Don:" 3-day trading patterns "
...interesting observations, Don. I became motivated to follow so-called "T+3 T/A" a little while ago, inspired by Gersh, who usually times his 401K entry and exit from VFINX - S&P500 Fund with a prerequisite 3-day lag to transfer his capital in and out of his fund; in effect, forcing him to time waves or other turning points three days in advance.
METASTOCK has the facility to "aggregate" two sessions, each depicted by a daily candlestick, into a "summed" T+2 candlestick (as many textbooks show). After doing this, which was useful, I took it another step and began to keep a rolling, T+3 candlestick chart for SPX+NDX as well.
A weekly chart would be a T+5 summation of candles generated from a Friday session... FWIW the T+2 and T+3 T/A charts approach weekly charts in their "more solid" feeling. What's different, and somewhat fascinating is that the T+2 and T+3 T/A charts are "dynamic", their aggregated candlesticks "rolling forward" indications of sentiment, as it is evolving.
FWIW I also use a variety of intra-day time slices = candlestick summations to generate apparent indicators of sentiment, depending upon the perceived "speed" of the market; obviously a touchy-feely kinda thing (^_^)
There exist known patterns that have to do with duration; eg., 200d EMA, 50d EMA; 90d, 30d, 5d and 1-day accounting cycles; and "the T+3 rule" required for funds transfers.
I find that the T+2 and T+3 T/A "rolling" candlestick charts are useful for exploring the evolution of sentiment as opposed to the "snapshot" observation of sentiment provided by standard daily and weekly charts. I give more "weight", in my mind, to the daily, T+2, T+3 or weekly candlesticks depending upon my perception of apparent market speed, not unlike using a variety of intra-day candlestick "time slices".
One might say that, by definition - "a reversal" is 1, 2, 3... N observations of time slices that sum together to become a doji, followed by a new sentiment = a subsequent change in direction: Whether intra-day or T+2, T+3, etc., we can always resolve "a reversal" to some time or price where things stopped moving in one direction, and became essentially neutral before they began to move the other way.
However, I find less consistency in a 3-day implied duration - as your postings can suggest (perhaps to readers who have not followed your open, stream-of-consciousness analysis-style for a long time :-)
...such consistency cannot be found when viewing either a 3d EMA overlay or, 3d (or multiple thereof) sto, where it would be omnipresent. In this case, it is as if you are "finding" that which you are "looking for": it "feels" to you as if your CLASS BUY & SELL signals for waves typically have three-day characteristics.
IMHO what you are feeling is this rolling, T+3 duration of sentiment - perhaps having to do with the known funds transfer cycle ?
What I do suspect, though - is that the duration of your CLASS signals is not so much dependent upon calendar time, but upon price change extent.
There are two ways to explore this hypothesis further, that I can think of. One is to relate your CLASS signals to (extents) changes in fibonacci levels, where prices start and stop. Another, very intriguing thing would be to destroy the time-scale altogether; ie., as in a Point & Figure chart, where perceived waves have no "durations", only "extents".
...indeed, on a P&F chart, (converted to "lines" so that it just looks like "oscillations") your CLASS signals appear to correlate better to different "classes" of extents, than any consistent "durations" on a time-series chart.
Perhaps this is because volume/velocity ("boy, it's a fast market right now") is not as apparently consistent as changes in "support levels" = fibonacci functions motivated by option boundaries, 200d & 50d EMA lines acting as strange attractors.
Try "fitting" various P&F increment-by-change settings to your CLASS signals, until you resolve a "best fit" pattern where your "turning point signals" are tops and bottoms of various waves defined not by their duration, but by their extent.
What would be useful to me, in this regard, is a record of your past CLASS signals in simple table/spreadsheet format, showing SIGNAL : DATE and PRICE where you believe that the signals best correlated (not "when you generated/posted the CLASS signal" but, "when and where you think past signals belonged" on the index) .
-Steve
geocities.com |
