OPEN-SOURCE SCRIPT
BioSwarm Imprinter™
BioSwarm Imprinter™ — Agent-Based Consensus for Traders
What it is
BioSwarm Imprinter™ is a non-repainting, agent-based sentiment oscillator. It fuses many short-to-medium lookback “opinions” into one 0–100 consensus line that is easy to read at a glance (50 = neutral, >55 bullish bias, <45 bearish bias). The engine borrows from swarm intelligence: many simple voters (agents) adapt their influence over time based on how well they’ve been predicting price, so the crowd gets smarter as conditions change.
Use it to:
• Detect emerging trends sooner without overreacting to noise.
• Filter mean-reversion vs continuation opportunities.
• Gate entries with a confidence score that reflects both strength and persistence of the move.
• Combine with your execution tools (VWAP/ORB/levels) as a state filter rather than a trade signal by itself.
⸻
Why it’s different
• Swarm learning: Each agent improves or decays its “fitness” depending on whether its vote matched the next bar’s direction. High-fitness agents matter more; weak agents fade.
• Multi-horizon by design: The crowd is composed of fixed, simple lookbacks spread from lenMin to lenMax. You get a blended, robust view instead of a single fragile parameter.
• Two complementary lenses: Each agent evaluates RSI-style balance (via Wilder’s RMA) and momentum (EMA deviation). You decide the weight of each.
• No repaint, no MTF pitfalls: Everything runs on the chart’s timeframe with bar-close confirmation; no request.security() or forward references.
• Actionable UI: A clean consensus line, optional regime background, confidence heat, and triangle markers when thresholds are crossed.
⸻
What you see on the chart
• Consensus line (0–100): Smoothed to your preference; color/area makes bull/bear zones obvious.
• Regime coloring (optional): Light green in bull zone, light red in bear zone; neutral otherwise.
• Confidence heat: A small gauge/number (0–100) that combines distance from neutral and recent persistence.
• Markers (optional): Triangles when consensus crosses up through your bull threshold (e.g., 55) or down through your bear threshold (e.g., 45).
• Info panel (optional): Consensus value, regime, confidence, number of agents, and basic diagnostics.
⸻
How it works (under the hood)
1. Horizon bins: The range [lenMin … lenMax] is divided into numBins. Each bin has a fixed, simple integer length (crucial for Pine’s safety rules).
2. Per-bin features (computed every bar):
• RSI-style balance using Wilder’s RMA (not ta.rsi()), then mapped to −1…+1.
• Momentum as (close − EMA(L)) / EMA(L) (dimensionless drift).
3. Agent vote: For its assigned bin, an agent forms a weighted score: score = wRSI*RSI_like + wMOM*Momentum. A small dead-band near zero suppresses chop; votes are +1/−1/0.
4. Fitness update (bar close): If the agent’s previous vote agreed with the next bar’s direction, multiply its fitness by learnGain; otherwise by learnPain. Fitness is clamped so it never explodes or dies.
5. Consensus: Weighted average of all votes using fitness as weights → map to 0–100 and smooth with EMA.
Why it doesn’t repaint:
• No future references, no MTF resampling, fitness updates only on confirmed bars.
• All TA primitives (RMA/EMA/deltas) are computed every bar unconditionally.
⸻
Signals & confidence
• Bullish bias: consensus ≥ bullThr (e.g., 55).
• Bearish bias: consensus ≤ bearThr (e.g., 45).
• Confidence (0–100):
• Distance score: how far consensus is from 50.
• Momentum score: how strong the recent change is versus its recent average.
• Combined into a single gate; start filtering entries at ≥60 for higher quality.
Tip: For range sessions, raise thresholds (60/40) and increase smoothing; for momentum sessions, lower smoothing and keep thresholds at 55/45.
⸻
Inputs you’ll actually tune
• Agents & horizons:
• N_agents (e.g., 64–128)
• lenMin / lenMax (e.g., 6–30 intraday, 10–60 swing)
• numBins (e.g., 12–24)
• Weights & smoothing:
• wRSI vs wMOM (e.g., 0.7/0.3 for FX & indices; 0.6/0.4 for crypto)
• deadBand (0.03–0.08)
• consSmooth (3–8)
• Thresholds & hygiene:
• bullThr/bearThr (55/45 default)
• cooldownBars to avoid signal spam
⸻
Playbooks (ready-to-use)
1) Breakout / Trend continuation
• Timeframe: 15m–1h for day/swing.
