Analytics Suite

Comprehensive analytics framework for backtest analysis, risk assessment, performance attribution, and trade diagnostics.

Overview

Purpose: Understand strategy performance beyond basic returns: - Risk Metrics: VaR, CVaR, stress tests, tail risk, beta analysis - Performance Attribution: Alpha/beta, factor exposures, timing/selection skill - Trade Analysis: Entry/exit quality, MAE/MFE, holding periods, cost impact - Visualization: Charts, heatmaps, distributions, timelines - Report Generation: Comprehensive backtest reports

When to Use: - ✅ After backtest completion for comprehensive analysis - ✅ Before deployment to understand risk profile - ✅ For investor reporting and documentation - ✅ To diagnose strategy weaknesses and improve execution

---

Quick Start

Risk Analysis

from rustybt.analytics.risk import RiskAnalytics
import pandas as pd

# Load backtest results
backtest_df = pd.read_parquet("backtest_results.parquet")  # Must have 'returns' or 'portfolio_value'

# Initialize risk analytics
risk = RiskAnalytics(
    backtest_result=backtest_df,
    confidence_levels=[0.95, 0.99],  # VaR confidence levels
    benchmark_returns=spy_returns,    # Optional: for beta analysis
    positions=positions_df            # Optional: for risk decomposition
)

# Calculate Value at Risk
var_results = risk.calculate_var(method='historical')
print(f"95% VaR: {var_results['var_95']}")  # Max expected loss at 95% confidence
print(f"99% VaR: {var_results['var_99']}")  # Max expected loss at 99% confidence

# Calculate Conditional VaR (tail risk)
cvar_results = risk.calculate_cvar(method='historical')
print(f"95% CVaR: {cvar_results['cvar_95']}")  # Average loss beyond VaR

# Run stress tests
stress_results = risk.run_stress_tests()
print(f"2008 Crisis Loss: {stress_results['2008_financial_crisis']}")
print(f"COVID Crash Loss: {stress_results['covid_crash']}")

# Comprehensive risk report
risk_report = risk.analyze_risk()
print(risk_report['var'])
print(risk_report['cvar'])
print(risk_report['tail_risk'])
print(risk_report['beta'])  # If benchmark provided

# Visualize risk
risk.plot_var_distribution(method='historical', confidence=0.95)
risk.plot_stress_test_results()

---

Performance Attribution

from rustybt.analytics.attribution import PerformanceAttribution

# Initialize attribution analyzer
attrib = PerformanceAttribution(
    backtest_result=backtest_df,
    benchmark_returns=spy_returns,
    factor_returns=ff_factors_df,  # Optional: Fama-French factors
    risk_free_rate=0.02             # Optional: risk-free rate (2%)
)

# Comprehensive attribution analysis
results = attrib.analyze_attribution()

# Alpha and beta
print(f"Alpha: {results['alpha_beta']['alpha']:.4f}")
print(f"Beta: {results['alpha_beta']['beta']:.4f}")
print(f"Alpha significant: {results['alpha_beta']['alpha_significant']}")
print(f"Information Ratio: {results['alpha_beta']['information_ratio']:.2f}")

# Factor attribution (if factors provided)
if 'factor_attribution' in results:
    print("\nFactor Loadings:")
    print(results['factor_attribution']['factor_loadings'])
    print(f"\nR-squared: {results['factor_attribution']['r_squared']:.3f}")

# Timing attribution
if 'timing' in results:
    print(f"\nTiming ability: {results['timing']['timing_coefficient']:.4f}")
    print(f"Timing significant: {results['timing']['timing_significant']}")

# Visualize attribution
attrib.plot_alpha_over_time()
attrib.plot_factor_exposures()

---

Trade Analysis

from rustybt.analytics.trade_analysis import TradeAnalyzer

# Initialize trade analyzer
analyzer = TradeAnalyzer(backtest_result)

# Analyze all trades
analysis = analyzer.analyze_trades()

# Summary statistics
print(f"Total trades: {analysis['summary_stats']['total_trades']}")
print(f"Win rate: {analysis['summary_stats']['win_rate']:.2%}")
print(f"Profit factor: {analysis['summary_stats']['profit_factor']:.2f}")
print(f"Average win: ${analysis['summary_stats']['avg_win']}")
print(f"Average loss: ${analysis['summary_stats']['avg_loss']}")
print(f"Largest win: ${analysis['summary_stats']['largest_win']}")
print(f"Largest loss: ${analysis['summary_stats']['largest_loss']}")

