Excuse me if I have to fudge a bit, as I seem to have misplaced my (not cheap!) copy of the book...
Anyway, to summarize, Harris's approach is to define one or a set of price pattern relationships - such as, say, "Three Higher Closes" - and test them against historical data for tradability. Another theoretical three-day pattern could be "Lower Close, Higher Open, Second Higher Open." He also describes how one might work with "delayed" and "split" patterns. The method operates without reference to indicators, volume patterns, intermarket factors, or mathematical measures of the various differences between the terms of the given pattern - his patterns are all restricted to simple relationships between Opens, Highs, Lows, and Closes - period. Thus, in the "Three Higher Closes" example, it wouldn't typically matter whether the closes were gigantically "up" or minimally "up." Of course, nothing prevents someone from applying Harris's ideas differently than Harris would, but, as soon as you started qualifying the patterns, you would be on the slippery slope to curve-fitting, in theory anyway.
The book itself includes many insights into strategy-testing and development (such as his caution against trusting tests based on strategies designed to capture very small moves in stocks), but is mainly concerned with laying out a framework, really a set of frameworks, for the systematic, more or less comprehensive and exhaustive definition and testing of patterns, whose complexity are theoretically limited only by the trader-investigator's imagination and, crucially, computing power, data, and programming ability. Also, though Harris does take time to discuss specific day-trading issues, all of his concrete examples focus on the daily time frame. The latter include Tradestation-generated performance results indicating Profit Factors in the range of 1.5 or so to 3 - for systems that produce on the order of 70 trades over 10 years of data. That the 10 years in question happen to be the 90s Bull Market doesn't seem to bother him. Similarly, though he refrains from optimizing the patterns themselves, he doesn't refrain from optimizing profit and stop-loss levels.
I suppose a true price pattern purist (plus royaliste que le roi), would insist on rigorous, comprehensive "exit pattern" tests to complement the entries. Somewhere, in trading strategy heaven or hell, someone (with lots of computers, data, and time) might successfully combine his methods and theoretical rigors with those put forward by Stridsman and Chande, or maybe throw in some intermarket stuff a la Ruggiero. In my own work, I continue to ricochet between a systems approach and a discretionary one. Right now, I've suffered enough as a result of my own excursions into system-trading to leave it aside for a while, but I imagine that some day I'll see if there's something I can do with it again. |