• Filter: Take longs only when consensus > 55 and confidence ≥ 60.
• Execution: Use your ORB/VWAP/pullback trigger for entry. Trail with swing lows or 1.5×ATR. Exit on a close back under 50 or when a bearish signal prints.
2) Mean reversion (fade)
• When: Sideways days or low-volatility clusters.
• Setup: Increase deadBand and consSmooth.
• Signal: Bearish fades when consensus rolls over below ≈55 but stays above 50; bullish fades when it rolls up above ≈45 but stays below 50.
• Targets: The neutral zone (~50) as the first take-profit.
3) Multi-TF alignment
• Keep BioSwarm on 1H for bias, execute on 5–15m:
• Only take entries in the direction of the 1H consensus.
• Skip counter-bias scalps unless confidence is very low (explicit mean-reversion plan).
⸻
Integrations that work
• DynamoSent Pro+ (macro bias): Only act when macro bias and swarm consensus agree.
• ORB + Session VWAP Pro: Trade London/NY ORB breakouts that retest while consensus >55 (long) or <45 (short).
• Levels/Orderflow: BioSwarm is your “go / no-go”; execution stays with your usual triggers.
⸻
Quick start
1. Drop the indicator on a 1H chart.
2. Start with: N_agents=64, lenMin=6, lenMax=30, numBins=16, deadBand=0.06, consSmooth=5, thresholds 55/45.
3. Trade only when confidence ≥ 60.
4. Add your favorite execution tool (VWAP/levels/OR) for entries & exits.
⸻
Non-repainting & safety notes
• No request.security(); no hidden lookahead.
• Bar-close confirmation for fitness and signals.
• All TA calls are unconditional (no “sometimes called” warnings).
• No series-length inputs to RSI/EMA — we use RMA/EMA formulas that accept fixed simple ints per bin.
⸻
Known limits & tips
• Too many signals? Raise deadBand, increase consSmooth, widen thresholds to 60/40.
• Too few signals? Lower deadBand, reduce consSmooth, narrow thresholds to 53/47.
• Over-fitting risk: Keep learnGain/learnPain modest (e.g., ×1.04 / ×0.96).
• Compute load: Large N_agents × numBins is heavier; scale to your device.
⸻
Example recipes
EURUSD 1H (swing):
lenMin=8, lenMax=34, numBins=16, wRSI=0.7, wMOM=0.3, deadBand=0.06, consSmooth=6, thr=55/45
Buy breakouts when consensus >55 and confidence ≥60; confirm with 5–15m pullback to VWAP or level.
SPY 15m (US session):
lenMin=6, lenMax=24, numBins=12, consSmooth=4, deadBand=0.05
On trend days, stay with longs as long as consensus >55; add on shallow pullbacks.
BTC 1H (24/7):
Increase momentum weight: wRSI=0.6, wMOM=0.4, extend lenMax to ~50. Use dynamic stops (ATR) and partials on strong verticals.
⸻
Final word
BioSwarm is a state engine: it tells you when the market is primed to continue or mean-revert. Pair it with your entries and risk framework to turn that state into trades. If you’d like, I can supply a companion strategy template that consumes the consensus and back-tests the three playbooks (Breakout/Fade/Flip) with standard risk management.
What it is
BioSwarm Imprinter™ is a non-repainting, agent-based sentiment oscillator. It fuses many short-to-medium lookback “opinions” into one 0–100 consensus line that is easy to read at a glance (50 = neutral, >55 bullish bias, <45 bearish bias). The engine borrows from swarm intelligence: many simple voters (agents) adapt their influence over time based on how well they’ve been predicting price, so the crowd gets smarter as conditions change.
Use it to:
• Detect emerging trends sooner without overreacting to noise.
• Filter mean-reversion vs continuation opportunities.
• Gate entries with a confidence score that reflects both strength and persistence of the move.
