Back to Documentation

AI Scoring System

Complete guide to understanding Vibe Trading's AI scoring system, including how signals are generated, confidence levels, and performance metrics.

Overview

The AI Scoring System is the core intelligence engine that powers Vibe Trading's automated trading recommendations. It combines multiple machine learning models, technical analysis, sentiment analysis, and market data to generate high-confidence trading signals.

System Architecture

Core Components

  1. Data Ingestion Layer

    • Real-time market data feeds
    • Social media sentiment data
    • News and fundamental data
    • Technical indicator calculations

  2. Machine Learning Models

    • Price prediction models
    • Volatility forecasting
    • Pattern recognition
    • Risk assessment models

  3. Signal Generation Engine

    • Multi-model ensemble
    • Confidence scoring
    • Risk assessment
    • Time horizon analysis

  4. Output Layer

    • Trading recommendations
    • Confidence scores
    • Risk metrics
    • Explanatory reasoning

Scoring Methodology

Multi-Factor Analysis

The AI system evaluates multiple factors to generate comprehensive trading signals:

Technical Analysis (40% Weight)

  • Price Action: Support/resistance levels, trend analysis
  • Technical Indicators: RSI, MACD, Moving Averages, Bollinger Bands
  • Volume Analysis: Volume patterns, accumulation/distribution
  • Chart Patterns: Triangles, flags, head and shoulders, etc.

Sentiment Analysis (25% Weight)

  • Social Media Sentiment: Twitter, Reddit, Discord analysis
  • News Sentiment: Financial news and press releases
  • Market Sentiment: Fear & Greed Index, VIX, Put/Call ratios
  • Community Activity: Trading volume, user engagement

Fundamental Analysis (20% Weight)

  • Market Cap: Relative market position
  • Trading Volume: Liquidity and interest levels
  • Volatility: Historical and implied volatility
  • Correlation: Relationship with other assets

Machine Learning Models (15% Weight)

  • Price Prediction: LSTM neural networks
  • Pattern Recognition: CNN for chart patterns
  • Ensemble Methods: Random Forest, XGBoost
  • Deep Learning: Transformer models for sequence analysis

Confidence Scoring Algorithm

The confidence score is calculated using a weighted ensemble approach:

def calculate_confidence_score(factors):
    weights = {
        'technical_alignment': 0.30,
        'sentiment_consensus': 0.25,
        'volume_confirmation': 0.20,
        'model_agreement': 0.15,
        'risk_assessment': 0.10
    }

    confidence = 0
    for factor, weight in weights.items():
        confidence += factors[factor] * weight

    # Apply confidence modifiers
    confidence = apply_confidence_modifiers(confidence, factors)

    return min(max(confidence, 0), 100)  # Clamp between 0-100

Confidence Modifiers

Positive Modifiers

  • High Volume Confirmation: +5-10 points
  • Multiple Timeframe Alignment: +5-15 points
  • Strong Sentiment Consensus: +3-8 points
  • Model Agreement: +2-5 points

Negative Modifiers

  • Low Volume: -5-10 points
  • Conflicting Signals: -10-20 points
  • High Volatility: -3-8 points
  • Market Uncertainty: -5-15 points

Signal Types

Buy Signals

Strong Buy (90-100% Confidence)

Characteristics:

  • All technical indicators aligned
  • Strong volume confirmation
  • Positive sentiment across all sources
  • Multiple timeframe agreement
  • Low risk assessment

Example:

{
  "recommendation": "strong_buy",
  "confidence": 94,
  "reasoning": [
    "All technical indicators showing bullish momentum",
    "Volume 150% above average with strong accumulation",
    "Social sentiment extremely positive (0.85)",
    "Price breaking above key resistance with confirmation",
    "Risk assessment: Low (0.15)"
  ]
}

Buy (80-89% Confidence)

Characteristics:

  • Majority of indicators bullish
  • Good volume confirmation
  • Generally positive sentiment
  • Some timeframe agreement
  • Moderate risk assessment

Weak Buy (70-79% Confidence)

Characteristics:

  • Mixed technical signals
  • Moderate volume
  • Neutral to positive sentiment
  • Limited timeframe agreement
  • Higher risk assessment

Sell Signals

Strong Sell (90-100% Confidence)

Characteristics:

