OPEN-SOURCE SCRIPT
AetherEdge - Adaptive Volume Profile
🖊️ Overview
AE-AVP pairs an auction-theory volume profile (POC and value area) with an engine that learns whether a value-area-edge break gets accepted (a real breakout) or rejected (faded back into value). It shows where volume concentrated and puts a probability on the most tradeable moment — the fight at the value-area edge.
🔶 Key Features
A real volume profile — POC, value area (VAH/VAL), and node strength drawn as a right-side histogram
ML: acceptance/rejection — online-learns whether an edge break continues or reverts; low probability flags a fade to the POC
RL self-tuning threshold — a UCB bandit tunes the acceptance threshold per volatility regime
Signals — distinguishes accepted breakouts (follow) from edge rejections (fade to POC)
POC/VAH/VAL are usable horizontal levels
Learning/signals on bar close — no repaint
🧠 Technical Architecture
Volume profile: the recent lookback bars' price range is split into rows bins; each bar's volume is distributed across the bins its [low, high] spans. POC = highest-volume bin; value area = expand from the POC toward the heavier neighbour until vaPct% (default 70%) of volume is enclosed → VAH/VAL; node strength = bin volume / max.
ML (online logistic regression): when price breaks above VAH or below VAL, seven features (edge node strength, approach momentum, volume surge, stretch from POC, trend, volatility regime, VA width — oriented to the break direction) feed a model of P(acceptance). Events are held in an internal array and labeled by a triple barrier (accept = beyond the edge by k·ATR; reject = back inside; time-out discarded) — no future data.
RL (UCB contextual bandit): tunes the acceptance threshold per volatility regime (ATR-percentile terciles); the reward is tied to the same edge resolution, so ML and RL share one judgment loop.
Signals: acceptance (P ≥ threshold) → follow the break direction; rejection (P ≤ 1 − threshold) → fade toward the POC.
Honest scope: a linear classifier + a UCB bandit over standard profile/price features. Not deep learning, not a guarantee.
⚙️ Recommended Settings & Tuning Guide (crypto 15m–4H)
Key parameters: profile lookback, bins (rows), value-area %, acceptance/rejection distance (ATR), horizon M, threshold search range.
Larger lookback → big-picture POC (stable, slower); smaller → sensitive to recent volume
More bins → higher resolution (finer, heavier to draw)
Higher value-area % → a wider VA and stricter break conditions
Guide: BTC/ETH (15m–1H) defaults (lookback 150, rows 24, VA 70%); 4H/daily a longer lookback; high-vol alts slightly wider acceptance/rejection distances
P(acceptance), threshold, and accuracy are coarse until warmup plus enough edge events accumulate
💡 How to Use in Practice
The POC is a magnet price; VAH/VAL are the battle lines — range inside value, direction on an edge break
Acceptance signals (BRK) mean follow the break; rejection signals (REJ) mean fade the edge toward the POC
Use P(acceptance) and node strength to judge whether a break is real (breaking a weak node with high acceptance)
Leave the auto threshold to the learner by default
Combine with a higher-timeframe value area or a trend tool for added precision
⚠️ Important Notes
Needs a learning period (warmup); weights and the Q-table re-learn on input/symbol/timeframe change
The profile is computed from closed bars; the histogram shows the current window
An edge break does not guarantee acceptance (which is exactly why it's shown as a probability)
Probability, not a guarantee — always use stops and position sizing
🚨 Disclaimer
This indicator is for educational and informational purposes only and is not financial advice or a recommendation to buy or sell. No method guarantees future profits; past performance does not indicate future results, and trading carries the risk of loss. All trading decisions are your own — use proper backtesting and disciplined risk management.
AE-AVP pairs an auction-theory volume profile (POC and value area) with an engine that learns whether a value-area-edge break gets accepted (a real breakout) or rejected (faded back into value). It shows where volume concentrated and puts a probability on the most tradeable moment — the fight at the value-area edge.
