OPEN-SOURCE SCRIPT
AetherEdge Multi-Feature Neural Divergence Hunter
🖊️ Overview
AetherEdge Multi-Feature Neural Divergence Hunter is a revolutionary divergence engine that simultaneously detects divergences across 6 independent oscillators and self-evolves neural weights based on each feature's empirical accuracy. Every feature (RSI/MACD/Stoch/CCI/MFI/MOM) detects both Regular and Hidden divergences in parallel, and Softmax-normalized dynamic weights evaluate confluence. With its 3D visual encoding (color × size × shape), the type and strength of every signal is identifiable at a glance — a next-generation divergence hunter.
🔶 Key Features
6 Parallel Oscillator Detection: RSI / MACD Hist / Stochastic / CCI / MFI / Momentum
Up to 12 Divergence Variants (2 types × 6 features) tracked simultaneously
Softmax Neural Weight Learning: Auto-redistribution based on empirical accuracy
Reward-Based Weight Updates: Bonus for correct, penalty for wrong
Confluence Counting: Separate Regular/Hidden tracking, combined for trigger
3D Visual Encoding:
Color: REG Bull/Bear, HID Bull/Bear, MIXED — 5-class system
Size: tiny→huge based on confluence count (5 levels)
Shape: Regular=triangle, Hidden=arrow, Mixed=label
Composite Neural Oscillator: Weighted average centered display
Confluence Glow: Background opacity reflects strength
Per-Feature Accuracy Tracking: Individual hit-rate records
Dynamic Divergence Lines: Width and style identify type
🧠 Technical Architecture
This indicator is a neural-weight-learning system processing 6 oscillators in parallel.
Feature Normalization:
RSI/Stoch/MFI: Native 0-100
MACD Hist/CCI/Momentum: Normalized to 0-100 over 100-bar range
Pivot Storage Method (Pine v5 compatible):
Eliminates dynamic history access by storing previous-pivot oscillator values
12 slots managed: prevPL_rsi, prevPL_macd, ...
Divergence Detection (per feature):
Regular Bull: Price LL + Oscillator HL
Hidden Bull: Price HL + Oscillator LL
Regular Bear: Price HH + Oscillator LH
Hidden Bear: Price LH + Oscillator HH
Distance filter: minBarsBet ≤ pivot gap ≤ maxBarsBet
Neural Weight Learning:
Evaluates return after rewardLook bars per signal
Correct (ret > 0): w += learnRate × |ret| × 100
Wrong (ret < 0): w -= learnRate × |ret| × 50 (asymmetric penalty)
Weight constraint: [0.2, 10.0]
Softmax Normalization (temperature τ):
w_i = exp((rawW_i - max) / τ) / Σexp(...)
Low τ → concentrate on best; High τ → uniform
Composite Neural Oscillator:
nnOsc = Σ(value_i × w_i) / Σw_i - 50
3-bar EMA smoothing, 5-zone state detection
3D Visual Encoding:
Color = type, Size = strength, Shape = category — 3-dimensional information transmission
⚙️ Recommended Settings & Tuning Guide
Crypto Defaults:
BTC (4H): pivotLen=5, minConfluence=2, learnRate=0.08 (standard)
ETH (1H): pivotLen=4, minConfluence=2, rewardLook=8
SOL (high-vol): pivotLen=6, minConfluence=3, rewardLook=12
XRP (short-term): pivotLen=3, minConfluence=2, rewardLook=5
Pivot Sensitivity:
pivotLen=3: high-sensitivity, noisy
pivotLen=5: standard
pivotLen=8–10: HTF, certainty-focused
Confluence Threshold:
minConfluence=1: many signals, false positives
minConfluence=2: standard (recommended)
minConfluence=3: curated, fewer
minConfluence=4+: ultra-curated, rare
Learning Parameters:
learnRate=0.05: conservative, stable
learnRate=0.08: standard
learnRate=0.15: fast, unstable risk
softmaxTemp=0.5: elite concentration (focus on best 1-2)
softmaxTemp=1.5: standard (balanced)
softmaxTemp=3.0: uniform (diversity)
Feature Selection Strategy:
Oscillator-focus: RSI + Stoch + MFI (momentum group)
Trend-focus: MACD + CCI + Momentum
Full parallel: All 6 ON (recommended, NN auto-curates)
💡 How to Use in Practice
5+ Confluence + REG MIXED: Strongest reversal signal (huge size emphasis)
3 REG Bull + NN 80%: High-probability reversal-buy candidate
3 HID Bull: Trend-continuation pullback buy
