Add research methodology for EA-076 Round Number Break

This document outlines the research methodology for EA-076 Round Number Break, detailing the purpose, strategy specifications, backtesting procedures, and validation criteria.

docs/EA-076_Round_Number_Break/methodology.md

@@ -0,0 +1,117 @@
+
+# EA-076 — Research Methodology
+
+## 1. Purpose
+
+This document defines the research, testing and validation process for EA-076 Round Number Break.
+
+The purpose is to maintain a reproducible record of strategy development, historical testing and subsequent validation.
+
+## 2. Source of Truth
+
+Every reported result must be traceable to:
+- An identifiable EA source-code version.
+- A complete input configuration.
+- An original MT5 Strategy Tester report.
+- The instrument and broker environment.
+- The historical test period and execution settings.
+
+Never replace a historical result with an optimized result without preserving the original evidence.
+
+## 3. Strategy Specification
+
+EA-076 is an intraday round-number breakout strategy for MetaTrader 5.
+
+The active implementation calculates relevant upper and lower round-number boundaries from the preceding completed candle's closing price.
+
+A buy signal is generated when the next completed candle closes above the upper boundary plus the configured breakout buffer.
+
+A sell signal is generated when the next completed candle closes below the lower boundary minus the configured breakout buffer.
+
+Signals are evaluated once per new candle.
+
+The default round-number interval is 10.0 price units.
+
+The breakout lookback parameter is retained in the source but is not used by the active round-number signal calculation.
+
+## 4. Baseline Configuration
+
+- Symbol: XAUUSD.PRO
+- Timeframe: M1
+- Round Number Step: 10.0
+- Lot Size: 0.01
+- Stop Loss: 300 points
+- Take Profit: 600 points
+- Breakout Buffer: 0 points
+- Break Even: Enabled
+- Trailing Stop: Enabled
+- Maximum Spread: 30 points
+
+The original source and complete MT5 report contain the authoritative parameter definitions.
+
+## 5. Backtesting
+
+Use MT5 Strategy Tester and preserve the selected tick-model configuration.
+
+For every test, record:
+1. Source-code version.
+2. Parameter set.
+3. Broker and symbol.
+4. Date range.
+5. Initial deposit and leverage.
+6. Tick-model and history quality.
+7. Performance metrics.
+8. Original HTML report and charts.
+
+The supplied baseline used 100% real ticks over January–March 2026.
+
+## 6. Experiment Control
+
+Modify a defined parameter group for each experiment.
+
+Record the hypothesis, baseline, modified inputs and resulting metrics.
+
+Preserve failed experiments. Do not retrospectively present an optimized configuration as the original baseline.
+
+## 7. Validation
+
+Separate development data from validation data.
+
+A configuration selected using one historical period must be evaluated on previously unused data.
+
+Where practical, investigate:
+- Different historical market conditions.
+- Different spread and slippage assumptions.
+- Different broker environments.
+- Sensitivity to small parameter changes.
+
+A positive in-sample result alone is not sufficient for live deployment.
+
+## 8. Acceptance Criteria
+
+Define numerical acceptance thresholds before conducting optimization.
+
+At minimum, evaluate profitability, drawdown, trade count, monthly consistency and out-of-sample performance.
+
+A candidate must not be marked as validated solely because it satisfies one performance metric.
+
+The current baseline is unprofitable and is not approved for live deployment.
+
+## 9. Reproducibility
+
+Preserve original artifacts in their designated directories.
+
+Do not modify historical reports or overwrite the original source without version control.
+
+Any new experiment must reference its corresponding source revision and input configuration.
+
+## 10. Current Status
+
+Historical baseline: Documented.  
+Independent reproduction: Pending.  
+Optimization: Pending.  
+Out-of-sample testing: Pending.  
+Forward testing: Pending.  
+Live deployment approval: Not granted.  
+
+All future changes and conclusions must be supported by reproducible test evidence.