There has been a big (and good/useful) discussion of cycles in my blog comments recently. What I want to do in this post is to take a step back and offer a statistical look at the cycle signals from my Daily Trend System
Note: I am referring to my Daily System, not my 60-minute System. I have backtested my Daily System to 1982. And like with my 60-minute system, I trade the signals based on the sub-cycles. The last Daily System trading signal was a Cover Short/Establish Long signal at 1181 on 10/10/11 on the SPX (see here and see here)
There are multiple sub-cycles within a cycle in my system. And while I don't show the cycle signals (because I don't trade off of them, I use them to 'reset the boundaries' of my system, and they are used as confirming indicators), they do have some analytical usefulness and might give some insight here.
The reason being that we are likely going to get a long cycle signal sometime this week, assuming we don't have a major breakdown here (1226 should hold for a triangle continuation).
My cycles are not defined on some strict time relationship, in fact there is no time requirement at all. It is based completely on indicator moves (mostly MA based). Since 1982 there have been 24 cycles (with a long/up half and a down/short half). Minimum Duration: 154 days, Maximum Duration: 951 days, Median Duration: 390 days, Average Duration: 444 days.
So let me pose this hypothetical: The triangle is valid and resolves to the upside (continuation pattern). Let me outline some statistics associated with such a move.
I measure a cycle from long cycle signal to the next long cycle signal. This would make the last cycle ~420 days (if it ends this week), which means that it would be right around the average and median cycle durations. That is good and would make sense.
So what would be the odds that it is a move up and a 'fakeout' (i.e. that it would reach its triangle objective sometime in late Nov / early Dec and then crash, in a Wave 3 down or something similar)?
Even though I don't trade the cycle signals, I ran a study that was formulated like this: From the price the long cycle signal was issued, what is the highest sub-cycle signal price before the next cycle short signal was issued? This gives a way to quantify how many cycles did not 'fail'. The stats are pretty compelling:
24 Long cycle signals
21 cycles made higher high sub-cycle signals before next cycle sell (88%)
Max Gain: 68.0% (Max sub-cycle peak, not necessarily the next sub-cycle peak)
Avg Gain: 19.7% (Avg sub-cycle peak, not necessarily the next sub-cycle peak)
Avg Loss: -2.2% (Max value of failure)
Max Loss: -2.8% (Max value of failure)
Here is what I see, and here is the edge that I think the stats are giving me:
If we resolve the triangle to the upside, there is an 88% chance that we made a legitimate bottom in October and that we should be looking for more upside (and on average a large amount of upside, 20% gains).
Again, these are statistics, not prognostications.
But given the way this bottom has developed on my Daily System, and the bullish seasonality, the fact that this looks like a triangle (continuation), the fact that my 60-min system is behaving exactly like it does in a triangle (based on extensive backtesting), etc. etc. then I think if there is any edge to be gained from this analysis, it is to go long on a triangle break up and specifically to not 'expect' a reversal (end of Minor 2 and looking for a Minor 3 down, or something equivalent).