# Entry/exit quality
print(f"\nAverage MAE: {analysis['mae_mfe']['avg_mae']:.2%}")
print(f"Average MFE: {analysis['mae_mfe']['avg_mfe']:.2%}")

# Holding period distribution
print(f"\nAverage holding period: {analysis['holding_period']['avg_holding_hours']} hours")

# Cost impact
print(f"\nTotal commission: ${analysis['costs']['total_commission']}")
print(f"Total slippage: ${analysis['costs']['total_slippage']}")
print(f"Commission impact: {analysis['costs']['commission_pct_of_pnl']:.2%}")

# Visualizations
analyzer.plot_mae_vs_pnl()  # MAE scatter plot
analyzer.plot_mfe_vs_pnl()  # MFE scatter plot
analyzer.plot_trade_timeline()  # Trade timeline
analyzer.plot_holding_period_distribution()  # Holding periods

---

Core Modules

Risk Analytics

• Module: rustybt.analytics.risk
• Purpose: VaR, CVaR, stress testing, beta analysis, tail risk
• Key Features: Multiple VaR methods, scenario analysis, risk decomposition
• Documentation: Risk Metrics

Performance Attribution

• Module: rustybt.analytics.attribution
• Purpose: Decompose returns into alpha, beta, factors, timing, selection
• Key Features: Multi-factor models, rolling attribution, timing tests
• Documentation: Performance Attribution

Trade Analysis

• Module: rustybt.analytics.trade_analysis
• Purpose: Trade-level diagnostics and execution quality analysis
• Key Features: MAE/MFE, entry/exit quality, cost impact, trade clustering
• Documentation: Trade Analysis

Visualization

• Module: rustybt.analytics.visualization
• Purpose: Charts, plots, heatmaps for backtest visualization
• Key Features: Returns distribution, equity curve, drawdown chart, correlation heatmap
• Documentation: Visualization Tools

Report Generation

• Module: rustybt.analytics.reports
• Purpose: Generate comprehensive backtest reports
• Key Features: PDF/HTML export, customizable templates, multi-strategy comparison
• Documentation: Report Generation

---

Complete Analysis Workflow

from rustybt.analytics.risk import RiskAnalytics
from rustybt.analytics.attribution import PerformanceAttribution
from rustybt.analytics.trade_analysis import TradeAnalyzer
from rustybt.analytics.reports import ReportGenerator
import pandas as pd

# Run backtest
backtest_result = run_backtest(strategy, data)

# 1. Risk Analysis
risk = RiskAnalytics(
    backtest_result=backtest_result.to_dataframe(),
    confidence_levels=[0.95, 0.99],
    benchmark_returns=spy_returns
)

risk_report = risk.analyze_risk()
print("\n=== Risk Analysis ===")
print(f"95% VaR: {risk_report['var']['var_95']}")
print(f"95% CVaR: {risk_report['cvar']['cvar_95']}")
print(f"Beta: {risk_report['beta']['beta']}")
print(f"Max drawdown: {risk_report['tail_risk']['max_loss_1d']}")

# 2. Performance Attribution
attribution = PerformanceAttribution(
    backtest_result=backtest_result.to_dataframe(),
    benchmark_returns=spy_returns
)

attrib_report = attribution.analyze_attribution()
print("\n=== Attribution Analysis ===")
print(f"Alpha: {attrib_report['alpha_beta']['alpha']}")
print(f"Beta: {attrib_report['alpha_beta']['beta']}")
print(f"Information Ratio: {attrib_report['alpha_beta']['information_ratio']}")

# 3. Trade Analysis
trade_analyzer = TradeAnalyzer(backtest_result)
trade_report = trade_analyzer.analyze_trades()

print("\n=== Trade Analysis ===")
print(f"Win rate: {trade_report['summary_stats']['win_rate']:.2%}")
print(f"Profit factor: {trade_report['summary_stats']['profit_factor']}")
print(f"Average MAE: {trade_report['mae_mfe']['avg_mae']:.2%}")
print(f"Average MFE: {trade_report['mae_mfe']['avg_mfe']:.2%}")

# 4. Generate Comprehensive Report
report_gen = ReportGenerator(
    backtest_result=backtest_result,
    risk_analysis=risk_report,
    attribution_analysis=attrib_report,
    trade_analysis=trade_report
)

report_gen.generate_report(
    output_path='backtest_report.html',
    format='html',
    include_plots=True
)

print("\n✅ Comprehensive report saved to backtest_report.html")

