SynAIpse EA
The SynAIpse EA analyzes multiple entry patterns in real-time by utilizing internal AI Decision Engine trained models combined with a comprehensive set of filters and with price movement confirmation. In cases where the market moves against the trading cycle, the EA automatically and strategically initiates recovery trade grids.
The AI decision engine is built in and fully API independant; this means there is no requirement to access any GPT APIs over the internet (which is also not possible via the strategy tester).
What sets the SynAIpse EA apart is its innovative approach to trade recovery management; We have taken a different approch with AI models, we utilise a set of models for entry and exit, but, also, we employ the AI Decision Engine models to manage risk effectively by dynamically mapping market structure in real-time to determine activation and optimal timing and placement for recovery trades during adverse market conditions.
General EA Settings
| Symbol | Timeframe | Lot Factor | Model Set |
|---|---|---|---|
| AUDNZD | M15 | 1.0 | m=1 |
| AUDCAD | M15 | 1.0 | m=1 |
| NZDCAD | M15 | 1.0 | m=1 |
| USDCHF | M15 | 0.5 | m=3 |
| EURCAD | M15 | 0.5 | m=3 |
EA Parameters (set these by using the relevant Setfile)
[Strategy Settings]
Use Symbol: (true/false) Enable or Disable the Strategy for the relevant symbol.
Symbol: Strategy instrument name without prefix or suffix. The prefix and suffix are automatic from chart symbol name.
Timeframe: Strategy timeframe (default M15).
Magic Number: Magic number for the strategy.
Lot Factor: Factor of the lotsize for each strategy. (Main pair default=1.0, extra pair default=0.5).
Model Set: This is the AI model to be used for the strategy. (Main pair default=m1, extra pair default=m3).
[Risk Settings]
Lot Selection Type: (Fixed Lots/Lots Per Equity) Use this to select the type of lot sizing.
Fixed Lot Size: When fixed lot-sizing is selected, this sets the fixed lot size for each trade. For example, setting it to 0.01 lot means each trade will use 0.01 lot size.
Lots Per Equity: When lot selection is lot per equity this will set 0.01 lots per “Lot Per Equity” setting. i.e. If set to 3000, then for every 3000 base currency equity units the start lot will be 0.01.
Max Live Trading Symbols: This sets the maximum number of individual symbols that are allowed to be trading at the same time (default=5).
[Day Filter]
Trade On
[Holiday Filter (Thin Markets, High Volatility)]
Use US Fed Holiday Filter: (true/false) This allows you to stop trading when US stock market is closed (default=true).
All US Fed Holidays Filter: (true/false) This allows you to stop trading on extra fed holidays i.e. Veterans and Columbus day (default=true).
Skip Trading Us Election Week: (true/false) This allows you to stop trading for the US election week in (default=true).
Skip Trading over Xmas / New Year: (true/false) This allows you to stop trading over Xmas and New Year (default=true).
Day to Stop Trading in December: If “Skip Xmas Trading”, then this is the date in December to stop opening new trades (default=21).
Day to Start Trading in January: If “Skip Xmas Trading”, then this is the date in January to start opening new trades (default=21).
[Equity]
Draw Down Limit in %: Close all trades if draw down reaches the draw down limit in % of equity.
Emergency Close Profit Bucket if #Trades>Limit: (true/false) Enable max open trades close by total bucket profit (default=true).
Emergency Max Trade Limit to Close Profit Bucket: Number of open trades to enable the total bucket closing process (default=30).
Emergency Close Bucket Profit %: Percent profit across all instruments/symbols to close all open trades (default=4.1).
Emergency Limit #Symbols if #Trades>Limit: (true/false) Enable max open trades max trading symbol limit (default=true).
Emergency Max Trade Limit to Limit Symbols: Number of open trades to enable the max trading symbols limit (default=50).
Emergency Limit Symbols Number: The maximum number of trading symbols once max trade limit is hit (default=1).
FAQs
Q. Are systems that employ grid systems as part of the overall operatiopn dangerous ?
A. No, well, ….. sometimes ( or most of the time ), it depends. If a grid recovery system is not rigorously tested across decades of different market conditions and is just coded to doiuble the risk and add more trades, then yes, that is not a great way to approach the problem. But, if the differeing adverse markets are studied and understood and you can calculate the best places to place the grid trades conbined with decades of testing, then, you can run a trading ystem with a grid element and it works really well. We have used many different types of machine learning techniques to make sure they system is robust as possible.
Q. How many years were your models traned for ?
A. We have trained hundreds of models over the years, we usually train for at least a decade, but we also concentrate on recent markets, then run out of sample tests on older markets to test efficiency.
Q. How often do you train the models ?
A. Great question. We are always looking for better models, we are training some as you read this ( we are always training ) and when we find better performing models we will add them to the EA code.
Q. Why are there recovery trades above or below the initial trade ?
A. Another great question. The answer is usually the market has not settled after high volatility. The model understands this, locks out the trades to protect them, and then starts the recovery grid once the market has settled down. This means the grid could be above or below where you would expect it. This is normal.
Q. How did you implement ML/AI into mql4 as most of the functions are not available ?
A. With great difficulty. We have had to do a lot of coding and storing data manually by looking at some of the open source modules. With all the best will in the world, it is not exactly the same and we are not sure we will keep the MT4 version going forward due to the work required to maintain.
Links
Anaconda https://www.anaconda.com/
Tensorflow https://www.tensorflow.org/
PyTorch https://pytorch.org/
Keras https://keras.io/
Scikit-learn https://scikit-learn.org/stable/
MQL5 Machine Learning https://www.mql5.com/en/articles/machine_learning