  • All technical indicators bearish
  • High volume selling pressure
  • Negative sentiment across sources
  • Multiple timeframe agreement
  • Low risk of false signals

Sell (80-89% Confidence)

Characteristics:

  • Majority of indicators bearish
  • Good volume confirmation
  • Generally negative sentiment
  • Some timeframe agreement
  • Moderate risk assessment

Weak Sell (70-79% Confidence)

Characteristics:

  • Mixed technical signals
  • Moderate selling volume
  • Neutral to negative sentiment
  • Limited timeframe agreement
  • Higher risk assessment

Hold Signals

Hold (60-69% Confidence)

Characteristics:

  • Mixed or neutral signals
  • Low conviction in direction
  • Conflicting indicators
  • High uncertainty
  • Wait for clearer signals

Time Horizons

Short-Term (1-7 days)

  • Focus: Intraday and swing trading
  • Indicators: 1m, 5m, 15m, 1h charts
  • Models: High-frequency price prediction
  • Risk: Higher volatility, faster changes

Medium-Term (1-4 weeks)

  • Focus: Position trading
  • Indicators: 4h, 1d charts
  • Models: Trend following and momentum
  • Risk: Moderate volatility, trend changes

Long-Term (1+ months)

  • Focus: Investment decisions
  • Indicators: 1d, 1w charts
  • Models: Fundamental and macro analysis
  • Risk: Lower volatility, major trend shifts

Risk Assessment

Risk Levels

Low Risk (0-25)

  • Characteristics: Strong signals, high volume, clear trends
  • Volatility: Below average
  • Drawdown Risk: Minimal
  • Recommended Position Size: 3-5% of portfolio

Medium Risk (26-50)

  • Characteristics: Good signals, moderate volume, some uncertainty
  • Volatility: Average
  • Drawdown Risk: Moderate
  • Recommended Position Size: 2-3% of portfolio

High Risk (51-75)

  • Characteristics: Weak signals, low volume, high uncertainty
  • Volatility: Above average
  • Drawdown Risk: Significant
  • Recommended Position Size: 1-2% of portfolio

Very High Risk (76-100)

  • Characteristics: Conflicting signals, very low volume, extreme uncertainty
  • Volatility: Very high
  • Drawdown Risk: High
  • Recommended Position Size: 0.5-1% of portfolio

Risk Factors

Market Risk

  • Volatility: Historical and implied volatility
  • Liquidity: Trading volume and spread
  • Correlation: Relationship with market indices

Model Risk

  • Model Uncertainty: Confidence in predictions
  • Data Quality: Reliability of input data
  • Overfitting: Model performance on unseen data

Execution Risk

  • Slippage: Price movement during execution
  • Timing: Optimal entry/exit timing
  • Market Impact: Effect of trade on market price

Performance Metrics

Signal Accuracy

Overall Accuracy

  • Target: >70% accuracy across all signals
  • Measurement: Percentage of profitable signals
  • Timeframe: Rolling 30-day window

Confidence-Based Accuracy

  • High Confidence (80%+): >85% accuracy target
  • Medium Confidence (60-79%): >70% accuracy target
  • Low Confidence (<60%): >55% accuracy target

Risk-Adjusted Returns

Sharpe Ratio

  • Target: >1.5 for high-confidence signals
  • Calculation: (Return - Risk-free rate) / Volatility
  • Timeframe: 90-day rolling window

Maximum Drawdown

  • Target: <15% for high-confidence signals
  • Measurement: Largest peak-to-trough decline
  • Timeframe: 90-day rolling window

Signal Distribution

Signal Frequency

  • High Confidence: 5-10 signals per day
  • Medium Confidence: 10-20 signals per day
  • Low Confidence: 20-30 signals per day

Asset Coverage

  • Major Pairs: BTC/USD, ETH/USD, etc.
  • Altcoins: Top 50 by market cap
  • Traditional Assets: Major stocks and indices

Model Performance

Individual Model Accuracy

Technical Analysis Model

  • Accuracy: 65-70%
  • Best Performance: Trending markets
  • Weakness: Sideways markets

Sentiment Analysis Model

  • Accuracy: 60-65%
  • Best Performance: High-volatility events
  • Weakness: Low-activity periods

Machine Learning Models

  • LSTM Price Prediction: 70-75%
  • CNN Pattern Recognition: 65-70%
  • Ensemble Methods: 72-78%

