Adaptive RSI Pro

中文说明 / Chinese README

Pine Script v6 | v7.3

Dynamic thresholds + multi-timeframe analysis + divergence detection + stats filtering + smart alerts

---

Core Concept

Traditional RSI uses fixed 30/70 thresholds, but different assets have different volatility regimes. This indicator uses Z-Score statistics to calculate adaptive thresholds based on each asset's historical distribution.

Z-Score	Percentile	Meaning
±2σ	P2 / P98	Extreme zone
±Nσ	Dynamic	Dynamic normal overbought/oversold reference

---

Signal Legend

Buy Signals (shown near the bottom)

Icon	Name	Condition	Priority
🌟	MTF Resonance	Multi-timeframe oversold alignment + Z < -2σ	★★★★★
💎	Divergence + Extreme	Bullish divergence inside the extreme oversold zone	★★★★☆
🔥	Extreme Oversold	Z-Score breaks below -2σ (about P2)	★★★☆☆
⬆️	Normal Oversold	Z-Score breaks below -Nσ (normal_threshold)	★★☆☆☆
↗️	Bullish Divergence	Price makes a new low while RSI does not	★☆☆☆☆

Sell Signals (shown near the top)

Icon	Name	Condition	Priority
🌟	MTF Resonance	Multi-timeframe overbought alignment + Z > +2σ	★★★★★
💎	Divergence + Extreme	Bearish divergence inside the extreme overbought zone	★★★★☆
❄️	Extreme Overbought	Z-Score breaks above +2σ (about P98)	★★★☆☆
⬇️	Normal Overbought	Z-Score breaks above +Nσ (normal_threshold)	★★☆☆☆
↘️	Bearish Divergence	Price makes a new high while RSI does not	★☆☆☆☆

Priority system: When multiple conditions are true on the same bar, only the highest-priority signal is shown.

Status Indicators

Icon	Status	Z-Score Range
🟢	Extreme oversold	Z < -2σ
🟡	Oversold	-2σ ≤ Z < -Nσ*
⚪	Neutral	-Nσ ≤ Z ≤ +Nσ
🟠	Overbought	+Nσ < Z ≤ +2σ
🔴	Extreme overbought	Z > +2σ

*N = the dynamic normal threshold (normal_threshold). The script calculates it from volatility, so it can be near 1.0σ in high-volatility markets and around 1.8σ in lower-volatility markets.

Quality Grades

Each signal carries a quality grade based on multi-factor scoring:

Grade	Score	Interpretation
[A]	≥80	High quality, tradable
[B]	60-79	Good, tradable
[C]	40-59	Mixed, trade cautiously
[D]	<40	Low quality, usually skip

Scoring factors: MTF resonance (+25) | divergence (+25) | RSI pivot confirmation (+10) | weekly trend alignment (+15) | volume surge (+10) | deeper extremes (+10/+20) | ADX counter-trend penalty (-10)

Display Marks

Mark	Meaning	Notes
✓	Passed the stats filter	Can appear in dashboard signal rows and alert messages
⚠️	Failed the stats filter but still shown	Common in Alert Only or Soft mode
🚫	Signal exists but is hidden	Can be caused by Smart normal-signal hiding, trend protection, or Hard filtering
None	Not triggered or stats filter disabled	For example persistent state text such as 🔥持续 or 超卖区

Note: Alerts fire only for signals that pass the stats filter. In practice, alert text usually shows ✓ or no mark at all, and does not send ⚠️.

---

Key Features

1. Adaptive Thresholds

• Automatically calculates the lookback window from the statistical formula n = (Z × σ / E)²
• Adjusts lookback_min and lookback_max from asset volatility
• Three precision presets: High / Normal / Low
• Includes health checks for sample coverage, distribution width, and statistical validity

2. Threshold Lines and Visual Modes

• Four line modes: Unified, Z-Score, Percentile, and Both
• Optional gradient fills and custom bull/bear colors
• Dashboard supports both Mobile and Full layouts with four size presets

3. Multi-Timeframe (MTF) Analysis

• Automatic fractal timeframe selection or manual selection of three timeframes
• Weighted resonance detection, with the highest timeframe counting double
• 🌟 resonance signals are the top-priority signals

4. Divergence Detection

• Volatility-adaptive behavior for Low Vol / Normal / High Vol / Crypto
• Single-anchor divergence detection, plotted at the structure pivot
• Extreme-zone divergence 💎 versus regular divergence ↗️ / ↘️
• Alerts can append ⚡实时背离 when realtime divergence is present

5. Trend Protection

• Weekly trend filter to reduce counter-trend trades
• Three protection levels: Aggressive / Moderate / Loose
• Percentile Confirm can require extreme signals to satisfy both Z-Score and P5/P95 conditions
• Smart normal-signal mode automatically hides normal signals during weekly extreme conditions

6. Cooldown Upgrade Awareness

• High-priority signals (🌟 / 💎 / 🔥 / ❄️) use a 1-bar cooldown
• Normal signals ⬆️ / ⬇️ use fixed or adaptive cooldown bars
• Same-side signal upgrades can bypass cooldown, such as ⬆️ -> 🔥 -> 🌟

7. Signal Statistics and Filtering

• Stats modes: Signal Type, Grade, and Ranking
• Uses Bayesian adjustment to reduce small-sample bias before filtering by sample count and adjusted win rate
• Three filter modes:
  • Alert Only: chart signals stay visible, alerts are filtered
  • Soft: failed signals are downgraded visually
  • Hard: failed signals are hidden

