July Commentary

This month will be more theoretical as we will deal with the sensitivity, a word AlphOmega uses to describe the amount of retracement required to deem that a peak or a trough is detected. Generally the percentage of retracement is associated with a zigzag function where the last portion is dynamic thus unreliable until this retracement is attained. We have to face two issues, the first being the use of a percentage rather than point and the percentage being applied to the peak requires a greater absolute retracement than the same applied to the trough. If we try to overcome the first issue by selecting points, we need to adjust the amount of retracement points with regards to the price of each security as a 5 points retracement is negligible for IBM and material for a one dollar security like Transmeta. On the other hand we have a consistent gauge for all the waves of the said security, yet the gauge will become obsolete as the security price moves to another range. The second issue corrects both cons of the points but the use of a different base to apply the percentage leads to failure in consolidation ranges. When the price travels back and forth between the same two values and the range is small, we will get failed patterns. Another aspect of the percentage is the relationship between volatility and time frame. When we open a chart, we look at a preset number of bars where a set percentage may cover the whole range of the security or only a third. If it covers the whole range, then we hardly see a wave. This may seem like a handicap but it provides homogeneity in the observed waves across securities. The time will vary from one to the other but the relative size of the waves will be the same since it is a function of a percentage.

Double click on any image to enlarge

There is another variation of AlphOmega that can be used called the horizon method. We select a number of bars for which the program finds the highest and the lowest price, computes a percentage of retracement that is likely to produce observable patterns for the span of bars.The drawbacks are the percentage used varies as new high or low come and the wave count is purely a function of the percentage triggered by the latest bar range even when you look at older history, thus not as relevant. The use is obviously for real time as the expert is lighter in calculations.

AlphOmega will release version 5.7 in September 2006, with its Gartley patterns detection. The detection is very selective hence we get a very short list as it must conform to strict rules. The chart above is just a random selection of an exploration run during the tests. AlphOmega will close for vacations from August 10th to August 25th 2006.