Ensemble Performance

  • Combined Accuracy: 75-80%
  • Risk Reduction: 20-30% lower volatility
  • Consistency: More stable performance

Signal Validation

Backtesting Results

Historical Performance (2023)

  • Overall Accuracy: 76.3%
  • High Confidence Signals: 84.7% accuracy
  • Average Return: 12.4% annually
  • Sharpe Ratio: 1.67
  • Maximum Drawdown: 8.2%

Asset-Specific Performance

  • Bitcoin: 78.1% accuracy, 15.2% return
  • Ethereum: 74.8% accuracy, 11.8% return
  • Major Altcoins: 72.3% accuracy, 9.4% return

Live Performance Tracking

Real-Time Metrics

  • Signal Generation: Every 5 minutes
  • Performance Updates: Daily
  • Model Retraining: Weekly
  • System Optimization: Monthly

Performance Dashboard

  • Accuracy Tracking: Rolling 7, 30, 90-day windows
  • Return Analysis: By signal type and confidence level
  • Risk Metrics: Volatility, drawdown, correlation
  • Model Health: Individual model performance

Continuous Improvement

Model Updates

Retraining Schedule

  • Daily: Incremental updates with new data
  • Weekly: Full model retraining
  • Monthly: Architecture optimization
  • Quarterly: Complete system review

Performance Monitoring

  • Drift Detection: Monitor model performance degradation
  • A/B Testing: Compare new models against current
  • Feature Engineering: Identify new predictive features
  • Hyperparameter Tuning: Optimize model parameters

Feedback Loop

User Feedback

  • Signal Rating: Users rate signal quality
  • Outcome Tracking: Track actual vs predicted outcomes
  • Behavioral Analysis: Understand user trading patterns
  • Preference Learning: Adapt to user risk tolerance

Market Adaptation

  • Regime Detection: Identify market regime changes
  • Model Switching: Use different models for different regimes
  • Parameter Adjustment: Adapt to changing market conditions
  • Risk Management: Adjust risk parameters dynamically

API Integration

Signal Endpoints

Get Current Signal

GET /api/v1/signals/{symbol}

Response:

{
  "symbol": "BTC/USD",
  "recommendation": "buy",
  "confidence": 87,
  "timeHorizon": "short_term",
  "riskLevel": "medium",
  "reasoning": [
    "Strong bullish momentum detected",
    "High volume confirmation",
    "Positive sentiment from social media"
  ],
  "technicalAnalysis": {
    "rsi": 35.2,
    "macd": "bullish",
    "support": 48500.0,
    "resistance": 50200.0
  },
  "sentimentAnalysis": {
    "socialSentiment": 0.75,
    "newsSentiment": 0.68,
    "fearGreedIndex": 45
  },
  "timestamp": "2024-01-15T12:00:00Z"
}

Get Signal History

GET /api/v1/signals/{symbol}/history

Get Performance Metrics

GET /api/v1/signals/performance

WebSocket Integration

Real-Time Signals

{
  "type": "signal",
  "symbol": "BTC/USD",
  "recommendation": "buy",
  "confidence": 87,
  "timeHorizon": "short_term",
  "timestamp": "2024-01-15T12:00:00Z"
}

Best Practices

Using AI Signals

Signal Interpretation

  • High Confidence: Strong conviction, larger position sizes
  • Medium Confidence: Moderate conviction, standard position sizes
  • Low Confidence: Weak conviction, smaller position sizes

Risk Management

  • Never Risk More: Than you can afford to lose
  • Diversify Signals: Don't rely on single signals
  • Use Stop Losses: Always protect against losses
  • Monitor Performance: Track signal accuracy over time

Position Sizing

  • Conservative: 1-2% of portfolio per signal
  • Moderate: 2-3% of portfolio per signal
  • Aggressive: 3-5% of portfolio per signal

Signal Validation

Cross-Reference

  • Multiple Timeframes: Check signals across timeframes
  • Different Assets: Compare related assets
  • Market Context: Consider overall market conditions
  • News Events: Factor in upcoming events

Performance Tracking

  • Keep Records: Track all signal outcomes
  • Analyze Patterns: Identify successful signal types
  • Adjust Strategy: Modify approach based on results
  • Continuous Learning: Improve signal interpretation

Ready to understand signal types? Check out our Signal Types guide or explore Confidence Levels.