Mad Turtle

This expert advisor is an ensemble system of models trained on a space of synthesized high-dimensional features.

Key features of Mad Turtle:

• Deposit safety
  No dangerous strategies like grids or martingale.
  Supports single-order trading and short stop-loss.

• Transparency
  You can see in real time how the model “thinks” and what probability assessment it gives (BUY/SELL).
  You can also adjust the activation threshold.

• Fundamentals
  Full 24-hour format on higher timeframes.
  Does not use micro-scalping, pre-rollover, or other broker/execution-dependent strategies.

• Uniqueness
  Built on my unique feature engineering:
  Time series are transformed with Fourier analysis to extract dominant frequencies, tested for stationarity (Dickey-Fuller), and enriched with nonlinear metrics (higher-order moments, entropy characteristics).
  This allows the model to operate not on raw quotes but on abstract multidimensional patterns.
  To improve robustness and reduce prediction variance, outputs are filtered through a cascade of meta-models (logistic regression, SVM), performing Bayesian regularization of the final signal.

Launch

Takes 1 minute:

1. Attach the EA to an XAUUSD H1 chart.

2. Choose the risk — I recommend starting low and then increasing.

3. Choose the model that suits you best. A full description of models will be provided below.
   If you’re unsure, just message me — I’ll happily explain and help with setup and risk selection.
   Don’t forget to enable auto-trading, and that’s it.

Models

• The number of models is growing, and I plan to add more. That’s why I created a dedicated channel for detailed descriptions.
  Turtle Channel: www.mql5.com/en/channels/mad_turtle
  
  
• WT: https://www.mql5.com/en/signals/2323073?source=Site+Signals+My
  ICM:https://www.mql5.com/en/signals/2335965?source=Site+Signals+My
  myfxbook.com/members/lleon

What to note before buying

• Stop Loss is part of the strategy and not a sign of malfunction.
  Profitability should be judged over a full trading month, not a few days.

• No trades for several days is also not a malfunction.
  Sometimes the model pauses for a few days if it detects anomalies in the price structure.

• In the world of algo trading, it’s easy to get lost among EAs with “perfect” balance curves — usually the result of heavy "overfitting".
  My models are designed not for flawless backtests but for stable performance in real, unpredictable markets.
  If your goal is purely to test “ideal” algorithms, you can check academic examples on my website.
  There you’ll find grails that backtest perfectly since the year 1800.