules ranging from the simple to the highly complex that can be defined for any given indicator. Thus, the analysis relates to the defined rule as much as it does to the indicator. I'd rather evaluate the indicator apart from any rules imposed on it.
Second, in the quest for better results, the analyst is often tempted to define complex, multiple-conditioned rules . Such over-fitted rules are more likely to prove profitable on past data, by chance alone, and are not likely to do as well in the future.
Does a given technical indicator have predictive value or not? This fundamental question must be answered before using it to forecast market trends or as input to trading decisions. An intuitively appealing way to answer this question is what I call the signal event method. The signal event method evaluates the net profit or loss that would have resulted over some period of past data had a trader acted on the buy and sell signals generated by the indicator. Applying the signal event method, the analyst defines rules (i.e., one or more conditions) that generate buy and sell signals from the indicator. For example, assume the indicator is a price momentum oscillator calculated by the difference between a 3-day and a 10-day moving average. The indicator takes on positive values when the value of the 3-day moving average is greater than the 10-day moving average and negative values when the reverse is true. One possible signal rule could be: buy when the oscillator becomes positive; sell when it becomes negative.
The signal event method has limitations. First, it requires the analyst to define a signal ruleand there are literally an infinite number of rules ranging from the simple to the highly complex that can be defined for any given indicator. Thus, the analysis relates to the defined rule as much as it does to the indicator. I'd rather evaluate the indicator apart from any rules imposed on it.
Second, in the quest for better results, the analyst is often tempted to define complex, multiple-conditioned rules . Such over-fitted rules are more likely to prove profitable on past data, by chance alone, and are not likely to do as well in the future.
Share
