The filter settling time is an important criteria. as long as the underlying wave can persist its cycle longer than a factor of the settling time, you may expect a possible good 'snap shot'. The other equally important criteria is the impulse response. Then, the Q factor, ... etc. These requirements are so different from the Fourier based analysis. ... ... To say the less, the signal’s Randomness ( versus unpredictability, see http://en.wikipedia.org/wiki/Randomness#In_statistics ) is the key. usually the daily Open/Close discontinuity noise may smear the signal quality.
You would need 3 or 4 cycles to be able to reasonably identify a cycle. Filters wich use previous outputs that are fed back need time to settle. Filters which use only present inputs pick up harmonics, as much as 25% of the nearest cycle. By using many cycles worth of data, the parasitic effects of harmonics is lessened. There are sooo many tradeoffs going on.
News 
Market Data 
Discover 
