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
柴亮太
09/22/2026 21:33
Release： 09/22/2026 21:33
Update： 09/22/2026 21:33
柴亮太
Investment style : Trading System
Investment career : More than 10 years
Investment amount : 10 million - 50 million JPY
Representative of RO LLC. My main line of work is AI agent research and development. I have about 25 years of investing experience, but it’s not a brag-worthy track record. I am not the type to sit still in front of a chart, and I would often cut losses when I looked away during critical moments. Through that repeated pattern, I have melted tens of millions of yen. What remains is knowledge and a record of my failures. “If I can’t watch it myself, I’ll build something that can watch for me.” With that in mind, I am also working on developing AI agent EAs. Not to predict the market, but to tell me when not to enter a position. To avoid silently expanding losses when prices move against me. Listing on GoGoJungle is a testing ground for that research and development. The tools I publish here are, as of now, completed products, but not the final finished version. They are things that run on actual accounts, record losses, and are cultivated while being improved each time. What didn’t go well is also shown as is. Therefore, for those who purchase, it isn’t a “buy and end” relationship, but a partnership to accompany the research. It isn’t that you are not winning because of your efforts. I have faced the same points and repeated the same losses many times. Those who stumbled in the same place, this time I hope we can turn to the winning side together and continue with verification. If you have questions or requests, please don’t hesitate to share. The feedback we receive will be reflected in the next version.
Article cover image

Even on Japanese holidays, it does not stop all day: a design that only in Tokyo time passes a soft caution to AI

Financial AI / AI Agent Research and Development; Market Context

Even on Japanese holidays, never stop all day: a design that passes soft caution to AI only during Tokyo time

Rather than simply stopping trading on holidays, we supply the AI with auxiliary information only during the Tokyo core time when liquidity is prone to thinning, leaving the higher-timeframe structure as the primary basis for decisions.

This article is a development log of the indicators and automated trading software (EA) I developed, being researched and validated with an AI agent and a live account. The verification process is also shared in a weekday 24-hour trading live stream.

Why not treat “holiday = stop”?

Even on Japanese holidays, GOLD can move significantly during London or New York hours. Stopping 24 hours based on a single attribute like a holiday would drag the Tokyo market’s caveats into overseas time and cause you to miss valid opportunities.

On the other hand, the Tokyo core time from 09:00 to 15:59 JST may have participant composition and liquidity different from normal days. Thin price moves, widened spreads, failed short-term breakouts, and slower chasing deserve more cautious scrutiny than usual.

Implemented time boundaries

  • 08:59 JST: soft caution invalid
  • 09:00 JST: valid
  • 15:59 JST: valid
  • 16:00 JST: invalid
  • Non-holidays: valid all day
  • Holiday library anomaly: neutral continuation

We set soft_caution_active=true only when it is a Japanese holiday and within the Tokyo core time. Outside these conditions, it is false. If holiday information cannot be retrieved, trading does not stop entirely; we revert to a neutral state with available=false.

Auxiliary context passed to AI

When soft caution is active, Gemini carefully evaluates from existing data the spread expansion, reduced tick ratio, higher-timeframe competition, single M5 breakout without H1/M15 confirmation, lack of retests, and long wicks as signs of potential failure.

Crucially, this information alone should not force a skip. PRIMARY=D1/H4, SECONDARY=H1/M15, ENTRY=M5 fixed pivot column are the main determinants, with holidays serving as auxiliary information. Do not trade solely based on the calendar while ignoring market structure.

Relation to multi-timeframe Dow structure

The Dow theory passed to the AI is a fixed pivot column, not a single candlestick. Look at the major direction on the daily and H4, intermediate reversals on H1 and M15, and entry on M5. Exclude ongoing candles and compare only confirmed information.

Tokyo time soft caution adds the context of “watch the chasing risk strictly” to this three-layer structure. When the higher and intermediate timeframes align and spreads and data freshness are normal, we do not close the entry merely because it is a holiday.

Prioritize mechanical hard guards

Even if AI interprets flexibly, mechanical guards such as warning signals, maximum spread, duplication prevention, account status, and order safety nets always take priority. Soft caution is not meant to weaken hard guards; it adds an extra layer of caution to the model.

If the higher timeframe is COUNTER_TREND, and Gemini has REVERSAL_ATTEMPT with no WITH_TREND or REVERSAL_BREAK on the intermediate timeframe, we mechanically skip whether it is a holiday or not. Treat calendar context and reversal confirmation as separate responsibilities.