8. Smart Alert

• A single alert aggregates all signal types
• Includes RSI, Z-Score, approximate percentile, and quality grade
• Alert icons match the script output: 🌟MTF共振 / 💎背离 / 🔥极端 / ❄️极端 / ⬆️超卖 / ⬇️超买
• Can append ✓确认, ↩反转, and ⚡实时背离 when conditions apply
• Optional ATR-based risk hints for stop-loss and take-profit suggestions

9. Strategy Harness

• Separate file: adaptive_rsi_strategy_harness.pine
• Trade Side: Long Only / Short Only / Both
• Backtest Mode
  • Baseline: raw production signals
  • Production: only the signals that would make it into production alerts
• Full guide: docs/STRATEGY_REPORT.md

---

Dashboard

Full Mode (Desktop)

┌────────────────────────────────────────┐
│ ADAPTIVE RSI                    35.2   │
├────────────────────────────────────────┤
│ Z-Score     -2.15σ (≈P2)               │
│ Percentile  P5 (−1.5σ ~ −2σ)           │
│ Signal      🔥[A]✓                     │
│ Status      🟢 EXTREME OVERSOLD        │
│ Protection  ✓ W.RSI:45 📊↑             │
│ Lookback    456(150-800) ✅✅✅         │
│ Normal      ⬆️1.50σ ✓                  │
├────────────────────────────────────────┤
│ MTF 1m|5m|15m  🟢|⚪|🟢                │
│ Resonance    🟢 3/4                    │
│ Divergence Auto  🟢 BULL (5/60)        │
├────────────────────────────────────────┤
│ Ranking      (20b)                     │
│ 🌟[A]📈(12)✓  +4.5%|85%                │
│ 💎[A]📈(8)✓   +4.2%|82%                │
└────────────────────────────────────────┘

Mobile Mode

┌─────────────────┐
│  RSI      35.2  │
│  Signal   🔥[A]✓│  Signal + grade + mark
│  Status   🟢EXT │  Pure status
└─────────────────┘

---

Quick Start

1. Add the Indicator

1. Open TradingView and go to Pine Editor.
2. Paste the contents of adaptive_rsi.pine.
3. Click Add to chart.

2. Configure Alerts

1. Right-click the indicator and choose Add Alert.
2. Set the condition to Any alert() function call.
3. If you want ATR hints in alerts, enable Include Risk Hints in Alerts.

3. Suggested Presets

Scenario	Dashboard	Normal Signals	Protection Level	Filter Mode
Intraday trading	Full	Smart	Moderate	Alert Only
Swing trading	Full	Off	Moderate	Hard
Mobile monitoring	Mobile	Off	Loose	Alert Only

4. Filter Mode Tips

• Alert Only: best default for most users; keep all chart context while filtering alerts
• Soft: useful when you want context but lower-quality signals de-emphasized
• Hard: use when you only want historically qualified signals and a cleaner chart

5. Strategy Validation

1. Open a separate Pine script.
2. Paste adaptive_rsi_strategy_harness.pine.
3. Use TradingView Strategy Tester with Trade Side and Backtest Mode.

---

Alert Examples

AAPL: 🟢 BUY → 🌟MTF共振 | RSI:25.3 Z:-2.1σ (≈P2) [A]✓
AAPL: 🔴 SELL → ❄️极端 | RSI:78.5 Z:2.3σ (≈P98) [B]✓

With risk hints enabled:

AAPL: 🟢 BUY → 🔥极端 ✓确认 ⚡实时背离 | RSI:25.3 Z:-2.1σ (≈P2) [A]✓ | SL:-1.5% TP:+3.0%

Note: Alerts do not send ⚠️ signals that fail the stats filter. If stats filtering is disabled, the filter mark at the end of the alert also disappears.

---

Current Version

v7.3

• Keeps the public v7.2 signal model as the main baseline
• Retains only minimal correctness fixes:
  • lookback floor stays above the statistical lower bound
  • weekly protection uses confirmed weekly data
  • lower-timeframe MTF uses proper lower-TF aggregation
• Adds adaptive_rsi_strategy_harness.pine for Strategy Tester validation

v7.2

• Tiered cooldown with upgrade exemption
• High-priority signals (🌟 / 💎 / 🔥 / ❄️) use a 1-bar cooldown
• Same-side upgrades can bypass cooldown, such as Normal -> Extreme -> MTF
• Faster response when signal quality deteriorates quickly
• Includes percentile confirm, stats filtering, ranking, and ATR risk hints

---

Code Quality

This project uses a custom Pine Script Static Analyzer for local and CI checks.

GitHub CI

GitHub Actions runs lint checks automatically when changes touch .pine, .pine-lint.yml, or tools/pine_linter/**.

Local Check

python3 tools/pine_linter/cli.py --config .pine-lint.yml adaptive_rsi.pine
python3 tools/pine_linter/cli.py --config .pine-lint.yml --format markdown --output lint-report.md adaptive_rsi.pine

Lint Rules

Rule	Severity	Description
SEC001	error	request.security() must use lookahead=barmerge.lookahead_off
SEC002	warning	request.security() inside conditional branches may repaint
SYN001	warning	Multi-line ternary expressions are fragile in Pine v6
SYN002	info	switch statements should include a default branch
SYN003	warning	table.clear() must receive an explicit range
NAM001-003	info	Naming convention checks for constants, functions, and types
QUA001	info	Tooltips should include bilingual text
QUA002	warning	request.security() outputs should be checked for na

Config file: .pine-lint.yml