OPEN-SOURCE SCRIPT
ForgeQuant AI | Regime + LSTM + R:R [MarketFragments]
ForgeQuant AI [MarketFragments]
A regime-aware overlay that combines four systems into a single read on the
chart: a regime classifier, a deep recurrent LSTM-style memory cell, a 0-100
confluence edge score, and a dynamic ATR-based risk/reward engine.
Important up front: this is a research preview. It has NOT been forward tested,
and no performance claims are made. See the notes at the bottom.
─────────────────────────────────────────────────────────────
HOW IT WORKS
─────────────────────────────────────────────────────────────
STEP 1 -- REGIME DETECTION
Volatility percentile (ATR rank) and a directional-movement trend-strength
reading classify each bar into one of six regimes: Strong Trend Bull, Strong
Trend Bear, High-Vol Chop, Mean Reversion, Breakout Potential, or Neutral. The
chart background shades to match.
STEP 2 -- DEEP LSTM MEMORY CELL
A 2-layer recurrent memory cell processes an engineered input vector built from
normalized volume, momentum vs the 21 EMA, volatility-of-volatility, and RSI
deviation. Layer 1 runs a fast cell and a slow cell; Layer 2 is a stacked
context cell. Gate weights adapt online via a clipped, momentum-smoothed
pseudo-BPTT update, with optional attention-weighted readout and layer-norm
input scaling. The cell outputs a hidden state, a confidence reading, a regime
bias (via an adaptive bias band), and an anomaly flag when price behavior
diverges sharply from the learned state.
This is a heuristic recurrent filter written in Pine -- not a trained neural
network. Treat it as an adaptive smoother, not a forecast.
STEP 3 -- CONFLUENCE EDGE SCORE
A 0-100 score blends five inputs: market structure (fair-value gaps and volume
absorption), volume behavior, momentum (RSI and the 9 EMA), higher-timeframe
trend (60-minute), and the LSTM bias/anomaly state.
STEP 4 -- DYNAMIC RISK & REWARD
Suggested risk % scales by regime (reduced in chop, trimmed further on an
anomaly). Stop and target are ATR multiples, with the target stretched by LSTM
confidence. A forward-projected R:R cloud draws the target (green) and stop
(red) zones from the latest close.
─────────────────────────────────────────────────────────────
SIGNALS
─────────────────────────────────────────────────────────────
Long Edge score >= 62, RSI > 52, a bullish FVG or absorption bar, and no
active LSTM anomaly.
Short Edge score >= 62, RSI < 48, a bearish FVG or absorption bar, and no
active LSTM anomaly.
Triggers evaluate on confirmed bars only.
─────────────────────────────────────────────────────────────
WHAT YOU SEE
─────────────────────────────────────────────────────────────
Background shade Current regime
Cyan line LSTM hidden state plotted around price (× ATR amplitude)
Green/Red cloud Dynamic R:R -- target (green) and stop (red)
Triangles Long (L, up) and Short (S, down) signals
Dashboard panel Regime, edge score, LSTM confidence, suggested risk %,
dynamic R:R, anomaly status, attention split, bias band
Gate plots Optional forget/input/output gate activations (debug)
─────────────────────────────────────────────────────────────
SETTINGS
─────────────────────────────────────────────────────────────
Display Regime background, panel position/size, diagnostics, gates
LSTM Memory Cell Enable, memory speed, anomaly sensitivity, attention,
layer-norm, hidden-state plot amplitude
Regime-Bias Band Auto (stdev-scaled) vs manual, sensitivity, lookback,
floor/ceiling
Dynamic Risk Base risk %, ATR stop/target multiples, chop multiplier,
R:R cloud toggle and forward extend
─────────────────────────────────────────────────────────────
IMPORTANT NOTES
─────────────────────────────────────────────────────────────
-- This indicator has NOT been backtested or forward tested
-- No performance claims are made
-- Shared as a research tool for community review
-- Risk, stop, and target values are illustrative sizing suggestions, not
trade instructions
-- Results will vary by instrument, timeframe, and market conditions
-- This is not financial advice
-- Trading involves substantial risk of loss
-- Past results do not guarantee future performance
-- Use for educational and research purposes only
─────────────────────────────────────────────────────────────
Free for public use
A regime-aware overlay that combines four systems into a single read on the
chart: a regime classifier, a deep recurrent LSTM-style memory cell, a 0-100
confluence edge score, and a dynamic ATR-based risk/reward engine.
