Backtesting depicted a beautifully rising asset curve for an EA (expert advisor / automated trading logic). The moment it was deployed to a real account, funds were gradually eroded, or sometimes rapidly. In the automated trading world, this phenomenon is commonplace, and many developers and users treat it as a “bug” in the program. However, from the viewpoint of a process design engineer, most of this breakdown is not a bug. Friction inherent in the real market environment, and the essential structural design (process design) required for real operation, are phenomena that inevitably arise from the design stage itself. This paper整理 (organizes) the verification philosophy that the lab uses as a foundation for development—typical design flaws that prevent EA from functioning in real markets, and three evaluation axes to mechanically uncover them—and formalizes the development stance.

1. Absence of Absolute Friction—Vulnerability of logic built in an ideal environment
Markets always have physical frictions such as spreads and slippage. If you omit these frictions in calculations and base expectations on a frictionless ideal environment, the advantage seen in backtests starts to erode from the inside as soon as it is deployed to a real environment. This is a typical pattern where backtest superiority vanishes in live trading.

2. Asymmetry of Risk Management—Structure that postpones collapse
Not placing a stop loss (SL), or averaging down while in drawdown. These designs superficially raise win rates, but they merely postpone the risk of account ruin due to a single market reversal to the future—a flawed structure. If you do not treat defeats as a current process, but instead accumulate them for the future, collapse is inevitable.

3. Loss of Macroscopic Perspective—Local optimization and curve fitting
Reliance solely on short-term signals such as 1-minute or 5-minute bars without viewing the longer-term trend walls indicated by daily (D1) or weekly (W1) charts. Or, over-optimization (curve fitting) by adjusting numbers to a past period, leading to loss of adaptability to unknown markets. Locally strong, but powerless against the larger picture.

4. Gap between Test Environment and Real World—Dependence on tool specifications
Logic that depends on tick interpolation in backtesting tools or millisecond-level processing assuming zero latency is realistically unimplementable in the real world; it becomes an impractical theory on paper rather than executable in a live account.

5. The Measure of Verification—Three absolute standards: survivability, logic, and real-world adaptability
To mechanically uncover these gaps, the lab evaluates the injected logic along three axes below.
   Axis 1: Survivability and the physical laws of friction. Check whether there is resilience to both short-term and long-term advantages and drawdown, whether the expected value withstands real absolute friction, and whether there is control of position sizing and margin maintenance based on a single-position principle (eliminating averaging down / martingale).
   Axis 2: Integrity of logic. Audit for curve fitting that only fits a past market era, whether the structure holds a shallow win rate by overly deep stops, intentional exclusion of unrealistic premises or inconvenient data, and look-ahead bias that references future data.
   Axis 3: Adaptation to real markets and robustness. Verify that there is logical basis capable of withstanding durability tests such as Walk-Forward Analysis (WFA), ensure that you do not ignore market liquidity and volatility realities, and confirm that millisecond processing or tool-specific phantom fills are not relied upon.
All three axes are not about subjectively judging the quality of a logic. They are instruments to mechanically extract the fact that there are structural gaps. The purpose is to identify where the holes are in the structure, not to determine superiority. This stance underpins the lab’s entire verification process and will continue to guide future logic development.

Semura Lab.'s defined development stance
We at Semura Lab. do not value the beautiful numerical results of backtests themselves. What we pursue is whether, in light of the three axes of survivability, logic, and real-world adaptability, the structural gaps in the logic can be uncovered as facts, and whether they can be rebuilt into a robust architecture based on those facts. We do not think all gaps discovered through verification should be discarded. If there is a skeletal framework containing raw material (the market’s possible edge), preserving that framework while redesigning only the missing layers is the logically correct approach. Verification to discover facts and reconstruction to adapt to real markets—a two-step process—is the stance we place at the foundation of our development, and we will consistently apply it to future logic development.