here's a picture of what I'm talking about:
|weekly chart of spy with pps indicator|
secondly, the pps (person's proprietary study, available in the free thinkdesktop platform from td ameritrade) indicator is the small up and down arrows. my observation is that if i had put on a vertical spread at the time and direction of the arrows i would have been generally happy with the results. with hindsight, i certainly would have been better off had i at least used the circled arrow as an opportunity to close my longs (woulda, coulda, shoulda.)
john person, the proprieter of the pps study, does not disclose its workings (you can listen to john on the sept 28, 2011 recorded seminar.) there is speculation that pps measures this or that moving average crossover but jp has denied this. i prefer to think that pps is what is known as a feed-forward network. a feed-forward network is the static, non-learning, part of a neural network - it emulates the neural workings of a highly trained savant. if pps is a feed-forward network then it wouldn't help me to know its inner workings because it would be just a matrix of weighting numbers. in any case, it is my observation that pps does identify reversals sooner than most other indicators and i plan to weight it more heavily in my own neural network.
lastly, i am looking at weekly charts now because they work better with my seasonality studies, sdi_seapro5 and sdi_seapro7. the seapro (seasonal projection) studies plot the average seasonal move starting on a specified anchor date. the reason i think they work better on a weekly rather than a daily chart is because of calendar issues. you see, the number of trading days in a year can vary on the year (leap years have 253 bars, and weekends don't always fall on the same dates.) going back 252 bars on daily chart does not always get you back to the same, year-ago date on the calendar, just more-or-less the same date. also, the number of trading days per year varies by the product. futures and forex products do not take all of the holidays that the us stock market does, so they have 8 more trading days in a year, or 260, more or less. however, there are far fewer deviations from a 52 week trading year*, so it is a better comparison to average the performance of the weeks that are multiples of 52 ago.
so here's how my weekly chart looks with these seapro studies:
|weekly spy with pps and seasonal projections|
*2007 had 53 weeks because the 53rd week began on december 31, according to the way tos delineates their weeks.