• Combine with your execution tools (VWAP/ORB/levels) as a state filter rather than a trade signal by itself.
⸻
Why it’s different
• Swarm learning: Each agent improves or decays its “fitness” depending on whether its vote matched the next bar’s direction. High-fitness agents matter more; weak agents fade.
• Multi-horizon by design: The crowd is composed of fixed, simple lookbacks spread from lenMin to lenMax. You get a blended, robust view instead of a single fragile parameter.
• Two complementary lenses: Each agent evaluates RSI-style balance (via Wilder’s RMA) and momentum (EMA deviation). You decide the weight of each.
• No repaint, no MTF pitfalls: Everything runs on the chart’s timeframe with bar-close confirmation; no request.security() or forward references.
• Actionable UI: A clean consensus line, optional regime background, confidence heat, and triangle markers when thresholds are crossed.
⸻
What you see on the chart
• Consensus line (0–100): Smoothed to your preference; color/area makes bull/bear zones obvious.
• Regime coloring (optional): Light green in bull zone, light red in bear zone; neutral otherwise.
• Confidence heat: A small gauge/number (0–100) that combines distance from neutral and recent persistence.
• Markers (optional): Triangles when consensus crosses up through your bull threshold (e.g., 55) or down through your bear threshold (e.g., 45).
• Info panel (optional): Consensus value, regime, confidence, number of agents, and basic diagnostics.
⸻
How it works (under the hood)
1. Horizon bins: The range [lenMin … lenMax] is divided into numBins. Each bin has a fixed, simple integer length (crucial for Pine’s safety rules).
2. Per-bin features (computed every bar):
• RSI-style balance using Wilder’s RMA (not ta.rsi()), then mapped to −1…+1.
• Momentum as (close − EMA(L)) / EMA(L) (dimensionless drift).
3. Agent vote: For its assigned bin, an agent forms a weighted score: score = wRSI*RSI_like + wMOM*Momentum. A small dead-band near zero suppresses chop; votes are +1/−1/0.
4. Fitness update (bar close): If the agent’s previous vote agreed with the next bar’s direction, multiply its fitness by learnGain; otherwise by learnPain. Fitness is clamped so it never explodes or dies.
5. Consensus: Weighted average of all votes using fitness as weights → map to 0–100 and smooth with EMA.
Why it doesn’t repaint:
• No future references, no MTF resampling, fitness updates only on confirmed bars.
• All TA primitives (RMA/EMA/deltas) are computed every bar unconditionally.
⸻
Signals & confidence
• Bullish bias: consensus ≥ bullThr (e.g., 55).
• Bearish bias: consensus ≤ bearThr (e.g., 45).
• Confidence (0–100):
• Distance score: how far consensus is from 50.
• Momentum score: how strong the recent change is versus its recent average.
• Combined into a single gate; start filtering entries at ≥60 for higher quality.
Tip: For range sessions, raise thresholds (60/40) and increase smoothing; for momentum sessions, lower smoothing and keep thresholds at 55/45.
⸻
Inputs you’ll actually tune
• Agents & horizons:
• N_agents (e.g., 64–128)
• lenMin / lenMax (e.g., 6–30 intraday, 10–60 swing)
• numBins (e.g., 12–24)
• Weights & smoothing:
• wRSI vs wMOM (e.g., 0.7/0.3 for FX & indices; 0.6/0.4 for crypto)
• deadBand (0.03–0.08)
• consSmooth (3–8)
• Thresholds & hygiene:
• bullThr/bearThr (55/45 default)
• cooldownBars to avoid signal spam
⸻
Playbooks (ready-to-use)
1) Breakout / Trend continuation
• Timeframe: 15m–1h for day/swing.
• Filter: Take longs only when consensus > 55 and confidence ≥ 60.
• Execution: Use your ORB/VWAP/pullback trigger for entry. Trail with swing lows or 1.5×ATR. Exit on a close back under 50 or when a bearish signal prints.
2) Mean reversion (fade)
• When: Sideways days or low-volatility clusters.
• Setup: Increase deadBand and consSmooth.
• Signal: Bearish fades when consensus rolls over below ≈55 but stays above 50; bullish fades when it rolls up above ≈45 but stays below 50.
