Imagine a market operating on multiple time frames. Each larger time frame is inclusive of, but also independent of, the shorter ones. Imagine each time varying in its degree of trending, from strong downtrending to non-trending to strong uptrending. Imagine each time frame also varying in its volatility, from highly volatile to very quiet.
In such a market, any trading method based upon rules and signals would vary in its results, from highly effective to highly ineffective. Traders utilizing such a method, whether it be mechanical or discretionary, would not necessarily obtain random results (because the method might accurately capture some portion of trending and volatility at particular time frames), but would undergo periods of profitability and slump as the market varied vis a vis the method's parameters.
Because the market's variation around the method's parameters would also occur across multiple time frames, the periods of slump and outperformance would also occur over varying time periods.
In such a regime, the long-term success of the method would be almost wholly dependent upon the trader's ability to recognize periods of outperformance and maximize those and recognize periods of underperformance and minimize those. It could be argued that the mechanical accuracy of the trading method would be less important to long-term success than the trader's ability to adapt to market shifts with risk management that takes maximum advantage of periods of valid signals and minimizes risk during periods of invalid ones.
To the extent, in such a regime, that shifts in markets lead to shifts in profitability and emotional shifts that range from overconfidence to underconfidence, traders would tend to run maximum risk just as their methods were degrading and minimum risk as their results were improving. This would ensure losing results over time, making the market seem rigged against human nature. It is not just the trading method, but the trader's management of this method, that might be crucial to long-term success.