V.13:07 (282-287): Using Filtered Waves for Trend Analysis by Scott Barrie
Filtering is simply processing price data to remove extraneous, noisy, information. What's left after filtering can be considered to be the more important and perhaps tradable information. Here's a method to filter price data for reliable patterns as well as some suggested trading plans.
Technical analysts use past market behavior as a guidepost for future behavior. By studying the past, an analyst can glean patterns that can be used as the basis for trades. The challenge is determining which past behaviors are appropriate to examine and then building trading schemes on those patterns. Here are the methods I employ to discern patterns, and the trading schemes I use in the Treasury bond futures market.
To compare present price action with the past, we must examine three facets of price behavior: first, magnitude, or the difference between starting and ending values; second, the duration, or how long it takes to move from beginning to end; and finally, the path, or which direction, up or down, from start to finish. The best process I have found for breaking down moves is the filtered wave method discussed by Arthur Merrill in his Filtered Waves, Basic Theory.
Filtering is the process of ignoring some data to discern a more telling picture from the remaining data: Ignore all moves less than a certain percentage, x%, and only look at moves greater than x%. I use a 1% filter to create a makeshift actuarial table of all absolute price movements greater than 1% (Figure 1). By using filters, I solve two of the three problems - magnitude and path. Then I use observed duration of the historical moves as my timing method, looking for moves that are overdue according to past average duration.
Before I go into the results of my analysis, these are the steps I used to analyze the T-bond market. The data was from Technical Tools daily data for the Chicago Board of Trade (CBOT) 30-year Treasury bond contract, from January 1, 1979, to December 17, 1993. I exported the data and used a Microsoft Excel spreadsheet to analyze it. I only used the December contract from the beginning of the contract year to contract expiration; for example, for the December 1979 contract, I used the first trading day in January 1979 until the contract expired. In Excel, I floored all percentage waves to the nearest 1%, meaning that all moves from 1% to 1.99% were grouped together as a 1% wave. This process was performed for all moves greater than 1% all the way to 7%, while all moves greater than 8% were lumped together. The same process was applied for negative waves. The results of the data I analyzed appear in Figure 1. The information that follows is illustrated there: