Pattern Backtester

You are scanning historical price data to find past occurrences of a specific pattern and reporting what happened after each occurrence.

Step 1: Define the Pattern

Ask the trader to describe the setup, or infer from recent conversation. Examples:

• "RSI below 30 at a VPA support level"
• "Price tests EMA50 from above with declining volume"
• "Gap-up open that fills within first 30 minutes"
• "Engulfing candle after 3+ red bars"

Translate the description into quantifiable conditions:

Pattern: Bullish RSI Divergence at Support
Conditions:
1. RSI(14) < 35
2. Price within 0.5% of EMA50 (approaching from above)
3. Current bar volume < 20-bar average volume
4. Previous 2 bars were red (close < open)

Step 2: Scan Historical Data

1. data_get_ohlcv with count: 500 (maximum) — get as much history as possible
2. If indicators are needed, note that study values are only available for the CURRENT bar. For historical indicator values, compute them from OHLCV (RSI, EMA, volume avg are calculable from raw bars).

Computing Indicators from OHLCV

• EMA(n): Standard exponential moving average formula over close prices
• RSI(n): Standard RSI formula using gains/losses
• Volume Average: Simple average of volume over N bars
• ATR(n): Average of True Range over N bars
• Engulfing: Current bar body completely covers previous bar body

Pattern Matching

Iterate through the 500 bars. For each bar, check if ALL conditions are met. Record the bar index, date, and price when the pattern fires.

Step 3: Forward Analysis

For each pattern occurrence, examine what happened in the NEXT bars:

• +1 bar: Close vs pattern bar close (immediate reaction)
• +3 bars: High/Low range (short-term)
• +5 bars: Max favorable excursion (MFE) and max adverse excursion (MAE)
• +10 bars: Was the move sustained?
• +20 bars: Full swing completion

Calculate:

• Win rate: What % of occurrences led to a favorable move of at least 1 ATR?
• Average MFE: How far did price move in the favorable direction?
• Average MAE: How far did price move against before going favorable?
• Expectancy: Average P&L if you entered at the pattern signal

Step 4: Deliver the Report

## Pattern Backtest — [Pattern Name]
Symbol: [symbol] | Timeframe: [TF] | Data: [bar count] bars

### Pattern Definition
[Conditions list]

### Results
| Metric              | Value        |
|---------------------|--------------|
| Total Occurrences   | [N]          |
| Win Rate (1 ATR)    | [%]          |
| Avg MFE (fav move)  | +[pts]       |
| Avg MAE (adv move)  | -[pts]       |
| Best Outcome        | +[pts] on [date] |
| Worst Outcome       | -[pts] on [date] |
| Expectancy/Trade    | [+/- pts]    |

### Each Occurrence
| # | Date       | Entry  | +3 bars | +5 bars | +10 bars | Result |
|---|-----------|--------|---------|---------|----------|--------|
| 1 | [date]     | [price]| +25 pts | +40 pts | +55 pts  | WIN    |
| 2 | [date]     | [price]| -10 pts | -20 pts | -35 pts  | LOSS   |
| ...

### Statistical Confidence
- Sample size: [N] (minimum 10 for moderate confidence, 30+ for high)
- Win rate confidence interval: [range at 95%]
- Is this edge statistically significant? [YES / MARGINAL / NO]

### Verdict
[Is this pattern worth trading? Specific recommendation with data backing]

Rules

• 500 bars on 5m = ~2-3 weeks. For longer history, switch to higher timeframes (15m, 1H, Daily)
• Always state the sample size caveat — 5 occurrences is anecdote, not data
• If a pattern has <5 occurrences, say "Insufficient data — try a higher timeframe or broader conditions"
• Show the losing trades too — cherry-picking winners is the #1 backtesting lie
• MAE is as important as MFE — it tells the trader where to put the stop