---

Key Metrics Reference

Risk Metrics

Metric	Description	Interpretation
VaR	Maximum expected loss at confidence level	Lower = less downside risk
CVaR	Average loss beyond VaR threshold	More conservative than VaR
Beta	Market sensitivity	1.0 = market, > 1.0 = more volatile
Skewness	Return distribution asymmetry	Negative = more extreme losses
Kurtosis	Fat tails indicator	High = more extreme events
Max Drawdown	Largest peak-to-trough decline	Lower = less severe losses
Downside Deviation	Volatility of negative returns	Lower = less downside risk

Performance Metrics

Metric	Description	Interpretation
Alpha	Excess return vs. benchmark	Positive = outperformance
Information Ratio	Alpha / tracking error	Higher = better risk-adjusted alpha
R-squared	Variance explained by benchmark	Higher = more market-driven
Tracking Error	Volatility of excess returns	Lower = closer to benchmark
Factor Loadings	Exposure to risk factors	Shows sources of return

Trade Metrics

Metric	Description	Interpretation
Win Rate	% of profitable trades	Higher = more consistent
Profit Factor	Gross profit / gross loss	> 1.0 = profitable
MAE	Max adverse excursion	Risk taken during trade
MFE	Max favorable excursion	Profit potential captured
Avg Win / Avg Loss	Reward/risk ratio	Higher = better risk/reward

---

Statistical Foundations

VaR Calculation Methods

1. Parametric VaR (assumes normal distribution):

VaR = μ + z * σ
where:
  μ = mean daily return
  σ = standard deviation of daily returns
  z = z-score for confidence level (e.g., -1.645 for 95%)

2. Historical VaR (empirical quantiles):

VaR = quantile(returns, 1 - confidence_level)

3. Monte Carlo VaR (simulation-based):

1. Simulate N returns from estimated distribution
2. VaR = empirical quantile of simulated returns

CVaR (Expected Shortfall)

CVaR = E[R | R ≤ VaR]
     = mean of returns below VaR threshold

Properties: - CVaR ≥ VaR (in absolute terms) - More conservative risk measure - Captures tail risk better than VaR

Alpha and Beta

Linear regression:

R_portfolio = α + β * R_benchmark + ε

where:
  α (alpha) = intercept (excess return)
  β (beta) = slope (market sensitivity)
  ε = error term (idiosyncratic risk)

Information Ratio:

IR = α / TE
where:
  TE (tracking error) = std(R_portfolio - R_benchmark)

MAE and MFE

MAE (Maximum Adverse Excursion):

For long:  MAE = max(0, (entry_price - min_price) / entry_price)
For short: MAE = max(0, (max_price - entry_price) / entry_price)

MFE (Maximum Favorable Excursion):

For long:  MFE = max(0, (max_price - entry_price) / entry_price)
For short: MFE = max(0, (entry_price - min_price) / entry_price)

---

Best Practices

✅ DO

1. Run comprehensive analysis after every backtest (risk + attribution + trade)
2. Use multiple VaR methods (parametric, historical, Monte Carlo) for robustness
3. Analyze tail risk (skewness, kurtosis) beyond VaR/CVaR
4. Review MAE/MFE to improve stop-loss and take-profit levels
5. Calculate beta to understand market dependency
6. Examine trade distribution for clustering and concentration risk
7. Document risk profile before deploying strategies

❌ DON'T

1. Rely on VaR alone (use CVaR and stress tests for tail risk)
2. Ignore statistical significance (check p-values for alpha)
3. Skip trade-level analysis (aggregate metrics hide execution issues)
4. Use parametric VaR for non-normal returns (use historical or Monte Carlo)
5. Forget transaction costs (commission/slippage impact profitability)
6. Ignore correlation risk (check asset correlation matrix)

---

Performance Considerations

Risk Analytics

• Time Complexity: O(n) for VaR/CVaR, O(n²) for correlation
• Memory: O(n) for returns storage
• Typical Runtime: < 1 second for 1000 observations

Performance Attribution

• Time Complexity: O(n) for alpha/beta, O(n × f) for factor models
• Memory: O(n × f) where f = number of factors
• Typical Runtime: 1-5 seconds for factor models with 1000 observations

Trade Analysis

• Time Complexity: O(t) where t = number of trades
• Memory: O(t) for trade storage
• Typical Runtime: < 1 second for 1000 trades

---

See Also

• Risk Metrics Documentation
• Performance Attribution
• Trade Analysis
• Visualization Tools
• Report Generation

---

Last Updated: 2025-10-16 | RustyBT v1.0