MIXED (REG+HID): Transition zone / unstable market
Neural Oscillator Usage:
EXTREME OB (>35): Overheated, watch for reversal
EXTREME OS (<-35): Bottom zone, watch for bounce
OS zero-cross + divergence: Direction confirmation
Learning Maturity Check:
Per-feature Acc% > 60%: feature is reliable
Acc% < 45%: feature is failing, ignore
Many Pending: still learning, wait
AetherEdge Synergy:
SMC AI Confidence Engine: A+ zone + 5× confluence = ultimate alignment
NeuraNet Predictor: Direction match + REG MIXED = double confirmation
RL Signal Optimizer: CONFIDENT + divergence = strong entry
All-in-One Dashboard: HIGH-CONVICTION + huge label = supreme reversal signal
⚠️ Important Notes
Pivot Lag: Divergence confirms pivotLen bars later (no real-time detection)
Weight Reset: Chart reload resets weight learning
Learning Maturity: Per-feature Acc% reliable only after Signals ≥ 20
Overfitting Risk: High learnRate + long runtime may over-skew weights
Hidden Divergence Nature: Continuation signal — avoid counter-trend misuse
MIXED Interpretation: Simultaneous REG+HID = transitional confusion, be cautious
Feature Disabling: Turning off unused features improves weight allocation
🚨 Disclaimer
This indicator is an advanced divergence-analysis tool for educational and research purposes only and does not constitute financial advice. Neural weight learning is a stochastic optimization method and does not guarantee future profits. Divergences suggest reversal or continuation possibilities but do not guarantee outcomes. Use with thorough validation and proper risk management.
AetherEdge Multi-Feature Neural Divergence Hunter is a revolutionary divergence engine that simultaneously detects divergences across 6 independent oscillators and self-evolves neural weights based on each feature's empirical accuracy. Every feature (RSI/MACD/Stoch/CCI/MFI/MOM) detects both Regular and Hidden divergences in parallel, and Softmax-normalized dynamic weights evaluate confluence. With its 3D visual encoding (color × size × shape), the type and strength of every signal is identifiable at a glance — a next-generation divergence hunter.
🔶 Key Features
6 Parallel Oscillator Detection: RSI / MACD Hist / Stochastic / CCI / MFI / Momentum
Up to 12 Divergence Variants (2 types × 6 features) tracked simultaneously
Softmax Neural Weight Learning: Auto-redistribution based on empirical accuracy
Reward-Based Weight Updates: Bonus for correct, penalty for wrong
Confluence Counting: Separate Regular/Hidden tracking, combined for trigger
3D Visual Encoding:
Color: REG Bull/Bear, HID Bull/Bear, MIXED — 5-class system
Size: tiny→huge based on confluence count (5 levels)
Shape: Regular=triangle, Hidden=arrow, Mixed=label
Composite Neural Oscillator: Weighted average centered display
Confluence Glow: Background opacity reflects strength
Per-Feature Accuracy Tracking: Individual hit-rate records
Dynamic Divergence Lines: Width and style identify type
🧠 Technical Architecture
This indicator is a neural-weight-learning system processing 6 oscillators in parallel.
Feature Normalization:
RSI/Stoch/MFI: Native 0-100
MACD Hist/CCI/Momentum: Normalized to 0-100 over 100-bar range
Pivot Storage Method (Pine v5 compatible):
Eliminates dynamic history access by storing previous-pivot oscillator values
12 slots managed: prevPL_rsi, prevPL_macd, ...
Divergence Detection (per feature):
Regular Bull: Price LL + Oscillator HL
Hidden Bull: Price HL + Oscillator LL
Regular Bear: Price HH + Oscillator LH
Hidden Bear: Price LH + Oscillator HH
Distance filter: minBarsBet ≤ pivot gap ≤ maxBarsBet
Neural Weight Learning:
Evaluates return after rewardLook bars per signal
Correct (ret > 0): w += learnRate × |ret| × 100
Wrong (ret < 0): w -= learnRate × |ret| × 50 (asymmetric penalty)
Weight constraint: [0.2, 10.0]
Softmax Normalization (temperature τ):
w_i = exp((rawW_i - max) / τ) / Σexp(...)