• Targets: The neutral zone (~50) as the first take-profit.
3) Multi-TF alignment
• Keep BioSwarm on 1H for bias, execute on 5–15m:
• Only take entries in the direction of the 1H consensus.
• Skip counter-bias scalps unless confidence is very low (explicit mean-reversion plan).
⸻
Integrations that work
• DynamoSent Pro+ (macro bias): Only act when macro bias and swarm consensus agree.
• ORB + Session VWAP Pro: Trade London/NY ORB breakouts that retest while consensus >55 (long) or <45 (short).
• Levels/Orderflow: BioSwarm is your “go / no-go”; execution stays with your usual triggers.
⸻
Quick start
1. Drop the indicator on a 1H chart.
2. Start with: N_agents=64, lenMin=6, lenMax=30, numBins=16, deadBand=0.06, consSmooth=5, thresholds 55/45.
3. Trade only when confidence ≥ 60.
4. Add your favorite execution tool (VWAP/levels/OR) for entries & exits.
⸻
Non-repainting & safety notes
• No request.security(); no hidden lookahead.
• Bar-close confirmation for fitness and signals.
• All TA calls are unconditional (no “sometimes called” warnings).
• No series-length inputs to RSI/EMA — we use RMA/EMA formulas that accept fixed simple ints per bin.
⸻
Known limits & tips
• Too many signals? Raise deadBand, increase consSmooth, widen thresholds to 60/40.
• Too few signals? Lower deadBand, reduce consSmooth, narrow thresholds to 53/47.
• Over-fitting risk: Keep learnGain/learnPain modest (e.g., ×1.04 / ×0.96).
• Compute load: Large N_agents × numBins is heavier; scale to your device.
⸻
Example recipes
EURUSD 1H (swing):
lenMin=8, lenMax=34, numBins=16, wRSI=0.7, wMOM=0.3, deadBand=0.06, consSmooth=6, thr=55/45
Buy breakouts when consensus >55 and confidence ≥60; confirm with 5–15m pullback to VWAP or level.
SPY 15m (US session):
lenMin=6, lenMax=24, numBins=12, consSmooth=4, deadBand=0.05
On trend days, stay with longs as long as consensus >55; add on shallow pullbacks.
BTC 1H (24/7):
Increase momentum weight: wRSI=0.6, wMOM=0.4, extend lenMax to ~50. Use dynamic stops (ATR) and partials on strong verticals.
⸻
Final word
BioSwarm is a state engine: it tells you when the market is primed to continue or mean-revert. Pair it with your entries and risk framework to turn that state into trades. If you’d like, I can supply a companion strategy template that consumes the consensus and back-tests the three playbooks (Breakout/Fade/Flip) with standard risk management.
סקריפט קוד פתוח
ברוח TradingView אמיתית, היוצר של הסקריפט הזה הפך אותו לקוד פתוח, כך שסוחרים יכולים לבדוק ולאמת את הפונקציונליות שלו. כל הכבוד למחבר! למרות שאתה יכול להשתמש בו בחינם, זכור שפרסום מחדש של הקוד כפוף לכללי הבית שלנו.
כתב ויתור
המידע והפרסומים אינם אמורים להיות, ואינם מהווים, עצות פיננסיות, השקעות, מסחר או סוגים אחרים של עצות או המלצות שסופקו או מאושרים על ידי TradingView. קרא עוד בתנאים וההגבלות.
סקריפט קוד פתוח
ברוח TradingView אמיתית, היוצר של הסקריפט הזה הפך אותו לקוד פתוח, כך שסוחרים יכולים לבדוק ולאמת את הפונקציונליות שלו. כל הכבוד למחבר! למרות שאתה יכול להשתמש בו בחינם, זכור שפרסום מחדש של הקוד כפוף לכללי הבית שלנו.
כתב ויתור
המידע והפרסומים אינם אמורים להיות, ואינם מהווים, עצות פיננסיות, השקעות, מסחר או סוגים אחרים של עצות או המלצות שסופקו או מאושרים על ידי TradingView. קרא עוד בתנאים וההגבלות.