kuntal-r-d Price Action
Prompt-first, pure-prompt skill. No script — read the raw OHLC bars directly. Patterns only matter in context (location vs structure), so always weigh where the pattern occurs.
Role & objective
Identify candlestick patterns and market structure and rate the resulting setup quality (−1 bearish .. +1 bullish), confidence and rating — purely from price, no indicators.
When to use
"Any candlestick pattern on GP?", "price action read", "market structure / swing points", "is that a pin bar / engulfing?", "no-indicator chart read".
Inputs you need
ohlcv — daily bars (need ≥3 for patterns; more for structure). Optional volume and
support/resistance levels for context.
Method (follow in order)
- Candlestick patterns — reversal (hammer, shooting star, engulfing, morning/evening star), continuation (three soldiers/crows), indecision (doji, spinning top, harami).
- Market structure — swing highs/lows; phase (trending vs ranging); last swing high/low; is structure intact or broken?
- Bar types — inside/outside bars, pin/rejection bars, momentum vs exhaustion bars.
- Context (weigh ~60%) — pattern location vs support/resistance and trend; a hammer at support in an uptrend is strong; the same mid-range is weak.
Scoring rubric
Base each pattern's signal by reliability, then multiply by context (location + trend alignment). score = clamp(context-weighted pattern signal, −1, 1). Rating: ≥0.5 strong_bullish · ≥0.15 bullish · −0.15..0.15 neutral · ≤−0.15 bearish · ≤−0.5 strong_bearish. Confidence rises with multi-bar confirmation and supportive context.
Output (emit this Thinking Card)
{ "skill": "price-action", "ticker": "..", "mode": "momentum", "as_of": "..",
"score": 0.0, "confidence": 0.0, "rating": "..",
"key_metrics": { "patterns": [{"pattern": "Bullish Engulfing", "location": "support",
"reliability": 0.0}], "phase": "uptrend|range", "last_swing_high": 0, "last_swing_low": 0 },
"reasoning": ["..."], "flags": ["insufficient_bars?"],
"disclaimer": "Educational analysis only. Not financial advice." }
DSE pitfalls
- Circuit-limited days produce distorted candles (truncated wicks/bodies) — don't read normal patterns into limit-up/limit-down bars; flag them.
- A pattern with no location/trend context is noise — context dominates the score.
- Thin trading creates erratic single bars — require confirmation before calling a reversal.
Optional precision helper
No bundled script — pure-prompt skill. The technical-analysis script scores the latest
candlestick (hammer/star/engulfing) numerically if you want a cross-check.
Worked example
Bullish engulfing at a tested support in an uptrend, confirmed next bar → reliability high × favourable context → score ≈ +0.6 → strong_bullish, confidence ≈ 0.7.
References
See technical-analysis/references/INDICATORS.md (candlestick section).
Output is educational analysis only, never financial advice.