Taking stock after months of testing: the road to an edge is narrower than I hoped

My recent research into simple technical indicators has confirmed that they lack a sustainable edge in the current market. I spent time testing a range of single-indicator strategies (including EMA, Donchian channels, trend-following with ADX, and RSI) across JPY-based FX pairs on H4 and D1 timeframes, as well as metals on the D1 timeframe. When I subjected these strategies to a walk-forward analysis (testing them on data they haven’t seen before to ensure they aren’t just memorizing the past), none showed robustness. Any instances of strong performance were merely period-dependent, meaning they were the result of over-optimization or pure luck. This outcome is within my expectations. In an efficient market, simple technical analysis rarely holds a persistent advantage. Continuing to churn through simple strategies is essentially data mining, which creates a high risk of finding false positives. These are results that look great on paper but fail the moment you trade them with real money.

Building a reliable filter

While the strategies themselves didn’t pass, the real asset I have established is a reliable testing infrastructure. I have built a framework that handles data conversion, execution engines, GPU-accelerated optimization, parallel processing, and prop-firm evaluation. In other words, the most valuable part of this project is my ability to quickly reject strategies that don’t work. This is the core of capital preservation. Knowing what doesn’t work is just as important as finding a strategy that does.

Where to go from here

Looking at the path forward, I see four potential directions for further development:

• Advanced methods: Exploring ensemble models, regime-based filtering, alternative data sources, or discretionary assistance.
• Goal recalibration: Distinguishing between passing a prop-firm challenge (which is mathematically achievable through risk management and diversification) and long-term, consistent withdrawals (which require a genuine, persistent market edge).
• Integration: Incorporating your own specific hypotheses, proprietary data, or unique market insights.
• Market selection: Looking beyond prop-firm models to other trading avenues. I have decided to start with the second point: focusing on the quantitative requirements for passing a challenge. Given the current data, this is the most logical next step for maintaining a disciplined, objective approach to trading.