🔶 Key Features
A real volume profile — POC, value area (VAH/VAL), and node strength drawn as a right-side histogram
ML: acceptance/rejection — online-learns whether an edge break continues or reverts; low probability flags a fade to the POC
RL self-tuning threshold — a UCB bandit tunes the acceptance threshold per volatility regime
Signals — distinguishes accepted breakouts (follow) from edge rejections (fade to POC)
POC/VAH/VAL are usable horizontal levels
Learning/signals on bar close — no repaint
🧠 Technical Architecture
Volume profile: the recent lookback bars' price range is split into rows bins; each bar's volume is distributed across the bins its [low, high] spans. POC = highest-volume bin; value area = expand from the POC toward the heavier neighbour until vaPct% (default 70%) of volume is enclosed → VAH/VAL; node strength = bin volume / max.
ML (online logistic regression): when price breaks above VAH or below VAL, seven features (edge node strength, approach momentum, volume surge, stretch from POC, trend, volatility regime, VA width — oriented to the break direction) feed a model of P(acceptance). Events are held in an internal array and labeled by a triple barrier (accept = beyond the edge by k·ATR; reject = back inside; time-out discarded) — no future data.
RL (UCB contextual bandit): tunes the acceptance threshold per volatility regime (ATR-percentile terciles); the reward is tied to the same edge resolution, so ML and RL share one judgment loop.
Signals: acceptance (P ≥ threshold) → follow the break direction; rejection (P ≤ 1 − threshold) → fade toward the POC.
Honest scope: a linear classifier + a UCB bandit over standard profile/price features. Not deep learning, not a guarantee.
⚙️ Recommended Settings & Tuning Guide (crypto 15m–4H)
Key parameters: profile lookback, bins (rows), value-area %, acceptance/rejection distance (ATR), horizon M, threshold search range.
Larger lookback → big-picture POC (stable, slower); smaller → sensitive to recent volume
More bins → higher resolution (finer, heavier to draw)
Higher value-area % → a wider VA and stricter break conditions
Guide: BTC/ETH (15m–1H) defaults (lookback 150, rows 24, VA 70%); 4H/daily a longer lookback; high-vol alts slightly wider acceptance/rejection distances
P(acceptance), threshold, and accuracy are coarse until warmup plus enough edge events accumulate
💡 How to Use in Practice
The POC is a magnet price; VAH/VAL are the battle lines — range inside value, direction on an edge break
Acceptance signals (BRK) mean follow the break; rejection signals (REJ) mean fade the edge toward the POC
Use P(acceptance) and node strength to judge whether a break is real (breaking a weak node with high acceptance)
Leave the auto threshold to the learner by default
Combine with a higher-timeframe value area or a trend tool for added precision
⚠️ Important Notes
Needs a learning period (warmup); weights and the Q-table re-learn on input/symbol/timeframe change
The profile is computed from closed bars; the histogram shows the current window
An edge break does not guarantee acceptance (which is exactly why it's shown as a probability)
Probability, not a guarantee — always use stops and position sizing
🚨 Disclaimer
This indicator is for educational and informational purposes only and is not financial advice or a recommendation to buy or sell. No method guarantees future profits; past performance does not indicate future results, and trading carries the risk of loss. All trading decisions are your own — use proper backtesting and disciplined risk management.
Open-source script
In true TradingView spirit, the creator of this script has made it open-source, so that traders can review and verify its functionality. Kudos to the author! While you can use it for free, remember that republishing the code is subject to our House Rules.
Disclaimer
The information and publications are not meant to be, and do not constitute, financial, investment, trading, or other types of advice or recommendations supplied or endorsed by TradingView. Read more in the Terms of Use.
Open-source script
In true TradingView spirit, the creator of this script has made it open-source, so that traders can review and verify its functionality. Kudos to the author! While you can use it for free, remember that republishing the code is subject to our House Rules.
Disclaimer
The information and publications are not meant to be, and do not constitute, financial, investment, trading, or other types of advice or recommendations supplied or endorsed by TradingView. Read more in the Terms of Use.