Backtest Analysis Skill

Triggering Contexts

• User shares TradingView strategy tester results (pasted or file)
• User asks "Is this strategy good?" or "Should I trade this?"
• User wants to compare strategy versions
• After implementing changes from coding-pinescript

Input Handling

Accept Any Format

Users may provide backtest data as:

1. Raw paste from TradingView Performance Summary (messy but parseable)
2. Excel export from TradingView (complete multi-sheet data)
3. CSV trade list (individual trade data)
4. Screenshot (extract visually)

Semantic Parsing Rules

Do NOT rely on exact formatting. Look for these concepts flexibly:

Value Extraction

After finding a metric name, extract:

• Number with $ or USD → Dollar value
• Number with % → Percentage
• Plain number → Raw value

Fallback

If parsing fails, ask user to confirm key metrics directly.

Core Analysis Framework

Step 1: Extract & Validate Metrics

Must Have (required for analysis):

• Net Profit ($ and %)
• Total Trades
• Win Rate (%)
• Profit Factor
• Max Drawdown ($ and %)
• Avg Win / Avg Loss

Nice to Have:

• Sharpe Ratio
• Sortino Ratio
• CAGR
• Avg Bars in Trade

Step 2: Sanity Checks

Flag if any of these are true:

• Profit Factor < 0 or > 10 → Parsing error or suspicious
• Win Rate > 100% or < 0% → Parsing error
• Total Trades < 30 → Statistically insignificant
• Max DD > Net Profit → Strategy may not be viable

Step 3: Quality Assessment

Statistical Significance

Profit Factor Interpretation

Drawdown Analysis

Expectancy Check

Expectancy = (Win% × Avg Win) - (Loss% × Avg Loss)

Must be positive. Calculate per-trade expected value.

Step 4: Red Flag Detection

Overfitting Indicators:

• Very high PF (> 3.0) with low trade count (< 100)
• Win rate extremely high (> 80%) for trend strategy
• Tested on single asset only
• No out-of-sample validation mentioned

Survivorship Bias:

• Only tested on assets that "worked"
• No mention of failed variations

Curve Fitting:

• Many optimized parameters
• Strategy logic too complex for trade count

Step 5: Probing Questions

Always ask:

1. "Was this tested on out-of-sample data (walk-forward or separate period)?"
2. "How many parameters were optimized?"
3. "Does this logic make sense fundamentally, or is it just pattern-matched?"

Output Format

Summary Card

📊 BACKTEST ANALYSIS: [Strategy Name]

Edge Quality: [STRONG/MODERATE/WEAK/NONE]
Trade Count: [N] trades → [Significance assessment]
Profit Factor: [X] → [Interpretation]
Max Drawdown: [X%] → [Risk assessment]
Expectancy: $[X] per trade

⚠️ WARNINGS:
- [List any red flags]

✅ STRENGTHS:
- [List positive aspects]

📋 RECOMMENDATIONS:
1. [Actionable next steps]

Verdict Categories

Cross-References

• After analysis, if improvements needed → coding-pinescript
• If strategy needs redesign → planning-trading-systems
• Track changes between versions → iterating-strategies