GBPJPY Anomaly v04 | Day-of-week and time-only trading time anomaly EA

■ Judging materials are only a clock and a calendar

GBPJPY Anomaly v04does not use any indicators at all.

Neither moving averages, RSI, nor Bollinger Bands are stacked.

What this EA watches is simply “what day of the week it is today and what time it is now”—that is all.

In the foreign exchange market, there are price-mushing habits that recur regardless of market sentiment.

Flows toward the Tokyo fix. Month-end settlement needs of the Fifty-day (go-to-bi). End-of-month rebalancing. Monday morning open.

These are not predictions of whether the market will rise or fall, buta structurally biased pattern born from the schedules of people and organizations participating in the market.

This EA cuts out the time zones where those biases are observed into 5-minute grid cells and only fires at110 surviving cellsas determined statistically.

■ Ten-year actual performance

This is the actual result from the MetaTrader 4 Strategy Tester (all ticks / no misalignment).

All figures are after deducting all trading fees.

【Testing environment】

AxioryAsia-02 Demo (Build 1471) / GBPJPY 1-minute chart / 2016.01.04 – 2025.12.30 (about 10 years)

All ticks (Modeling quality 25.00%)/ Misalignment chart error0/ Test bar count 3,719,556

Initial margin 10,000 / Spread 10 (1.0 pip) / 0.05 lots fixed

■ I will write out all the logic

We do not write “it is a black box.” Here is everything about this EA.

【110 cells breakdown】

• Mon 1 o'clock (buy) 12 cells — just after market open on Monday

• Tue–Fri 0 o'clock (sell) 48 cells — early in each trading day

• Tue 1–2 o'clock / Wed 1 o'clock (buy) 18 cells — Asia session ramp-up

• Goto-day 0–2 o'clock (buy/sell) 29 cells — Fifty-day settlement demand

• End of month Friday 5 o'clock / 20th 17:00 / 10th 13:00 3 cells

■ To avoid curve fitting

Each cell’s parameters are determined byforward testing (walk-forward).

Using only the most recent 3 years of data to select the optimal cell combinations → test on the next year → shift by one year and repeat

This is repeated for 10 cycles, and the profit/loss during the testing periods (which at selection time are all unknown data) waspositive in all 10 cycles.

The values built into this product are theexact combinations chosen by the latest cycle (learning period 2022–2024) are fixed as is.

We do not search for the best combination by selecting years with good results or by surveying all periods.

The selection is made under the discipline of “only use information available at the moment of firing.”

Furthermore, among the 110 cells,104 cells converge to the same parameters (take profit 20 pips / stop loss 20 pips / hold 120 minutes).

Forward testing continued to yield the same answer across years, indicating the parameters are not at sharp peaks but on abroad plateau.

■ Also note the weaknesses

Please know before purchase which scenarios this EA struggles with.

1. Profits skew toward the later period

2016 and 2018 were losses for the year. 2017 was almost flat.

The biases in the time windows captured by this EA depend on the composition of market participants and trading hours.

The most profitable decade was the most recent three years, so please do not assume ongoing profitability and set capital accordingly.

2. Depends on server time

This EA triggers on the account server time. Brokers with a time difference in the testing environment will fire in completely different time windows, and the above performance will not be reproduced. Please check your server time before deployment.

3. Firing times cluster around low-activity periods

Most activations occur around the time the date changes. This period tends to have wider spreads and may differ from testing results depending on account type.We recommend accounts with narrow spreads.

4. Stop loss does activate

During the testing period, 535 trades, or 17.7%, were closed by stop loss. It is designed to cap losses at 20 pips.

■ Recommended operating conditions

■ Included items

• GBPJPY_Anomaly_v04.ex4 (EA executable)

• User manual (Japanese / PDF)

• User Manual (English / PDF)

• Strategy Tester report

■ In conclusion

This EA does not rely on predicting the market.

What exists is a collection of observed facts: “at this time on this day of the week, market participants’ circumstances tend to bias prices.”

Therefore, if that bias disappears, performance disappears as well.

I have presented yearly performance and weaknesses openly so you can use it with full knowledge.

Please use ten years of actual results and the underlying patterns as your decision material.

SEMURA, Process Design Engineer

※ The product’s performance is based on historical data and is not a guarantee of future profits. Foreign exchange margin trading is a type of trading that can result in principal loss. Please use within your excess funds, based on your own judgment and responsibility.