MT5 CatBoost Angle EA Upgrade – Stage 2: Pattern Engine + Self-Training + MAE/MFE Tracking

This project is Stage 2 of the previously completed MT5 CatBoost Angle EA upgrade.
The previous job has been closed automatically by the system, so this new job is created specifically to continue work with the same developer: Jhay Are Budomo.

This job includes only Milestone 2, with the following required deliverables:

Scope of Work (Stage 2 / Milestone 2)

1. Pattern Engine Implementation

• Angle-based feature binning

• Feature serialization using fixed-width representation

• 64-bit hashing (xxHash or MurmurHash)

• Collision detection and safe handling

• PatternTable implementation using CArrayObj + map index

• Lookup + update workflow

2. SymbolLearner Development
Each symbol must have an independent learner instance containing:

• PatternTable

• MAE/MFE state

• Decay timestamps

• Fully isolated storage

• Independent CSV file: pattern_SYMBOL.csv
  (No cross-symbol data mixing)

3. MAE/MFE Real-Time Tracking

• Register open trades on OnTradeTransaction

• Update tick-level favourable/adverse excursion

• Finalize MAE/MFE on close

• Must be efficient (O(1) per open trade), not based on HistoryScan

4. CSV Persistence System

• Write-ahead buffer (no per-update file writes)

• Atomic save (temp file + rename)

• Periodic flush or threshold-based flush

• CSV must correctly store pattern entries

5. Performance Requirements

• Must run multi-symbol backtests smoothly

• No memory leaks, no CPU spike issues

• Logs must show correct pattern updates and MAE/MFE values

• Working Pattern Engine implementation

• Working SymbolLearner with full reinforcement update logic

• MAE/MFE tracking operating correctly in Strategy Tester

• CSV files generated for at least two symbols (example output required)

• Short backtest log or screenshot showing:

  • pattern updates

  • MAE/MFE values

  • per-symbol separation

• Full MQ5 source code