Financial Instruments Business Operator Kanto Finance Bureau Director (Kinsho) No. 1960 / Membership Association Investment Trusts Association, Japan Membership No. 012-02323 Go to GogoJungle Home
icon List
SEMURAラボ
09/27/2026 23:32
Release： 09/27/2026 23:32
Update： 09/27/2026 23:32
SEMURAラボ
Investment style : Day Trade
Investment career : More than 10 years
Investment amount : 10 million - 50 million JPY
Hello, Semura Lab. We are diligently pursuing logic development in our daily work. ■ Lab Policy "Build long-lasting, robust systems that reflect the realities of the market." This is the most important guiding principle for our lab. Even if backtests show clean, upward-trending performance, in real markets you can be exposed to slippage. Such stories are not rare. Therefore, we design with the assumption of real-world frictions like spreads and slippage from the outset. ■ What we are particular about - We value "why it works" more than appearance of performance - We avoid over-optimization on past data and emphasize reproducible grounds - We strive for rigorous verification that does not incorporate future information - We do not skip durability checks to see if it holds in unknown markets - Not only the reasons it moves, but also honestly communicate when market conditions are not favorable ■ Why we run this lab I myself have been disappointed many times by products with exaggerated claims. "Then we should deliver the facts as they are." That is the origin of continuing logic development and verification. We may not say flashy things, but we will not compromise on facing the facts. ■ What we provide We are developing tools and logic centered on MT4 that aid operational decisions and understanding of the environment. We aim to be tools that confront the market with solid grounds, not just a vague sense of something. ■ To everyone who purchases It’s not the end at purchase; we want to accompany you in ongoing operation as well. If you have questions or notice anything, please feel free to contact us. We would be glad to be of some help.
Article cover image

[Development Log] EA's "Exit Timing" Judgment — How to determine the lifespan of logic and the disappearance of its edge

When operating FX automated trading (EA), you inevitably face the question of “how long to keep this logic running.” There is abundant information on how to create and validate logic, but “when to stop” is surprisingly seldom discussed, and many people struggle with judgment. This time, we will整理 how to discern the lifespan of a logic and the erasure of its edge.


■ Logic has a lifespan

First, as a premise, no matter how excellent a logic is, there is no guarantee it will function forever. The makeup of market participants, the level of volatility, and the environment of monetary policy all change gradually over time. There is always a possibility that an edge found in past data merely happened to match the market structure of that era.

Possible reasons an edge fades include, among others:

・Changes in the market environment itself
・More participants using similar methods, diluting the edge
・Changes in broker spreads and execution conditions
・Overfitting during validation (cases where the model was too tailored to past data)

In other words, the disappearance of an edge is not a special event; it is a natural phenomenon that you will inevitably face in EA operation.


■ “Just losing more than winning” vs. “edge disappearing”

The most difficult part is making this distinction. Even a functioning logic will experience periods of drawdowns. In the backtesting stage, drawdowns and losing streaks are expected. Conversely, a logic whose edge has vanished may initially appear to be “just having a rough spell.” These two cannot be distinguished merely by short-term profits and losses.

Therefore, what matters is not the look of profit and loss, but whether it stays within the range you anticipated in backtests.


■ Points to use as decision criteria

Specifically, check the following points.

・Has the maximum drawdown clearly exceeded the level expected in the backtest?
・Are the expected value and the profit factor not deteriorating continuously compared to the validation (after enough trade instances)?
・Are win rate and average profit/loss still reflecting the “character of the logic,” not different from during validation?
・Is the change due to environment factors such as spreads or execution?

The last point is especially easy to overlook. A logic may appear to have deteriorated when, in fact, the broker’s changes in specifications or widened spreads were the cause. If you do not separate whether the cause lies in the logic or in the environment, you may take the wrong corrective action. Also, with few trades, the numbers are statistically volatile, so deciding “the edge has disappeared” after only a few dozen results is premature.


■ Set up the expected yardstick in advance

The foundation for decisions is the “yardstick” prepared before operation. Write down in advance the maximum drawdown, the maximum losing streak, and the monthly profit/loss variability from backtests, and be able to compare them side by side with actual operating numbers. Without a yardstick, you are left to judge whether something is within or outside the normal range by feel.

Also, it is desirable to decide in advance when to review, such as monthly or quarterly. If you swing between daily profits and losses, you will be swayed by short-term fluctuations. Regularly stabilizing your numbers and comparing with validation results will greatly improve the quality of your judgments.


■ Do not decide to stop while you are losing

The most to be avoided is making decisions in the heat of the moment while you are under the drawdown. During a losing streak, you are torn between “maybe it will recover if you wait a bit longer” and “let’s give up,” and both judgments are prone to bias.

What works is to decide the rules for stopping before you start operating. For example, “stop if maximum drawdown reaches 1.5 times the validation value” or “if, after a certain number of trades, the expected value remains negative, review it.” The exact numbers can vary by logic, but it is important to set them calmly in advance. If you have pre-set rules, you will avoid hesitation and being swayed by emotions.


■ Stopping does not mean ending

When you hear “stop,” you might imagine discarding the logic entirely, but that is not necessarily the case. Options include lowering the lot size and watching, pausing to review the environment, or re-running validation to re-evaluate. In some cases, you cannot determine whether edge has fully disappeared or you are temporarily out of sync with the market until you give it time.

Therefore, when stopping, also decide in advance under what conditions you will restart. For example, restart with a small amount once environmental factors are resolved; if re-validation shows the expected value turning positive, resume in earnest. Leaving it stopped and forgotten makes the logic fade away, so it is desirable that the stop decision and the next action be set together.


■ Summary: exit rules are part of the logic

Just as you carefully design entry and exit, you should also design rules for when to stop as part of the logic itself. Accept that edges have lifespans, and determine a path to exit calmly in advance. It’s a modest but essential way of thinking for sustaining EA operation over the long term.



Also, discussions on how to review logic and issues noticed during operation are available at Semura_Lab Yorozu Consultation Desk. If you’re interested, feel free to join.
https://www.gogojungle.co.jp/mypage/community/485