backtesting-trading-strategies

name	backtesting-trading-strategies
description	Backtest crypto and traditional trading strategies against historical data. Calculates performance metrics (Sharpe, Sortino, max drawdown), generates equity curves, and optimizes strategy parameters. Use when user wants to test a trading strategy, validate signals, or compare approaches. Trigger with phrases like "backtest strategy", "test trading strategy", "historical performance", "simulate trades", "optimize parameters", or "validate signals".
allowed-tools	Read, Write, Edit, Grep, Glob, Bash(python:*)
version	2.0.0
author	Jeremy Longshore <[e-mail]>
license	MIT

Backtesting Trading Strategies

Overview

Validate trading strategies against historical data before risking real capital. This skill provides a complete backtesting framework with 8 built-in strategies, comprehensive performance metrics, and parameter optimization.

Key Features:

• 8 pre-built trading strategies (SMA, EMA, RSI, MACD, Bollinger, Breakout, Mean Reversion, Momentum)
• Full performance metrics (Sharpe, Sortino, Calmar, VaR, max drawdown)
• Parameter grid search optimization
• Equity curve visualization
• Trade-by-trade analysis

Prerequisites

Install required dependencies:

pip install pandas numpy yfinance matplotlib

For A-share / ETF backtests (6-digit codes such as 510300 or 510300.SS), the scripts load history via this repo’s plugins/data_collection stack (AkShare, mootdx, project cache — same strategy as openclaw-data-china-stock). Install:

pip install akshare

Optional for advanced features:

pip install ta-lib scipy scikit-learn

Instructions

Step 1: Fetch Historical Data

Crypto / US (Yahoo):

python {baseDir}/scripts/fetch_data.py --symbol BTC-USD --period 2y --interval 1d --source yfinance

China ETF (plugin-aligned pipeline, default for --source auto when symbol is a 6-digit code):

python {baseDir}/scripts/fetch_data.py --symbol 510300.SS --period 2y --interval 1d --source auto

Data is cached to {baseDir}/data/{symbol}_{interval}.csv for reuse.

Step 2: Run Backtest

Basic backtest with default parameters:

python {baseDir}/scripts/backtest.py --strategy sma_crossover --symbol BTC-USD --period 1y --data-source yfinance

China ETF (uses plugins/data_collection automatically; no Yahoo):

python {baseDir}/scripts/backtest.py --strategy sma_crossover --symbol 510300.SS --period 1y

Advanced backtest with custom parameters:

# Example: backtest with specific date range
python {baseDir}/scripts/backtest.py \
  --strategy rsi_reversal \
  --symbol ETH-USD \
  --period 1y \
  --capital 10000 \
  --params '{"period": 14, "overbought": 70, "oversold": 30}'

Step 3: Analyze Results

Results are saved to {baseDir}/reports/ including:

• *_summary.txt - Performance metrics
• *_trades.csv - Trade log
• *_equity.csv - Equity curve data
• *_chart.png - Visual equity curve

Step 4: Optimize Parameters

Find optimal parameters via grid search:

python {baseDir}/scripts/optimize.py \
  --strategy sma_crossover \
  --symbol BTC-USD \
  --period 1y \
  --param-grid '{"fast_period": [10, 20, 30], "slow_period": [50, 100, 200]}'

Output

Performance Metrics

Metric	Description
Total Return	Overall percentage gain/loss
CAGR	Compound annual growth rate
Sharpe Ratio	Risk-adjusted return (target: >1.5)
Sortino Ratio	Downside risk-adjusted return
Calmar Ratio	Return divided by max drawdown

Risk Metrics

Metric	Description
Max Drawdown	Largest peak-to-trough decline
VaR (95%)	Value at Risk at 95% confidence
CVaR (95%)	Expected loss beyond VaR
Volatility	Annualized standard deviation

Trade Statistics

Metric	Description
Total Trades	Number of round-trip trades
Win Rate	Percentage of profitable trades
Profit Factor	Gross profit divided by gross loss
Expectancy	Expected value per trade

Example Output

================================================================================
                    BACKTEST RESULTS: SMA CROSSOVER
                    BTC-USD | [start_date] to [end_date]
================================================================================
 PERFORMANCE                          | RISK
 Total Return:        +47.32%         | Max Drawdown:      -18.45%
 CAGR:                +47.32%         | VaR (95%):         -2.34%
 Sharpe Ratio:        1.87            | Volatility:        42.1%
 Sortino Ratio:       2.41            | Ulcer Index:       8.2
--------------------------------------------------------------------------------
 TRADE STATISTICS
 Total Trades:        24              | Profit Factor:     2.34
 Win Rate:            58.3%           | Expectancy:        $197.17
 Avg Win:             $892.45         | Max Consec. Losses: 3
================================================================================

