[Laboratory Development Perspective] The Death of the System Caused by Adding Indicators and the Beauty of Subtraction
What Semura Lab. regards as most important in system construction is not showy profitability but survival in a live forward environment with real capital. I am not a programmer, but a process design engineer. Extracting pure causality from market noise and designing a robust architecture is my supreme mission.
In this article, we succinctly present the core philosophy of our lab, the "aesthetic of subtraction."
The Trap of Excessive Optimization (Curve Fitting)
Many EA developers, aiming to improve backtest results (profit factor and win rate), bind entries by summing multiple technical indicators to create increasingly complex conditions.
This is nothing but over-optimization that fits the system to past market data (in-sample). When faced with unknown markets (out-of-sample), the tangled conditions become a fatal bottleneck, and the system collapses in a brittle manner. Architectural complexity is the greatest enemy of portability and survival.
［ Figure 1: A tangle of complex code short-circuits, leaving only a single solid core of light.］
"Complexity from addition breeds fragility. A system gains robustness only through subtraction."
The Process Design of Subtraction
Our approach is the exact opposite. The system must be rebuilt not by addition, but by subtraction to the extreme.
| Comparison Item | Addition System (Conventional EA) | Subtraction System (Semura Lab. Standard) |
| Logic Structure | Layered combination of multiple indicators | Single condition rooted in market causality (causes such as squeezes or surges) |
| Resilience to Unknown Markets | Very low (panic triggered by unexpected noise) | High (depends only on essential causality, hence generalizability is preserved) |
| Backtest Appearance | Sterile, perfectly upward-sloping in a clean room | A transition that intentionally includes realistic noises such as drawdown |
| Maintainability | Poor (difficult to identify which variable is malfunctioning) | Good (identifying bottlenecks is extremely easy) |
If you strip away unnecessary filters, backtests naturally look worse, and you will also pick up unnecessary losses. But that is exactly the reality of live capital environments.
There is no holy grail in markets. Eliminating the black box and extracting only a complete causality through the “process of subtraction” creates true robustness.