Important up front: this is a research preview. It has NOT been forward tested,
and no performance claims are made. See the notes at the bottom.
─────────────────────────────────────────────────────────────
HOW IT WORKS
─────────────────────────────────────────────────────────────
STEP 1 -- REGIME DETECTION
Volatility percentile (ATR rank) and a directional-movement trend-strength
reading classify each bar into one of six regimes: Strong Trend Bull, Strong
Trend Bear, High-Vol Chop, Mean Reversion, Breakout Potential, or Neutral. The
chart background shades to match.
STEP 2 -- DEEP LSTM MEMORY CELL
A 2-layer recurrent memory cell processes an engineered input vector built from
normalized volume, momentum vs the 21 EMA, volatility-of-volatility, and RSI
deviation. Layer 1 runs a fast cell and a slow cell; Layer 2 is a stacked
context cell. Gate weights adapt online via a clipped, momentum-smoothed
pseudo-BPTT update, with optional attention-weighted readout and layer-norm
input scaling. The cell outputs a hidden state, a confidence reading, a regime
bias (via an adaptive bias band), and an anomaly flag when price behavior
diverges sharply from the learned state.
This is a heuristic recurrent filter written in Pine -- not a trained neural
network. Treat it as an adaptive smoother, not a forecast.
STEP 3 -- CONFLUENCE EDGE SCORE
A 0-100 score blends five inputs: market structure (fair-value gaps and volume
absorption), volume behavior, momentum (RSI and the 9 EMA), higher-timeframe
trend (60-minute), and the LSTM bias/anomaly state.
STEP 4 -- DYNAMIC RISK & REWARD
Suggested risk % scales by regime (reduced in chop, trimmed further on an
anomaly). Stop and target are ATR multiples, with the target stretched by LSTM
confidence. A forward-projected R:R cloud draws the target (green) and stop
(red) zones from the latest close.
─────────────────────────────────────────────────────────────
SIGNALS
─────────────────────────────────────────────────────────────
Long Edge score >= 62, RSI > 52, a bullish FVG or absorption bar, and no
active LSTM anomaly.
Short Edge score >= 62, RSI < 48, a bearish FVG or absorption bar, and no
active LSTM anomaly.
Triggers evaluate on confirmed bars only.
─────────────────────────────────────────────────────────────
WHAT YOU SEE
─────────────────────────────────────────────────────────────
Background shade Current regime
Cyan line LSTM hidden state plotted around price (× ATR amplitude)
Green/Red cloud Dynamic R:R -- target (green) and stop (red)
Triangles Long (L, up) and Short (S, down) signals
Dashboard panel Regime, edge score, LSTM confidence, suggested risk %,
dynamic R:R, anomaly status, attention split, bias band
Gate plots Optional forget/input/output gate activations (debug)
─────────────────────────────────────────────────────────────
SETTINGS
─────────────────────────────────────────────────────────────
Display Regime background, panel position/size, diagnostics, gates
LSTM Memory Cell Enable, memory speed, anomaly sensitivity, attention,
layer-norm, hidden-state plot amplitude
Regime-Bias Band Auto (stdev-scaled) vs manual, sensitivity, lookback,
floor/ceiling
Dynamic Risk Base risk %, ATR stop/target multiples, chop multiplier,
R:R cloud toggle and forward extend
─────────────────────────────────────────────────────────────
IMPORTANT NOTES
─────────────────────────────────────────────────────────────
-- This indicator has NOT been backtested or forward tested
-- No performance claims are made
-- Shared as a research tool for community review
-- Risk, stop, and target values are illustrative sizing suggestions, not
trade instructions
-- Results will vary by instrument, timeframe, and market conditions
-- This is not financial advice
-- Trading involves substantial risk of loss
-- Past results do not guarantee future performance
-- Use for educational and research purposes only
─────────────────────────────────────────────────────────────
Free for public use
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。