---

name: analyzing-backtests description: Analyzes algorithmic trading backtest results from Jupyter notebooks and generates summary reports. Use when the user wants to analyze or summarize backtest notebooks. allowed-tools: Read, Bash, Glob, Grep

Backtest Analysis Skill

Analyze a Jupyter notebook containing algorithmic trading backtest results and generate a comprehensive summary report.

Analysis Steps

1. Version Control Information

   • Run git status to check current state
   • Run git log -1 --format="%H %ci" for latest commit hash and date
   • Note any uncommitted changes

2. Read the Notebook

   • Use Read tool to load the specified .ipynb file
   • Parse cells for code, markdown, and outputs

3. Extract Key Information

   Model/Strategy Details:

   • Strategy name, type, and configuration
   • Key hyperparameters
   • Training and testing period information

   Date Coverage:

   • Backtest period (start, end, duration)

   Performance Metrics:

   • Monetary results: returns, capital, drawdowns, trade statistics
   • Statistical analysis: risk metrics, benchmark comparisons, distributions
   • Extract whatever metrics are available in the notebook

4. Generate Report

Output a structured markdown report:

Copy# Backtest Analysis Report

**Notebook:** [filename]
**Generated:** [date]
**Git Commit:** [hash] ([date])
**Uncommitted Changes:** [yes/no]

## Strategy
[Name and brief description]

**Configuration:**
- [Key parameters]

## Period
- **Dates:** [start] to [end] ([duration])

## Performance

| Metric | Value | Benchmark |
|--------|-------|-----------|
| Total Return | X% | X% |
| Annualized Return | X% | X% |
| Max Drawdown | X% | X% |
| Sharpe Ratio | X.XX | X.XX |
| Win Rate | X% | - |
| Total Trades | X | - |

## Risk Metrics
| Metric | Value |
|--------|-------|
| Volatility | X% |
| Alpha | X% |
| Beta | X.XX |

## Key Findings
- [Notable observations]
- [Strengths and weaknesses]

## Concerns/Recommendations
- [Any issues or suggestions]

Instructions

• Extract all available metrics from the notebook
• Mark unavailable metrics as "N/A"
• Provide brief analysis, not just data
• Flag unusual results or potential issues
• Keep report concise but comprehensive