OPEN-SOURCE SCRIPT
AetherEdge - Adaptive Volume Surge Detector
🖊️ Overview
AE-AVSD detects volume surges and uses a learning classifier to judge whether each surge is REAL or NOISE, printing a large BUY/SELL label only when a real surge aligns with a decisive candle direction. Unlike classical volume alerts, the AI filter structurally suppresses false signals — the chart never clutters with arrows.
🔶 Key Features
Auto-detects volume surges via statistical z-score significance
Candle-direction confirmation via body/range ratio — only directional spikes count
[AI Layer] Real/noise classification via online logistic regression learning follow-through
Fires only when three conditions align: volume surprise, direction, AI confidence
Big arrows on real surges only — everything else stays silent
Bar coloring visually emphasizes confirmed real surges
Integrated HUD showing verdict, features, AI probability, model accuracy
Per-direction alerts (BUY/SELL)
🧠 Technical Architecture
Surge detection uses the standard z-score method: bars where (volume − volMean) / volStd exceeds the threshold become candidates. The AI classification layer is a genuine online logistic regression: five features (bias, volume z, body magnitude in ATRs, body/range ratio, log return) feed a sigmoid that outputs P(real). Learning is delayed-label: surge contexts are buffered, and after the horizon a label is generated from whether an ATR-scaled continuation actually occurred, with weights updated by stochastic gradient descent. Signals fire only when all three — volume z, candle direction, and AI probability — pass their thresholds simultaneously, structurally excluding noise.
⚙️ Recommended Settings & Tuning Guide
Tuned with crypto in mind. The scalping-focused defaults (Surge Threshold 2σ, Follow-Through Horizon 5 bars, Required Follow-Through 0.7 ATR) are a starting point. If too many signals, raise Surge Threshold to 2.5–3.0σ and Real-Probability Threshold to 0.65–0.7. For day trading, raise Follow-Through Horizon to 10–15 bars and Required Follow-Through to 1.0–1.5 ATR, classifying only larger moves as "real." Min Body/Range Ratio excludes doji-like wick spikes; 0.5 is the standard default.
💡 How to Use in Practice
A BUY/SELL label appearing is a strong signal — volume confirmation + AI confidence all aligned. Scalpers can ride the next candle's direction; day traders can wait for the first pullback before entering. Pair with support/resistance and trend tools, and favor signals coinciding with key-level breaks or rejections. Consistently high P(real) on an instrument means volume signals work well there; consistently low values point to noisy instruments where other tools should be preferred.
⚠️ Important Notes
The AI layer requires accumulated data, so on new instruments or short windows P(real) stays near the initial 0.5 and the filter is weak. Static or near-50% model accuracy in the HUD indicates insufficient learning samples — meaningful classification needs roughly 30–50 surges. Volume signals tend to be more accurate in strong trends than in ranges, where fakeouts increase. Designed for scalping, intended for lower timeframes.
🚨 Disclaimer
This indicator is provided for educational and informational purposes only and does not constitute financial or investment advice. No signal guarantees future profit, and past performance does not indicate future results. All trading decisions are made at your own risk.
AE-AVSD detects volume surges and uses a learning classifier to judge whether each surge is REAL or NOISE, printing a large BUY/SELL label only when a real surge aligns with a decisive candle direction. Unlike classical volume alerts, the AI filter structurally suppresses false signals — the chart never clutters with arrows.
🔶 Key Features
Auto-detects volume surges via statistical z-score significance
Candle-direction confirmation via body/range ratio — only directional spikes count
[AI Layer] Real/noise classification via online logistic regression learning follow-through
Fires only when three conditions align: volume surprise, direction, AI confidence
Big arrows on real surges only — everything else stays silent
Bar coloring visually emphasizes confirmed real surges
Integrated HUD showing verdict, features, AI probability, model accuracy
Per-direction alerts (BUY/SELL)
🧠 Technical Architecture
Surge detection uses the standard z-score method: bars where (volume − volMean) / volStd exceeds the threshold become candidates. The AI classification layer is a genuine online logistic regression: five features (bias, volume z, body magnitude in ATRs, body/range ratio, log return) feed a sigmoid that outputs P(real). Learning is delayed-label: surge contexts are buffered, and after the horizon a label is generated from whether an ATR-scaled continuation actually occurred, with weights updated by stochastic gradient descent. Signals fire only when all three — volume z, candle direction, and AI probability — pass their thresholds simultaneously, structurally excluding noise.
⚙️ Recommended Settings & Tuning Guide
Tuned with crypto in mind. The scalping-focused defaults (Surge Threshold 2σ, Follow-Through Horizon 5 bars, Required Follow-Through 0.7 ATR) are a starting point. If too many signals, raise Surge Threshold to 2.5–3.0σ and Real-Probability Threshold to 0.65–0.7. For day trading, raise Follow-Through Horizon to 10–15 bars and Required Follow-Through to 1.0–1.5 ATR, classifying only larger moves as "real." Min Body/Range Ratio excludes doji-like wick spikes; 0.5 is the standard default.
💡 How to Use in Practice
A BUY/SELL label appearing is a strong signal — volume confirmation + AI confidence all aligned. Scalpers can ride the next candle's direction; day traders can wait for the first pullback before entering. Pair with support/resistance and trend tools, and favor signals coinciding with key-level breaks or rejections. Consistently high P(real) on an instrument means volume signals work well there; consistently low values point to noisy instruments where other tools should be preferred.
⚠️ Important Notes
The AI layer requires accumulated data, so on new instruments or short windows P(real) stays near the initial 0.5 and the filter is weak. Static or near-50% model accuracy in the HUD indicates insufficient learning samples — meaningful classification needs roughly 30–50 surges. Volume signals tend to be more accurate in strong trends than in ranges, where fakeouts increase. Designed for scalping, intended for lower timeframes.
🚨 Disclaimer
This indicator is provided for educational and informational purposes only and does not constitute financial or investment advice. No signal guarantees future profit, and past performance does not indicate future results. All trading decisions are made at your own risk.
开源脚本
秉承TradingView的精神，该脚本的作者将其开源，以便交易者可以查看和验证其功能。向作者致敬！您可以免费使用该脚本，但请记住，重新发布代码须遵守我们的网站规则。
免责声明
这些信息和出版物并非旨在提供，也不构成TradingView提供或认可的任何形式的财务、投资、交易或其他类型的建议或推荐。请阅读使用条款了解更多信息。
开源脚本
秉承TradingView的精神，该脚本的作者将其开源，以便交易者可以查看和验证其功能。向作者致敬！您可以免费使用该脚本，但请记住，重新发布代码须遵守我们的网站规则。
免责声明
这些信息和出版物并非旨在提供，也不构成TradingView提供或认可的任何形式的财务、投资、交易或其他类型的建议或推荐。请阅读使用条款了解更多信息。