🧠AI Optimization

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

Transform Infinity Algo into a self-improving system that adapts to markets automatically.

Quick Start: Enable AI → Select mode (Static/Walk-Forward) → Choose metric → Let AI optimize

---

🚀 Quick Setup

1

Enable AI

Turn ON 🧠 Enable AI Optimization in settings

2

Choose Mode

• Backtesting? → Static (Full History)

• Live Trading? → Walk-Forward

3

Select Signal Type

Choose AI or AI Sniper in Signal Mode

That's it! Default settings work for most users.

---

🎯 How AI Works

1️⃣ Simulate	2️⃣ Evaluate	3️⃣ Apply
Tests 100s-1000s of parameter combinations	Scores each using your metric	Implements best configuration

Walk-Forward: Periodically re-optimizes on a rolling in-sample window and validates out-of-sample, reducing overfitting.

---

⚙️ Core Settings

🧠 AI Modes

Optimization Modes

Mode	How It Works	Use For
Walk-Forward	Updates every N bars	Live trading
Static	Optimizes once, locks	Backtesting

Tip: Start with Static for testing, switch to Walk-Forward for live

🔄 Update Frequency

Walk-Forward Only

frequency-guide.txt

100 bars (default) → Ultra-responsive (high CPU)
200-1000 bars     → Balanced ✅
1000-5000 bars    → Very stable, slower to adapt

Examples on 1h chart:
- 100 bars = ~4 days
- 1000 bars = ~42 days
- 5000 bars = ~208 days

Lower = More responsive but intensive | Higher = More stable and efficient

🧪 Sensitivity Range

Parameter Space

Range	Values	Best For
Very Fast	5-9	Scalping
Fast	10-14	Day Trading
Balanced ✅	10-20	Most Strategies
Medium	15-21	Swing Trading
Slow	22-28	Position Trading
Auto	5-28	Full exploration

📊 Optimization Metric

Choose Your Goal

Quick Selection:

Your Style	Use This Metric	Why
Scalping	Win Rate	Consistency matters
Day Trading	Sharpe Ratio	Balance risk/return
Swing	Sortino Ratio	Downside protection
Position	Calmar Ratio	Avoid drawdowns

All Available Metrics:

• Classic: Total Profit, Win Rate, Average P&L, Gain-to-Pain

• Risk-Adjusted: Sharpe, Sortino, Calmar, Martin

• Advanced: SQN (System Quality Number), Robust ML Score

Important: High win rate ≠ profitability. A 90% win rate with large losses can be unprofitable.

Not sure? Use Total Profit for testing, Sharpe Ratio for live trading

---

📈 Simulation Settings

AI Sim TP% (Testing Only)

Note: These are internal simulation parameters - they do NOT create real orders

What they do:

• Help AI evaluate strategies

• Set internal profit targets

• Default: 1.0% both directions

sim-settings:

Long TP: 1.0%
Short TP: 1.0%
Purpose: AI testing only
Real trades: Not affected

---

📊 Dashboard Display

Live Monitoring

When enabled, see:

• ✅ Current optimal sensitivity

• ✅ Selected thresholds

• ✅ Win rate & metrics

• ✅ Confidence score

• ✅ Mode status

Status Indicators:

• STATIC (LOCKED) - One-time optimization complete

• OPTIMIZING - Currently calculating

• SIMULATED - Results ready

---

💡 Best Practices

🎯 Getting Started

1. Use Static for initial testing

2. Select Balanced sensitivity

3. Default 100 bar frequency

4. Match metric to goals

5. Walk-Forward needs ~535 bars for first optimization

6. Static needs ~5000 bars total

⚡ Optimization

• Lower timeframes → Complex calculations

• Monitor dashboard → Track selections

• Small adjustments → Better results

• Patience required → AI needs time

📈 Performance

Limits:

• Max lookback: 5000 bars

• Lower frequency = Higher CPU

• Static = One calculation only at bar 4900

• Higher TF = Better performance

---

🔧 Troubleshooting

Problem	Solution
Timeout	Use Static or increase frequency
No signals	Check AI Optimization is ON
Poor results	Try different metric/range
No dashboard	Enable in settings
Static fails	Need 5000+ bars data

---

⚡ Quick Reference

For Testing

• Mode: Static

• Range: Balanced

• Metric: Total Profit

• Frequency: N/A

• Min bars: 5000

For Live Trading

• Mode: Walk-Forward

• Range: Balanced

• Metric: Your preference

• Frequency: 100 (default)

• Min bars: 535

---

📚 Understanding Performance Metrics

Detailed Metric Explanations

Note: Infinity Algo computes metrics on per-trade returns with risk-free rate and MAR = 0. Industry definitions typically use time-series (daily/monthly) returns.

Classic Metrics

Metric	Formula	Best For
Total Profit	Sum of all P&L	Quick assessment
Win Rate	Wins ÷ Total trades × 100	Consistency check
Average P&L	Total P&L ÷ Trades	Trade quality
Gain-to-Pain	Σ gains / |Σ losses|	Risk/reward balance

---

Risk-Adjusted Metrics

Sharpe Ratio - Industry Standard

• Formula: Excess return (over risk-free) ÷ Standard deviation

• Infinity Algo: Uses risk-free = 0

• Pros: Most widely used, easy comparison, considers total volatility

• Cons: Penalizes upside volatility, assumes normal distribution

• Benchmarks: ~1 = Good | ~2 = Very good | 3+ = Outstanding

Sortino Ratio - Downside Focus

• Formula: Excess return (over target/MAR) ÷ Downside deviation

• Infinity Algo: Uses MAR = 0

• Pros: Only penalizes bad volatility, better for trend following

• Cons: Requires defining target return, less standardized

• Benchmarks: >1 = Good | >2 = Very good | >3 = Excellent

Calmar Ratio - Drawdown Protection

• Formula: CAGR ÷ Maximum drawdown (commonly 36 months)

• Pros: Focus on capital preservation, easy to understand

• Cons: Based on single worst event, backward-looking

• Benchmarks: >1 = Good | 3-5 = Strong

Martin Ratio - Ulcer Performance

• Formula: Excess return ÷ Ulcer Index (RMS of drawdowns)

• Pros: Considers all drawdowns, smooth equity curve focus

• Cons: Less known/comparable, complex calculation

• Use: Compare across your strategies

SQN - System Quality Number

• Formula: (Expectancy ÷ Std Dev) × √Number of trades

• Pros: Accounts for sample size, good for system comparison

• Cons: Requires sufficient trades for validity

• Benchmarks: >2 = Good | >3 = Excellent | >5 = Superb

---

Choosing by Trading Style

Style	Primary Metrics	Secondary Metrics
Scalping	Win Rate + Sharpe	Total Profit
Day Trading	Sharpe + Win Rate	Average P&L
Swing Trading	Sortino + Calmar	Gain-to-Pain
Position Trading	Calmar + Martin	Sortino

Remember: Win rate alone is misleading. A strategy with 30% win rate but 3:1 reward/risk is more profitable than 70% win rate with 1:3 reward/risk.

AI Mode Selection:

• Intraday/Mean-reversion: Optimizes for Sharpe + Win Rate

• Trend/Swing trading: Prioritizes Sortino + Calmar

• Multi-metric: Balances all metrics for robust performance

---

Bottom Line: Let AI handle optimization while you focus on trading decisions and risk management.