AI Optimization 🧠

Note: This is the initial release (V1) of the AI Clustering feature. Ongoing development will introduce further refinements and capabilities.

AI Optimization

The AI Optimization feature is the most advanced component of Infinity Algo V3.0. It uses machine learning algorithms to continuously test thousands of setting combinations in the background, automatically identifying and applying the most profitable configuration for current market conditions.

How the AI Works

The AI follows a powerful three-step optimization cycle:

  1. Simulate: Tests hundreds to thousands of parameter combinations within your chosen Sensitivity Range

  2. Evaluate: Scores each combination using your selected Performance Metric

  3. Apply: Automatically implements the best-performing configuration in real-time

AI Modes Available

  • AI Mode: Optimizes sensitivity and threshold parameters for adaptive signal generation

  • AI Sniper Mode: Combines AI optimization with precision entry algorithms for maximum accuracy

AI Settings Explained

🧠 Enable AI Optimization

Master switch for the AI engine. Must be enabled for AI and AI Sniper signal modes to function.

AI modes require enablement: To use AI or AI Sniper, turn ON 🧠 Enable AI Optimization. If OFF, AI signals won’t plot or alert.

🧠 AI Optimization Mode

Determines how the AI performs optimization:

  • Walk-Forward: Continuously reselects the best configuration every N bars (defined by Update Frequency). Adapts to changing market conditions in real-time. Recommended for live trading.

  • Static (Full History): Optimizes once using the first 4900 bars of historical data, then locks that configuration permanently. Provides consistent backtesting results with proper train/test split. Recommended for strategy validation and backtesting.

Mode Selection Guide:

Use Case
Recommended Mode
Why

Live Trading

Walk-Forward

Adapts to market changes

Backtesting

Static

Consistent, reproducible results

Strategy Development

Static

Eliminates lookahead bias

Market Analysis

Walk-Forward

Shows adaptation patterns

🔄 AI Update Frequency (Bars)

Controls how often the AI recalculates optimal settings (Walk-Forward mode only).

  • Lower Values (1-100): More responsive but computationally intensive

  • Medium Values (200-1000): Balanced performance and adaptation

  • Higher Values (1000-5000): Stable, efficient, recommended for most users

Note: In Static mode, this setting is ignored as optimization occurs only once.

🧪 Sensitivity Range

Defines the parameter space the AI explores:

  • Auto: Full range (5-28) - Most comprehensive but slower

  • Very Fast (5-9): Ultra-responsive for scalping

  • Fast (10-14): Active day trading

  • Balanced (10-20): Recommended - Optimal for most strategies

  • Medium (15-21): Swing trading focus

  • Slow (22-28): Position trading and trend following

📈📉 AI Sim Long/Short TP %

Internal simulation parameters only - These define profit targets used by the AI to evaluate strategies during backtesting. They do NOT affect your actual trades or create real TP orders.

  • Default: 1.0% for both directions

  • Adjust based on your typical profit targets for more accurate optimization

📊 AI Performance Metric

Determines how the AI ranks and selects the best configuration:

Classic Metrics:

  • Total Profit: Raw cumulative profit

  • Average Profit: Profit per trade

  • Win Rate: Percentage of winning trades

  • GPR (Gross Profit Ratio): Profit efficiency ratio

  • GPR × √Trades: Balanced profit and trade count

Advanced Risk-Adjusted Metrics (New):

  • Sharpe Ratio: Return relative to volatility

  • Sortino Ratio: Return relative to downside risk

  • Calmar Ratio: Return relative to maximum drawdown

  • SQN (System Quality Number): Statistical system quality

  • Martin Ratio: Risk-adjusted return using Ulcer Index

Composite Metrics (New):

  • Sortino + Calmar Composite: Balanced risk-adjusted performance

  • Robust ML Score: Machine learning score resistant to outliers

Trading Style
Recommended Metric
Why

Scalping

Win Rate or GPR × √Trades

Consistency matters most

Day Trading

Sharpe Ratio or Total Profit

Balance risk and return

Swing Trading

Sortino Ratio or Calmar Ratio

Downside protection important

Position Trading

Robust ML Score or Composite

Long-term stability

High Frequency

SQN or Average Profit

System quality crucial

Performance Considerations

Computational Limits

  • Maximum historical lookback: 4900-5000 bars

  • Processing time: Increases with lower Update Frequency (Walk-Forward mode)

  • Static mode: Single computation at bar 4900, then no further processing

  • Timeframe impact: Lower timeframes handle more complex calculations

Optimization Tips

  1. Start Conservative: Use default settings (Static mode for testing, Walk-Forward for live)

  2. Monitor Performance: Check the dashboard for current AI selections

  3. Adjust Gradually: Make small changes to Update Frequency (Walk-Forward mode)

  4. Match Your Style: Choose metrics aligned with your goals

  5. Be Patient: AI needs 50+ bars minimum to establish patterns

Dashboard Features

When enabled, displays:

  • Current optimal sensitivity period

  • Selected threshold levels

  • Simulated performance metrics

  • Trade count and win rate

  • Configuration confidence score

  • Mode indicator: Shows "STATIC (LOCKED)", "OPTIMIZING", or "SIMULATED"

Troubleshooting

Indicator Timeout? → Use Static mode or increase Update Frequency

No AI Signals? → Ensure "Enable AI Optimization" is checked

Poor Performance? → Try different Performance Metrics or wider Sensitivity Range

Dashboard Not Showing? → Enable "Show Optimization Dashboard" in settings

Static Mode Not Working? → Ensure you have at least 5000 bars of historical data loaded

Summary

AI Optimization transforms Infinity Algo into a self-improving system that adapts to market conditions automatically. Choose Walk-Forward mode for live trading adaptation or Static mode for consistent backtesting. By selecting appropriate settings for your trading style and allowing the AI sufficient time to learn, you can achieve consistently optimized performance without manual parameter tuning.

Last updated