Avega is a fully automatic intraday trading advisor based on the artificial intelligence algorithm.
The expert was trained on 6 popular currency pairs, using the most advanced and new technologies in the field of artificial intelligence. At the heart of the expert is the LSTM cell (neural layer with memory) and the DDQN learning algorithm .
Avega does not use dangerous strategies - martingale, a grid of orders and averaging. One deal is always open, and SL, TP is installed.
The peculiarity of the system is that the neural network itself selects the optimal settings, and also calculates Stop Loss, Take Profit and Lot, depending on the situation on the market.
The expert trades according to stable patterns in the market identified during training.
All weights for each currency pair are already written in the expert code for more convenient use, so the EA file will be large, a system with at least 1 GB of RAM is also recommended - as LSTM technology is very demanding on system resources. Each currency pair has been trained for a period of 10 years - this is the best volume / quality match for trading in forward tests.
- Recommended time frame H1 .
- Recommended pairs are EURUSD , USDCAD , GBPUSD , AUDUSD , NZDUSD , USDCHF .
- Recommended account type with 5 digit quotes.
- auto_lot - this parameter allows you to automatically calculate the lot from the risk parameters in the transaction and from the deposit.
- Lot - this parameter sets the lot size when the auto_lot parameter is disabled.
- PercentOfRisk - this parameter sets the amount of risk in a transaction from a deposit and SL. (in percents)
- Slippage - this parameter sets the allowed slippage.
- manual - this parameter enables manual control with rendering and signals for entering trades.
- auto_detect_settings - this parameter enables the mode of determining the currency pair in automatic mode and setting the optimal settings.
- symbol - this parameter sets the currency pair in manual mode if auto_detect_settings does not work.
2) добавлен слой softmax
3)обучены все 6 пар
4) добавлен фильтр волантильности
2) исправлена ошибка автоматического расчета лота
2) added information on the latest transactions on the info panel.
3) the parameter "days" is added, which is responsible for the period (days) when the total profit / loss was received.
2) the mode of increasing the lot is added.
3) the architecture of the LSTM neural network is changed, a significant number of neurons are added.
4) bug fixes and improved performance.