strategy-backtest

Strategy Backtest Skill

This skill runs a full backtest on any named strategy and compares it against the default baseline.

Usage

Invoke with: /strategy-backtest <strategy_name> [bars]

Examples:

• /strategy-backtest scalp_1m_sweep
• /strategy-backtest scalp_1m_sweep 2000

Parameters

• strategy_name (required): Name of strategy file in config/strategies/ (without .json)
• bars (optional): Number of bars to backtest (default: 1000)

Steps

// turbo-all

Step 1: Validate Strategy Exists

$strategyPath = "config/strategies/$strategyName.json"
if (-not (Test-Path $strategyPath)) {
    Write-Error "Strategy not found: $strategyPath"
    exit 1
}
Write-Host "Strategy found: $strategyPath"

Step 2: Run Baseline Backtest

Run default strategy first to establish baseline:

$env:PYTHONPATH="src"; .\.venv\Scripts\python.exe -m src.laptop_agents.run --mode backtest --source mock --backtest $bars

Save metrics from runs/latest/stats.json as baseline.

Step 3: Run Target Strategy Backtest

$env:PYTHONPATH="src"; .\.venv\Scripts\python.exe -m src.laptop_agents.run --strategy $strategyName --mode backtest --source mock --backtest $bars

Step 4: Extract Metrics

Parse runs/latest/stats.json for:

• net_pnl: Net Profit/Loss
• max_drawdown: Maximum Drawdown
• trade_count: Number of trades
• win_rate: Win percentage
• sharpe_ratio: Risk-adjusted return (if available)

Step 5: Generate Comparison Report

Output a comparison table:

Step 6: Pass/Fail Determination

The strategy PASSES if:

• Net PnL > 0
• Max Drawdown < 10%
• Trade Count > 10

Output Artifacts

• Comparison table printed to console
• Full results available in runs/latest/

On Failure

• Report which criteria failed
• Suggest parameter adjustments if drawdown is too high