Can technical analysis beat the market? Part II, Adaptive Market Hypothesis

In Part I, I surveyed a small sample of the literature regarding whether technical analysis works in the markets.  The literature ranged from discussions of candlesticks to genetic algorithms that evolve various trading rules.  The results are decidedly mixed.  Depending on which side you are on, you can probably find fault with any of the papers that don't support your view.  If you believe in technical analysis, you could say "these papers never examined Method x," whereas if you believe technical analysis does not work you might say "the authors data-snooped or the effect, if it existed, is gone."

I take three big themes from a synthesis of Part I:

1. Adaptive Market Hypothesis.  I'm fully a believer in the Adaptive Market Hypothesis (AMH) which is a much more logical explanation of what we see in the markets than the Efficient Market Hypothesis (EMH). The markets are decidedly adaptive, i.e. - they change over time.  Profitable inefficiencies are found and exploited away until they are no longer profitable for all but the most efficient players. The apparent decline of alpha is temporary until someone finds a new way to extract alpha.  If word gets out, everyone exploits it, but opportunities persist because of people have behaviorial biases, changing realities, new discoveries or a number of other possible factors.  EMH believers feel that all opportunities are already exploited, but never explain who exactly keeps the markets efficient for no apparent payback.  What these guys don't realize in my opinion is that the exploiters must get paid to do the exploiting.  The market is as efficient with respect to a particular area as the most marginally profitable exploiter makes it.
2. Technical analysis probably works, but it's fleeting.  It's a common theme in the papers of Part I and from anecdotes I've heard and personally experienced in some 20 years of investing that successful investment methodologies tend to disappear over time, consistent with the Adaptive Market Hypothesis.  Technical analysis, if it works, is likely no different.
3. Technical analysts must guard against datasnooping.  Data snooping is is the fool's gold of anyone trying to beat the market.  Turn over enough rocks, and you'll eventually find something that looks like gold, but is it really?  If you torture the data enough, eventually it will confess anything you want it to.  The papers in Part I provide methodologies to help protect against this bias.

I've never been a big fan of technical analysis due primarily to the usual inability of it's practitioners to explain why it supposedly works, and I completely reject any technical method that is subjective and cannot be computerized because it is simply not repeatable or even backtestable.

I have never used technical rules at all to trade or invest.  That said, I do believe that investor herding, insider trades, large money flows, smart vs. dumb money flows (perhaps), and even fundamental shifts in the economy would be, detected to some degree in shrewd combinations of technical trading indicators and exploited for all it's worth.  Combined with the fact that they are typically easy to program and even easier to get data for, and a lot less work typically than other methods, I can see the attraction.

Precisely because they are easy to program, they will be the first type of adaptive agent added to CASTrader.  The philosophy will be to let CASTrader sort out which, if any rules are valid.  In other words, if it works, I'm not going to be able to explain why it works either, I'm just going to know that adaptive traders using the Kelly Formula in CASTrader will exploit them the best they can.  In Part III, I'll discuss the details of implementation.