"any change in the market for any reason can make a profitable system unprofitable"
An overstatement: Certain particular changes, usually based on very particular reasons, can make a profitable system unprofitable. In many if not most instances, these particular changes are quite susceptible to simple observation.
More generally, system-traders and discretionary traders alike presume that what has previously been observed in price action in whatever tradable is more likely than not to persist to some sufficient extent. All trading approaches presume elements of non-randomness and, just as or more crucially, what statisticians call "stationarity" and "dependence" in price action. In this sense, system-traders remain "curve-seeking" in their activity, no matter how simply they define their rules and no matter how rigorously and extensively they conduct their testing in whatever effort to avoid "curve-fitting." In addition, system-traders perforce apply certain "filters" that cause or allow them to focus on certain tradables or markets, and to exclude others. Feeling no compunction about seeking whatever curves there might be to exploit, discretionary traders typically take advantage of multiple filters of different types: Ideally, they'd like to be trading the most tradable tradable in the most tradable way at any given moment.
The chief advantages of system-trading over discretionary trading, at least in theory, would have to lie elsewhere (less susceptibility to impulsive trading errors, for instance) than in some absolute, superior view of market behavior. By the same token, however, if (and only if) there are indeed multiple "curves" out there to be discovered, then there is no inherent reason why a skilled discretionary trader couldn't outperform or improve upon even a very well-designed system. That is not to say that doing so is very likely, and there would obviously be no way to prove that just because Trader X has outperformed System X for the last ten years, that he or she is going to do so for the next ten. |