What was tested

The new isolation test passed on 8/8. We verified boundaries at 08:59, 09:00, 15:59, 16:00; non-holiday; neutral operation when library is not installed. In audits that do not call real data APIs, we confirmed national holidays, Tokyo core time, soft caution active, and prompt generation.

Existing regressions were unchanged before and after. Known baseline 31/41, low-frequency 4/4, spreads 6/6. Due to remaining mismatch between old expectations and RELAX settings, we do not claim “all pass.”

Safe deployment procedure

Before deployment, the heartbeat confirmed open_count=0, tickets empty, ack_pending absent. While paused, EA5_ENTRY_ENABLED=False; the normal entrance was maintained, and after final deployment, EA5 entrance was restored. The 10:25 cron ran with NO_BREAKOUT and exit=0 to terminate normally; we confirmed the receiver and funda service are active.

We did not change SL, TP, lot size, HOLD, warning hard blocks, receiver/cron processing, AutoTrading, MT4, or GOLD13. The difference this time is the holiday context passed to AI and the basis saved.

Why keep it in the basis

If the AI defers a decision, you cannot improve without knowing whether it was because of the holiday or higher-timeframe competition. Thus, we save the holiday name, applicable Tokyo core time, soft caution status, and rule version to the basis.

The same applies to enters. If the caution context is active and you enter, you should verify which market structure and data justify the entry to have been beneficial. Not only the text but input values and the decision version should be reproducible.

Things not yet confirmed

Candidate natural Japan holidays × Tokyo core time have not been observed with actual Gemini judgments and saved bases containing new context. Therefore LIVE_VALIDATION is PENDING.

Even if the boundaries are correct in tests, real markets can be affected by data gaps, time zones, daylight saving time, library updates, and service restarts. In the next natural candidate, we will verify time, holiday name, judgment, and saved results across the entire cycle.

Evaluation metrics

  • Number of holiday Tokyo-time candidates and orders
  • Classification of reasons for deferral
  • Maximum forward and maximum reverse after deferral
  • Difference in spreads and tick ratio vs normal days
  • Number of misapplied cases outside Tokyo time
  • Basis gaps and time misalignments

Trading fewer on holidays alone is not a success. If the deferred candidates continue to expand, you lose opportunities. Compare in the same format as normal days and judge whether the caution contributed to mitigating real losses.

Handling of articles and videos

Only when live trading occurs will we embed recordings that match ticket, account, and time in the article. We will not use external trading charts merely because the holiday theme seems fitting. If there is no trading, we will not forcibly attach footage; we will publish the candidate and the operation log as verification results.

Conclusion

By not hard-stopping on holidays and by applying soft caution only during Tokyo core time, we can preserve overseas-time opportunities while communicating chasing risk in thin markets to the AI. We did not mix calendar, Dow structure, and mechanical guards into a single unit; each serves its own role.

The live deployment is complete, but live validation with natural candidates is still to come. We will distinguish deployed, tested, and live-market-verified statuses and update when the next relevant fact is obtained.

Explicit time zones

If the displayed time zones differ among the VPS, MT4, server logs, and article publication times, judgment at the 09:00 boundary will be incorrect. Holiday judgments should use explicitly converted times to Asia/Tokyo and record the original UTC, MT4 server time, and JST. Do not simply rely on the screen clock for boundary testing.

It is also important not to conflate London and New York times affected by daylight saving with Japan’s holiday judgments using a single fixed offset. Market sessions are calculated separately, and holiday core time is fixed in JST to separate responsibilities.

Holiday data updates and failures

The holiday library may be outdated, ad-hoc holidays may not be reflected, or loading can fail. If trading were stopped entirely in such cases, minor external data issues could halt 24-hour operation. This time we adopt fail-neutral, leaving the inability to obtain data saved in the basis and returning to normal market structure and hard guards.

However, to avoid missing continued inability to obtain data, we monitor the number and duration of available=false occurrences. Neutral continuation does not mean ignoring failures; it is designed to separate trading actions from operation alerts.

Next improvement decisions

Before gathering a sufficient sample size, we will not decide to stop all holidays or remove soft caution. We separate holiday Tokyo time, normal Tokyo time, and holiday overseas time, compare candidate quality, post-deferral price movement, and traded spread. We decide the evaluation period and minimum counts in advance and do not choose convenient periods afterward.

GOLD13 Sales Page
Weekday 24-hour Trade Live Drive

This article is a research-and-development and live-account verification log and does not guarantee profits or future performance.