Transform Your EA into an Autonomous ICT "Brain" — Create Custom Data for LLM Models with Smart Money Consciousness
Pending orders sit in invalidated zones while the market moves away, turning potential profits into guaranteed losses.
Your rigid rules can't adapt when volatility shifts, leaving stops exposed to institutional liquidity sweeps.
ICT setups invalidated in real-time while your robot keeps executing — no contextual intelligence, just rigid rules.
Your robot lacks contextual intelligence. Not logic — just rigid rules that explode the moment volatility shifts.
Meet the first professional-grade JSONL dataset engineered to fine-tune agents that dynamically manage orders in perfect harmony with ICT/Inner Circle Trading principles.
{
"messages": [{
"role": "system",
"content": "You are the PENDING ORDER MANAGEMENT Agent
using Smart money concepts...
VALIDATION CHECKLIST:
1. Is the pending order still aligned with
current market structure?
2. Has market structure changed
(new MSS/BOS invalidating setup)?
3. Are we still in/near the OTE zone..."
}]
}
Every entry replicates exactly how institutional traders think: MSS/BOS validation, OTE zone analysis, liquidity sweep detection, dynamic risk adjustment. No blunt "BUY/SELL" commands — only context-aware decisions with explicit reasoning.
Your agent detects invalidating MSS/BOS events and proactively cancels pending orders — like the USDCHF example where a bullish BOS invalidated a sell limit before it could trigger a loss.
Zero trades against active BOS/MSS: your LLM understands real-time market structure and maintains alignment with every decision it makes.
Integrates spread safety factors, live drawdown, and recent win rate into every decision — not afterthoughts, but core reasoning components.
Dataset optimized for the multi-agent architecture of the flagship EA — Order Management, Entry Validation, Risk Orchestration all working in concert.
Fine-tune with YOUR data → an agent that speaks your trading language, not some vendor's generic template. Customization is the edge.
After fine-tuning with this dataset, my agent reduced losses from stale limit orders by 63% after liquidity sweeps. Finally — an EA that thinks like an ICT trader.
Marco T.
Retail Quant — Germany
Everything you need to fine-tune an ICT-aware trading agent that thinks like a prop trader
Ready-to-run notebook that generates custom JSONL datasets from your own trading data and scenarios.
training.jsonl + validation.jsonl — 500+ real-world examples with market state, risk metrics & explicit reasoning.
Complete fine-tuning guide using Together.ai LoRA workflow. Works with Llama 3, Mistral, Qwen, and more.
Get the complete JSONL dataset with 500+ real-world examples, market states, risk metrics, and explicit ICT reasoning.
Use the step-by-step guide to fine-tune Llama 3, Mistral, or Qwen with the Together.ai LoRA workflow.
Connect with the multi-agent EA architecture for professional order management, entry validation, and risk orchestration.
Fine-tune any open-source model with the standardized JSONL format
This is a professional development toolkit for traders who understand that edge comes from customization — not buying someone else's generic model.
If you want a zero-effort "auto-trade" button → this product isn't for you.
This is for you if:
500+ real-world examples. Market state analysis. Risk metrics. Explicit reasoning. All in one professional JSONL dataset.