Peceptron_Mult
A multi-layer perceptron (MLP) neural network-based Expert Advisor that uses supervised learning to predict price direction and generate buy/sell trading signals. The EA trains its neural network on historical price features and executes trades based on the network's forward-pass predictions, combining machine learning classification with automated order placement.
Open the interactive view (equity curve, symbol switch)Key facts
Backtest results (H1, about 1 year)
| Symbol | P&L | Annual | Max DD | Trades | Win rate | Sharpe | Recovery | Losing streak | Per trade | Status |
|---|---|---|---|---|---|---|---|---|---|---|
| EURUSD | 0.00% | 0.00% | 0.00% | 0 | — | — | — | — | — | Not convertible yet |
| GBPUSD | 0.00% | 0.00% | 0.00% | 0 | — | — | — | — | — | Not convertible yet |
| USDJPY | 0.00% | 0.00% | 0.00% | 0 | — | — | — | — | — | Not convertible yet |
| XAUUSD | 0.00% | 0.00% | 0.00% | 0 | — | — | — | — | — | Not convertible yet |
Why this EA never traded: no entry rule survived the conversion.
Converted from MQL5 to Python and run with the same rules for every EA. Positions use MT5 lot sizes (100,000 units per FX lot, 100 oz per gold lot) with 1:100 leverage, one order at a time, and a stop-out below 10% of the starting balance. Sharpe is computed from the equity curve. Results marked as unreliable are shown for completeness only.
Blow-up risk: No blow-up pattern found
None of the usual blow-up patterns were found in the code, and the account was not stopped out in our backtests. That is not a guarantee: any EA can lose money.
High: stopped out in our backtest, or martingale/grid sizing without a stop loss. Elevated: martingale or grid sizing with a stop loss, or several positions at once without a stop loss. What the label means and what we found →
Strategy analysis
This EA employs a multi-layer perceptron (MLP) neural network as its core strategy engine, classifying it firmly in the machine learning trading category. Entry conditions are determined by the neural network's forward-pass output—opening long positions when the output exceeds a buy threshold and short positions when it falls below a sell threshold. The network input features are typically derived from technical indicators computed on price series (such as moving averages, momentum, and volatility), which are normalized before being fed into the network. The training phase uses historical data with backpropagation to update weights, operating under a supervised learning framework where labeled price direction data guides the optimization. Exit conditions generally rely on fixed take-profit and stop-loss levels or reverse signal triggers for position closing. Position sizing is configurable through input parameters, but the neural network itself does not dynamically control lot sizes. Money management considerations must account for overfitting risk on training data and the network's degraded adaptability during regime shifts in market conditions.
Strengths
- Machine learning approach can theoretically capture non-linear price patterns, offering greater adaptability than traditional indicator-based strategies
- Neural network has self-learning capability, continuously optimizing weight parameters through historical data training
- Multi-layer perceptron architecture allows flexible adjustment of network depth and width to match varying market complexity
- Supervised learning framework provides clear training objectives and quantifiable prediction accuracy evaluation
- Input features can be flexibly selected and combined, supporting multi-dimensional technical indicator fusion analysis
Weaknesses
- Neural networks are prone to overfitting on historical training data, with live performance often significantly worse than backtesting results
- Poor generalization during market regime shifts, as changes in data distribution outside the training period cause prediction failures
- Lacks dynamic stop-loss and adaptive position sizing, resulting in monolithic risk control measures
- Network training is computationally expensive, and real-time training on the MT5 platform may introduce latency
- Black-box model makes trading logic difficult to interpret, hindering strategy diagnosis and optimization adjustments
- Does not account for slippage and spread impacts on short-term trading, where live costs may erode profits
Possible improvements
- Add regularization techniques (L1/L2 regularization, Dropout) to prevent overfitting, and introduce cross-validation to evaluate model generalization
- Implement dynamic stop-loss and adaptive position sizing, such as adjusting stop distance based on ATR or position size based on prediction confidence
- Add a market regime identification module to switch or retrain models under different market conditions
- Introduce feature importance analysis and feature selection pipeline to reduce noise feature interference with the model
- Add slippage, spread filtering, and trading cost estimation to bring the strategy closer to live trading conditions
Risk assessment
Overall risk: Caution
- High neural network overfitting risk, with potentially large gap between backtesting and live performance
- Lacks market regime shift detection, model may incur consecutive losses during trend reversals
- Fixed stop-loss and take-profit do not account for market volatility changes, potentially causing unexpected losses in extreme conditions
- Improper training data window selection may cause the model to learn incorrect patterns
Evaluation scores
Parameters
| Parameter | Default | Description |
|---|---|---|
InputLayerSize | N/A | Neural network input layer nodes, determines feature dimensions |
HiddenLayerSize | N/A | Hidden layer nodes, affects model complexity |
OutputLayerSize | N/A | Output layer nodes, typically equals number of classes |
LearningRate | N/A | Learning rate, controls weight update step size |
TrainPeriod | N/A | Training period, number of bars for backtesting |
LotSize | 0.1 | Lot size per trade |
StopLoss | N/A | Stop loss in points |
TakeProfit | N/A | Take profit in points |
BuyThreshold | 0.5 | buysignal threshold |
SellThreshold | 0.5 | sellsignal threshold |
Similar Machine Learning EAs
Backtests are historical simulations and do not guarantee future results. Not financial advice.