Auto Trade AI-EURUSD with state-of-the-art AI-equipped high-frequency trading A

～Features of the AI Equipped～

①A deliberately harsh-learning, rugged AI

This EA places trades based on a model learned and built with the latest optimization algorithms.

Learning uses 22 years of historical data to ensure robustness in training.

In addition, a variable spread is used in learning,by setting the minimum spread value to 3 pips, we recreate a harsher environmentto ensure durability

so that it can withstand any market condition.

②Learning evaluation criteria

In AI learning, it is necessary to provide an evaluation criterion for “what makes a good EA.”

For example, if the criterion is “win rate,” learning will proceed to maximize win rate, but focusing only on win rate can lead to extreme scores like

“1 win and 0 losses per trade.”

Therefore, this EA constructs a composite evaluation metric that emphasizes the following elements and trains accordingly.

1. Total gained pips

A fundamental evaluation metric.

Ultimately, the most important aspect of an EA is “how much profits can be extended.”

Hence, the basic axis of AI learning is the profits obtained.

2. Profit Factor

Profit Factor (total profit ÷ total loss) is also one of the most important factors.

A small PF means large total losses.

With a small PF, certain short periods can skew results toward losses, increasing risk.

Therefore, higher PF is trained to indicate a better EA.

3. Maximum Drawdown

Max DD is also a crucial measure of EA stability.

Maximum DD is calculated as the greatest among DDs not only at the moment of settlement but including positions held.

(computed as “the largest DD among all periods including holding positions”)as described.

In other words, the logic is built to genuinely avoid large drawdowns.

4. Time Efficiency

This is not a common indicator, but an original metric introduced to emphasize stability.

Time efficiency is the measure of how much profit was earned relative to the time a position was held,and the learning aims to maximize this time efficiency.

Often seen in martingale-type or high-win-rate EAs, where profits are realized if a position is held waiting just before a stop-out, it tends to result in idle positions when profits are small.

This metric greatly helps avoid “position holding.”

③Prevention of overfitting

Overfitting is the biggest risk in AI development.

Overfitting means the model adapts too much to historical data from training and cannot respond to new data at all.

Backtests may look good, but in real operation it performs poorly.

In tasks with many random elements like FX, overfitting is particularly likely.

Therefore, this EA does not train on all historical data; instead, it uses validation data separated into training and validation sets.

During learning, only training data is used, and the validation data is used to check for overfitting.

This EA further divides validation data and test data for a two-stage test.

In other words, the latter part of the shown backtests (the green section) are results on unknown data that the AI has not seen. (Approximately 7% of all data is unknown data.)

④How MT4 connects with AI

This EA includes a simplified AI model built-in.

Because of this feature, compared to standard EAs, backtesting takes longer, and the main ex4 file size is larger.

However, installing the EA is simple and can be completed in a few minutes, so please rest assured.

～Backtest result introduction～

Backtests use TickDataSuite data with a variable spread for testing.

It is almost as close to live environments as historical data, and model quality is very high at 99%.

Below isthe result of testing with an initial margin of 1 million JPY and a fixed 0.1 lot.

① Avoiding gradual losses

Maximum DD marks a very low 2.53%.

Also the average winning trade is 982 yen, and the average losing trade is 1,675 yen; although losing trades are larger,

the design prevents risk-reward from deteriorating and avoids losing accumulated profits in one go, steadily increasing profits.

②High PF, high win rate, high trade frequency

This EA also features a high PF, high win rate, and high trade frequency.

PF is 3.95, win rate 87.25%, and number of trades 11,749, boasting results achievable only through AI learning.

On average, around 44 trades per month.

～Trading logic～

The logic is not entirely left to AI; it combines the know-how from previous EAs with AI performance.

The basic logic is contrarian scalping, with finer decisions made by AI.

No further details are disclosed. Please understand.

～Three selectable modes

This EA can select from three modes via parameters: “small profit aiming mode,” “normal mode,” and “high profit aiming mode.” (Default is “Normal mode.”)

In “Small profit aiming mode,” settlements occur with smaller profit margins than in “Normal mode.” The win rate is high and stable, but profits are smaller.

In “High profit aiming mode,” settlements occur with larger profit margins than in “Normal mode.”

Waiting for profits reduces trading frequency and can increase max DD, but targets larger profits.

～Compounding feature～

Below are backtest results using the compounding feature.

To compare with the no-compounding version, the first position is entered at 0.1 lot by adjusting the parameters.

Backtests show that from 1 million JPY, funds grew to 127 million JPY over 22 years.

Also notable is the maximum DD at a very low 4.09%.

This is because the loss amount remains stable over 22 years, so enabling compounding has little impact on overall performance.

～Frequently Asked Questions～

Q. What specs are recommended for the trading environment?

A. A PC or VPS with 3 GB of memory or more is recommended.

Q. Does this EA perform machine learning?

A. The model (the result of learning) built by machine learning (AI) is included.

Q. Can it be used on any timeframe?

A. This EA is for GBPUSD on the 5-minute chart only.

Q. Is installation easy?

A. A clear manual explaining the installation is included.

Q. Should I stop during major news announcements?

A. It is not necessary to stop during news announcements.

Q. Are there any broker restrictions?

A.Available only with brokerages where server time is GMT+3 in summer, GMT+2 in winter.

Also use is prohibited with pro-type brokers, and some proprietary trading brokerages may restrict usage.

Q. Can I use it with multiple brokers?

A. Only one account per purchase. If you want to operate on additional accounts, please consider purchasing separately.

Q. How many positions can I hold?

A. By default, up to 3 positions are held. Hedging is not performed.

The maximum number of positions can be changed via parameters, but 3 positions is recommended by default.

Q. Do you use Martingale or grid averaging?

A. We do not use Martingale (lot size is fixed or determined by compounding).

Default is to hold up to 3 positions, with slight grid averaging adjustments if the price moves unfavorably.

Q. Will the backtest results match real trading?

A. The EA is trained to avoid overfitting to historical results, but during certain periods it may be negative, and future trading results cannot be guaranteed.

The forward test results are also published, so please refer to them as well.

Q. How is the compounding feature calculated?

A. The compounding feature automatically sets the lot size that can be entered based on the risk (percentage) of the excess margin.

For example, if excess margin is 1,000,000 and risk parameter is 1%, the maximum lot size that can be entered with 10,000 units is set.

The exact lot size for each entry is shown in the MT4 Expert Advisor display, so please check.

Q. What is the logic?

A. The fundamental logic is contrarian, but the detailed logic is determined by AI learning.

Q. The trading frequency is low

A. Depending on market conditions, there may be several days without entries.

If you loosen the maximum spread parameter, entries may occur, but this is not recommended for long-term stable operation.

There may also be errors in MT4. Please periodically check the terminal’s Experts tab for errors.