Sunday, November 18, 2007

Cycle detection



On some occasions, cycle detection which should always be done cautiously for reasons of non-stationarity, seems to be valid at a particular point in time. In this instance, the error level is close to our optimal Lag128 algorithm's error level. We then check dominant cycles (F1 and F2) both in periods and amplitude. F1 seems to have a reasonable amplitude relative to all other frequencies in the spectrum, and F2 can be considered as noise (no amplitude). There is virtually never more than 1 or 2 dominant frequencies in the L-S spectrum.

The usual caveat has been repeated over and over: there is no reason to believe the dominant cycle (period of 72 bars) will last very long. So we will only maybe project it for a short while using a phase advance mechanism, and confirm it with other indicators such as support/resistance levels, entropy, MTFS lines etc.


Answer in about 2 to 3 weeks...