Low τ → concentrate on best; High τ → uniform
Composite Neural Oscillator:
nnOsc = Σ(value_i × w_i) / Σw_i - 50
3-bar EMA smoothing, 5-zone state detection
3D Visual Encoding:
Color = type, Size = strength, Shape = category — 3-dimensional information transmission
⚙️ Recommended Settings & Tuning Guide
Crypto Defaults:
BTC (4H): pivotLen=5, minConfluence=2, learnRate=0.08 (standard)
ETH (1H): pivotLen=4, minConfluence=2, rewardLook=8
SOL (high-vol): pivotLen=6, minConfluence=3, rewardLook=12
XRP (short-term): pivotLen=3, minConfluence=2, rewardLook=5
Pivot Sensitivity:
pivotLen=3: high-sensitivity, noisy
pivotLen=5: standard
pivotLen=8–10: HTF, certainty-focused
Confluence Threshold:
minConfluence=1: many signals, false positives
minConfluence=2: standard (recommended)
minConfluence=3: curated, fewer
minConfluence=4+: ultra-curated, rare
Learning Parameters:
learnRate=0.05: conservative, stable
learnRate=0.08: standard
learnRate=0.15: fast, unstable risk
softmaxTemp=0.5: elite concentration (focus on best 1-2)
softmaxTemp=1.5: standard (balanced)
softmaxTemp=3.0: uniform (diversity)
Feature Selection Strategy:
Oscillator-focus: RSI + Stoch + MFI (momentum group)
Trend-focus: MACD + CCI + Momentum
Full parallel: All 6 ON (recommended, NN auto-curates)
💡 How to Use in Practice
5+ Confluence + REG MIXED: Strongest reversal signal (huge size emphasis)
3 REG Bull + NN 80%: High-probability reversal-buy candidate
3 HID Bull: Trend-continuation pullback buy
MIXED (REG+HID): Transition zone / unstable market
Neural Oscillator Usage:
EXTREME OB (>35): Overheated, watch for reversal
EXTREME OS (<-35): Bottom zone, watch for bounce
OS zero-cross + divergence: Direction confirmation
Learning Maturity Check:
Per-feature Acc% > 60%: feature is reliable
Acc% < 45%: feature is failing, ignore
Many Pending: still learning, wait
AetherEdge Synergy:
SMC AI Confidence Engine: A+ zone + 5× confluence = ultimate alignment
NeuraNet Predictor: Direction match + REG MIXED = double confirmation
RL Signal Optimizer: CONFIDENT + divergence = strong entry
All-in-One Dashboard: HIGH-CONVICTION + huge label = supreme reversal signal
⚠️ Important Notes
Pivot Lag: Divergence confirms pivotLen bars later (no real-time detection)
Weight Reset: Chart reload resets weight learning
Learning Maturity: Per-feature Acc% reliable only after Signals ≥ 20
Overfitting Risk: High learnRate + long runtime may over-skew weights
Hidden Divergence Nature: Continuation signal — avoid counter-trend misuse
MIXED Interpretation: Simultaneous REG+HID = transitional confusion, be cautious
Feature Disabling: Turning off unused features improves weight allocation
🚨 Disclaimer
This indicator is an advanced divergence-analysis tool for educational and research purposes only and does not constitute financial advice. Neural weight learning is a stochastic optimization method and does not guarantee future profits. Divergences suggest reversal or continuation possibilities but do not guarantee outcomes. Use with thorough validation and proper risk management.
オープンソーススクリプト
TradingViewの精神に則り、このスクリプトの作者はコードをオープンソースとして公開してくれました。トレーダーが内容を確認・検証できるようにという配慮です。作者に拍手を送りましょう！無料で利用できますが、コードの再公開はハウスルールに従う必要があります。
免責事項
これらの情報および投稿は、TradingViewが提供または承認する金融、投資、取引、またはその他の種類の助言もしくは推奨であることを意図したものではなく、またこれらに該当するものでもありません。詳細は利用規約をご覧ください。
オープンソーススクリプト
TradingViewの精神に則り、このスクリプトの作者はコードをオープンソースとして公開してくれました。トレーダーが内容を確認・検証できるようにという配慮です。作者に拍手を送りましょう！無料で利用できますが、コードの再公開はハウスルールに従う必要があります。
免責事項
これらの情報および投稿は、TradingViewが提供または承認する金融、投資、取引、またはその他の種類の助言もしくは推奨であることを意図したものではなく、またこれらに該当するものでもありません。詳細は利用規約をご覧ください。