How the AI Makes Trading Decisions
The 7N7D trading agent uses multiple AI models and data sources to make informed trading decisions. Here's an inside look at how it works.
Decision Framework
Every trade goes through a structured analysis:
Market Data ──► Analysis ──► AI Decision ──► Risk Check ──► Execute
│ │ │ │
▼ ▼ ▼ ▼
Prices Patterns Confidence Position
Volume Sentiment Direction Sizing
News Indicators Entry/Exit Leverage
Analysis Inputs
Technical Analysis
- RSI (Relative Strength Index) - Overbought/oversold detection
- MACD - Trend direction and momentum
- Bollinger Bands - Volatility and price channels
- Chart Patterns - Historical pattern recognition
Market Regime Detection
The AI identifies three market conditions:
| Regime | Characteristics | Strategy |
|---|---|---|
| BULL | Uptrend, high momentum | Trend following, higher leverage |
| BEAR | Downtrend, risk-off | Shorter positions, lower exposure |
| NEUTRAL | Sideways, low volatility | Range trading, mean reversion |
Sentiment Analysis
- News headlines and social media
- Funding rates on exchanges
- Open interest changes
- Whale wallet monitoring
CME Gap Analysis
Institutional data from CME Bitcoin futures helps predict weekend gaps and institutional sentiment.
The AI Engine
7N7D uses DeepSeek AI for decision-making, chosen for:
- Reasoning ability - Complex market analysis
- Fast inference - Real-time decisions
- Cost efficiency - Sustainable operations
How Decisions Are Made
- Gather Data - All market data collected
- Build Context - AI receives market summary
- Generate Analysis - AI reasons about conditions
- Output Decision - Direction, confidence, timing
- Apply Constraints - Risk management limits applied
- Execute - Trade sent to Hyperliquid
Confidence Scoring
Every decision includes a confidence score (0-100%):
| Confidence | Action |
|---|---|
| 80%+ | Full position size |
| 60-80% | Reduced position |
| 40-60% | Minimal position |
| Below 40% | No trade |
Risk Management
The AI follows strict risk rules:
Position Sizing
Position Size = Account Balance × Risk Factor × Confidence
Where:
- Risk Factor = 0.5x to 1.5x (based on recent performance)
- Confidence = AI's conviction level
Leverage Limits
- Conservative: 3-5x leverage
- Normal: 5-10x leverage
- Aggressive: 10-20x leverage (only in high-confidence setups)
Stop-Losses
Every position has automated stop-losses based on:
- Technical support/resistance levels
- Maximum drawdown limits (typically 5-10%)
- Time-based exits for stale positions
Self-Improvement
What makes 7N7D unique is the AI improves itself:
Daily Strategy Loop (4AM UTC)
- Reviews previous 24 hours of trades
- Identifies patterns in winners/losers
- Generates code improvements
- Tests changes automatically
- Deploys if tests pass
Learning Metrics
The AI tracks and optimizes for:
- Win rate (target: 65%+)
- Profit factor (target: 1.5+)
- Sharpe ratio (target: 2.0+)
- Maximum drawdown (limit: 20%)
Failure Prevention
The agent learns from mistakes:
- Maintains database of failed patterns
- Checks new trades against past failures
- Avoids repeating the same errors
Performance Tracking
All decisions are logged for transparency:
{
"timestamp": "2025-12-06T14:30:00Z",
"asset": "BTC-PERP",
"direction": "LONG",
"confidence": 78,
"reasoning": "Bullish divergence on RSI...",
"entry_price": 98500,
"target": 101000,
"stop_loss": 97000
}
Investors can view all decisions in the dashboard.
Next: Learn about Transparency to see how everything is verified.