They don't always agree about the positioning of the cycles either. Drokes has the 10 year cycle as bottoming in 2004, whereas Ferrera has it bottoming in 2002 (projecting the next bottom in 2012 or 2014). All the analysts other than Curry and Barker tend to focus on cycles longer than 1 year. Barker covers the Wall (20 week) cycle. Curry covers the 45 trading day cycle (10 weeks?) and the 90 trading day cycle (18-20 weeks).
That leaves a niche of 20 weeks or shorter which is where I tend to focus most of my analysis as this is the time span that probably most interests swing traders (and day traders). I believe I have identified (and often use) the 20 week cycle (100 trading days or 4.5 months), the 33 trading days or 1 1/2 month cycle, the 11.2 trading days or 1/2 month cycle.
In deference to day traders I also look at the 5.6 trading day cycle (1/4 a month) and the 2.8 trading day cycle (1/8 of a month). I am not opposed to using a 45 trading day cycle (10 weeks) if the data suggests that is the cycle currently in dominance. I try to adjust as the data suggests I should as cycles fade (amplitude decays) or come to the forefront (amplitude expands).
I try to cover some of the longer cycles every month or so to give you a look at the overall environment in which these shorter cycles are functioning - as we should not ignore the big picture. So I am not in competition with these analysts (and their conclusions), but I believe I extend their analysis into the shorter cycle arena.
Additional cycle information (some mathematical relationships):
Gann liked 360 and factors of 360 (270, 135, 90,45, etc) as cycle lengths. These are the basis of Jim Curry's analysis (also described by JM Hurst).
- 45 calendar days = (45 - 12.85 weekend days) is 32.15 trading days
- Note: 46 calendar days works better. One month avg = 30.4 days. so 1.5 months is 45.6 days. 12 (months)/1.5 = 8 periods per year. 8 x 45.6 = 364.8 days (1 year). So 1 year is 2 and 2/3 Wall cycles.
- 92 calendar days (2 x 46) is about 66TDs, and 3 calendar months (92 / 30.4 = 3.02 months)
- 138 calendar days (3 x 46) = 4.5 months (138/30.4 = 4.53 mths)
- 138/7 = 19.7 weeks. 138 - 19.7 x 2 = 98.6 TDs (approx. 3 X 33 TDs)
I would suggest Curry with minor modifications to his described cycle lengths is on the same wave length as the other analysts
Finally - just for the heck of it (everybody seems to be pretty much in agreement on the longer cycles) let's consider one more cycle - the PI cycle.
Definitions:
- PI cycle = 3,142 calendar days.
- Kitchin cycle = 40.68 months
- Wall cycle = 20 weeks (100 trading days)
= 8.61 years (8.61/ 3.39 = 2.54
= 103 months (103/40.68 = 2.54
= 448 weeks (448/20 =22.38 ~ 22 Wall cycles (22/9 =2.50 Kitchin cycles)
Conclusions:
1 PI cycle = 2.5 Kitchin cycle
1 Kitchin cycle = 9 Wall cycles
9 Month cycle = 2 Wall cycles
1 Kitchin cycle = 4.5 X 9 month cycles
3 Kitchin Cycles = 123 months ~ 10 year cycle
As you can see there are mathematical relationships between all these cycle.... There seems to be a mathematical order to cycles. This by no means covers all cycles. For Example: a Super cycle is 30 years or 3 x 10 year cycle (or 9 Kitchin cycles). Some claim there is a 17 year cycle (which is approximately 2 PI cycles). Just something to think about.,,,
Just curious are cycles like technical charts where the bigger time frame over preceeds the smaller time frame. You can day trade using the smaller time and pattern but always look to the bigger time to see if they agree with the shorter time frame.
ReplyDeleteSteelie,
ReplyDeleteI believe that is a valid way to look at it.
Like with trend lines - over time the longer cycles will prevail, but in the shorter time frame you can take advantage of movements within the shorter (trend lines) cycle's amplitude.
Excellent post cycle dawg
ReplyDeleteCan you elaborate on amplitude decay and is there anyway known to measure this realtime or otherwise ??
ReplyDeletec_k
ReplyDeleteExcellent question, There are methods using regression analysis and fourier analysis that attempts to do this. One of these days (if I find the time) I plan to look at the DPO indicator and see if it provides any clues into answers for this question.
Detrend the price series to focus on cycle lows. Detrending can be done with the Detrended Price Oscillator (DPO). This indicator is based on a centered moving average. In other words, a moving average that has been displaced to the left by a factor (N/2 + 1). A 20-day DPO would be based on a 20-day moving average displaced to the left (past) by 11 days [(20/2 + 1) = 11)]. DPO would then be the closing price less the value of the displaced moving average. The resulting oscillator reflects price movements above and below this displaced moving average. We can then use oscillator dips to identify a cycle. Notice that the DPO ends before the last price. This is because the moving average is displaced and the DPO aligns with the displaced moving average. It often helps to set the DPO in line with the cycle length. Use a 10-period DPO when looking for 10-day cycle lows or a 40-day DPO when looking for 40-day cycle lows.
I know I did not answer your question, not sure there is a simple answer. At this time I have a couple of ideas, but no answers.