I am inclined to think that stock market cycles which used to exist may have been discounted by the market at this point, but I had not seen the 17 and 43 week EMA combination before, and I was impressed by how promptly the moving average crossovers identified major trend changes during this eight year period, and without giving false signals.
I have used the 50 and 200 day simple moving averages for this purpose; however, the 10 and 40 week moving averages give virtually identical results. This is based on research of mechanical trading systems I did years ago. I studied many simple systems, but only 50 and 200 simple moving average crossovers actually worked, when studied over 50 years. This simple trading system takes the long side only, since I also learned from my studies that selling short the S&P 500 is a losing proposition.
These crossovers correctly forecast big moves 78% of the time. The other 22% of the time, false crossovers will take you out of the market for a few weeks or months during shallow corrections, but the crossovers would have kept you fully invested during the great bulk of all bull markets, and out of the market during all major declines. I have used this simple system to trade mutual funds in my IRA account with good results.
My first impression is that the 17 and 43 week EMA crossovers may give better results. During the last three years, the 17 week EMA has stayed above the 43 week EMA of the S&P 500 the whole time, but the 50 and 200 day moving averages have given two crossover sell signals which proved to be incorrect, in August 2004 and in July 2006. In addition to causing small losses of profits by exiting mutual funds unnecessarily and then having to buy them back at higher prices a few weeks later, these two bearish moving average crossovers caused me to turn negative on the overall market just when I should have been most bullish, and thus I was late for a couple of good buying opportunities.
I believe the 17 and 43 EMA combination may be a superior tool and I plan to run historical tests with 50 years of data as soon as I can.