Why h&s's fail:
There are a number of possibilities, the most obvious being that, as argued dramatically on this board a few days ago, the whole thing is an illusion. Random walkers cite computer studies of chart patterns that are said to support such skepticism entirely: The predictive capacity, much less the minimal effective tradability, of such patterns is said to be nil. If the pattern and any other selection criteria were well enough defined to be computer-programmable, and the patterns were intrinsically reliable, then the program would become in effect a money-printing machine, and the programmers would shortly become frighteningly wealthy. (Along these lines, one might also wonder why securitytrader.com and Advanced GET don't, say, own their own TV stations, skyscrapers, and private island empires by now.)
On the other side, the definitions and precepts that underlie such studies and other attempts to codify and universalize trading rules may themselves be subject to criticism. If the rules and criteria cannot be unambiguously stated, then they don't qualify as truly "reliable" on their own terms, but it's also conceivable that proper selection criteria and other "filters" and methods of the sort employed by successful traders might remain too complex or "artistic" to be formulated in any hitherto objectively testable manner.
As to the specific question, citing a personal study of 500 5-year daily charts, as detailed in ENCYLOPEDIA OF CHART PATTERNS, Thomas Bulkowski claims that properly identified h&s patterns remain among the most reliably tradable, with very low failure rates and very high relative likelihood of reaching targets. (BTW, Bulkowski's publishers claim that he was successful enough trading his formations to "'retire' from his day job at age 36.")
In defense of such chart-pattern reading, one could argue that many supposed H&S failures could also result from numerous other factors (exceptions that prove the rule), including:
1) The very widespread tendency to ignore volume characteristics. I've commented on this before, but have given up trying to point it out every time I see it;
2) The tendency to identify formations in the wrong place - seeing a top at a place (i.e., amidst a downtrend) where only a bottom would make sense and vice versa;
3) The higher noise-to-signal ratio that occurs the less well-defined the pattern is, and the smaller the timeframe - a properly defined (volume, spacing, placement) h&s pattern on a daily or weekly chart, reflecting many millions of shares traded over a substantial period of time, would be less likely to be the product of random movement than one appearing on, say, a 5-minute chart reflecting a handful of upticks and downticks. At a certain point, the relationship might invert - a formation on a monthly or yearly chart might encompass too much time for meaningful interrelationships to hold.
For my own part, though I've encountered some fairly sophisticated material that seems to contradict key elements of the random walk/efficient market theory, I'm shifting to a position of agnostic on these and related issues until I've undertaken some further study. |