Supported Strategies

Strategy	Description	Key Parameters
sma_crossover	Simple moving average crossover	fast_period, slow_period
ema_crossover	Exponential MA crossover	fast_period, slow_period
rsi_reversal	RSI overbought/oversold	period, overbought, oversold
macd	MACD signal line crossover	fast, slow, signal
bollinger_bands	Mean reversion on bands	period, std_dev
breakout	Price breakout from range	lookback, threshold
mean_reversion	Return to moving average	period, z_threshold
momentum	Rate of change momentum	period, threshold

Configuration

Edit {baseDir}/config/settings.yaml (loaded automatically by backtest.py, optimize.py, fetch_data.py):

• data.provider: auto | china | yfinance | coingecko — used when CLI omits --data-source / --source and env BACKTEST_DATA_SOURCE is unset.
• data.cache_dir / reporting.output_dir: paths relative to the skill root unless absolute.
• BACKTEST_SKILL_SETTINGS: optional path to a different YAML (same schema).
• Precedence: CLI → BACKTEST_DATA_SOURCE → YAML data.provider → auto.

data:
  provider: auto
  cache_dir: ./data
  default_interval: 1d

backtest:
  default_capital: 10000
  commission: 0.001     # 0.1% per trade
  slippage: 0.0005      # 0.05% slippage

reporting:
  output_dir: ./reports

risk:
  max_position_size: 0.95
  stop_loss: null       # Optional fixed stop loss
  take_profit: null     # Optional fixed take profit

Implementation: {baseDir}/scripts/skill_settings.py.

Verification (smoke tests)

Full command log and ops notes live in the assistant repo: docs/ops/回测使用指导-自动任务与日常交互.md → §7（脚本实测摘要；全文仅覆盖本 Skill，不含涨停回马枪工具）。

Check	Command (from {baseDir}/scripts/)	Expected
List strategies	python3 backtest.py --list	Prints 8 strategies
SMA needs long history	python3 backtest.py --strategy sma_crossover --symbol 510300.SS --period 90d	Exit 1, message about ≥205 bars
CN ETF RSI	python3 backtest.py --strategy rsi_reversal --symbol 510300.SS --period 1y --quiet	Exit 0; data via plugins/data_collection
Crypto Yahoo	python3 backtest.py --strategy rsi_reversal --symbol BTC-USD --period 120d --data-source yfinance --quiet	Exit 0
Crypto CoinGecko	python3 backtest.py --strategy rsi_reversal --symbol ETH-USD --period 90d --data-source coingecko --quiet	Exit 0
Fetch CN	python3 fetch_data.py --symbol 510300.SS --period 120d	Exit 0; writes data/510300.SS_1d.csv
Fetch Yahoo	python3 fetch_data.py --symbol BTC-USD --period 60d --source yfinance	Exit 0
Tiny optimize grid	python3 optimize.py --strategy rsi_reversal --symbol 510300.SS --period 1y --param-grid '{"period":[12,14]}'	Exit 0
Repo wrapper	From scripts/: python3 ../../../scripts/run_backtest_trading_strategies.py --strategy rsi_reversal --symbol 510300 --period 6m --quiet (or use repo absolute path to that script)	Exit 0; same as backtest.py

Lightweight Backtest (MVP mode)

When users ask for “quick backtest / strategy optimization / Sharpe / max drawdown” but the environment lacks a dedicated tool_backtest_*, run a lightweight backtest by combining:

• tool_fetch_market_data (to get sufficient historical OHLCV)
• tool_calculate_technical_indicators (to derive signals/features)

Mandatory rules (MVP mode):

• If history is insufficient to evaluate the described strategy, return insufficient_evidence.
• Explicitly state this is MVP mode and describe capability boundaries.
• Do not output specific buy/sell points, position sizing, or leverage advice.
• Parameter search ranges must be read from config/strategy-backtester_config.yaml (skill-local config).

Output structure (fixed):

1. strategy_spec and backtest_window\n2) performance and risk_metrics\n3) trade_stats and parameter_sensitivity\n4) limitations and next_experiments

Remember: 510300 vs 510300.SS → two different cache files under data/; pick one symbol style per workflow.

Error Handling

See {baseDir}/references/errors.md for common issues and solutions.

Examples

See {baseDir}/references/examples.md for detailed usage examples including:

• Multi-asset comparison
• Walk-forward analysis
• Parameter optimization workflows

Files

File	Purpose
scripts/backtest.py	Main backtesting engine
scripts/china_stock_loader.py	CN ETF path via plugins/data_collection
scripts/skill_settings.py	YAML + env + CLI precedence for paths and data.provider
scripts/fetch_data.py	Historical data fetcher
scripts/strategies.py	Strategy definitions
scripts/metrics.py	Performance calculations
scripts/optimize.py	Parameter optimization

Resources

• yfinance - Yahoo Finance data
• TA-Lib - Technical analysis library
• QuantStats - Portfolio analytics