Add research methodology for EA-096 Intraday Range Extension

This document outlines the research methodology for EA-096 Intraday Range Extension, detailing the purpose, baseline principles, test parameters, performance metrics, and testing strategies.
2026-10-05 08:18:06 +07:00
parent 622fdaabfd
commit 1c52cee193
1 changed files with 281 additions and 0 deletions
@@ -0,0 +1,281 @@
+# Research Methodology
+
+## 1. Purpose
+
+This document defines the methodology used to research and validate EA-096_Intraday_Range_Extension.
+
+The objective is to separate:
+
+* strategy discovery
+* baseline measurement
+* parameter research
+* robustness testing
+* out-of-sample validation
+* forward testing
+
+---
+
+## 2. Baseline Principle
+
+The first completed test is treated as the baseline:
+
+`EA096-M1-BASELINE-001`
+
+The baseline must remain unchanged.
+
+All future experiments are compared against this reference.
+
+---
+
+## 3. Baseline Test
+
+### Market
+
+* Symbol: XAUUSD.PRO
+* Timeframe: M1
+
+### Period
+
+`2026.01.02 – 2026.03.31`
+
+### Data
+
+`100% real ticks`
+
+### Initial Capital
+
+`$1,000`
+
+### Core Strategy Parameters
+
+* ATR Period = 14
+* Extension ATR = 1.0
+* Minimum Range Bars = 30
+
+### Execution
+
+* Lot = 0.01
+* SL = 300 points
+* TP = 600 points
+* Break-Even = 150 points
+* Trailing Start = 200 points
+* Trailing Distance = 100 points
+
+---
+
+## 4. Baseline Performance
+
+| Metric          | Baseline |
+| --------------- | -------: |
+| Net Profit      |  +$30.80 |
+| Profit Factor   |     1.37 |
+| Expected Payoff |    $0.36 |
+| Recovery Factor |     2.11 |
+| Sharpe Ratio    |    49.56 |
+| Total Trades    |       86 |
+| Win Rate        |   53.49% |
+| Max Equity DD   |    1.44% |
+
+## The tester report records 86 trades and a maximum equity drawdown of 1.44%.
+
+## 5. Controlled Experiment Principle
+
+A research experiment should change as few variables as possible.
+
+For example:
+
+### Experiment A
+
+Change:
+
+`ExtensionATR`
+
+Keep everything else equal.
+
+### Experiment B
+
+Change:
+
+`InpMinRangeBars`
+
+Keep everything else equal.
+
+### Experiment C
+
+Change:
+
+`StopLoss / TakeProfit`
+
+Keep entry logic unchanged.
+
+This makes it possible to identify which component actually changes performance.
+
+---
+
+## 6. Primary Metrics
+
+Every experiment should record:
+
+### Profitability
+
+* Net Profit
+* Gross Profit
+* Gross Loss
+* Profit Factor
+* Expected Payoff
+
+### Risk
+
+* Maximum Balance Drawdown
+* Maximum Equity Drawdown
+* Relative Drawdown
+* Recovery Factor
+
+### Trade Quality
+
+* Total Trades
+* Win Rate
+* Average Profit Trade
+* Average Loss Trade
+* Largest Profit Trade
+* Largest Loss Trade
+
+### Consistency
+
+* Consecutive wins
+* Consecutive losses
+* Average holding time
+* Long versus short performance
+
+---
+
+## 7. Statistical Discipline
+
+A positive backtest with a small number of trades should not be treated as proof of a stable edge.
+
+The current EA-096 baseline contains:
+
+`86 trades`
+
+Therefore additional data is required before making a strong statistical conclusion.
+
+The research should prioritize increasing the sample size before aggressive parameter optimization.
+
+---
+
+## 8. Parameter Robustness
+
+A robust strategy should continue to perform reasonably when parameters are changed slightly.
+
+For example, if:
+
+`ExtensionATR = 1.0`
+
+produces a strong result but:
+
+`0.9`
+
+and
+
+`1.1`
+
+produce poor results, the strategy may be overly sensitive.
+
+The research should therefore examine parameter neighborhoods rather than searching for one perfect value.
+
+---
+
+## 9. Out-of-Sample Testing
+
+After identifying a candidate configuration:
+
+1. Freeze all parameters.
+2. Select data not used during discovery.
+3. Run the EA without additional optimization.
+4. Record all performance metrics.
+5. Compare the result with the baseline.
+
+A configuration should not be accepted solely because it performs well on the data used to discover it.
+
+---
+
+## 10. Walk-Forward Testing
+
+A later research stage should use walk-forward testing.
+
+General process:
+
+`Historical Data`
+
+↓
+
+`Research / Optimization Window`
+
+↓
+
+`Parameter Freeze`
+
+↓
+
+`Out-of-Sample Window`
+
+↓
+
+`Record Result`
+
+↓
+
+`Move Forward`
+
+The objective is to determine whether the strategy can adapt without repeatedly optimizing itself to past data.
+
+---
+
+## 11. Robustness Testing
+
+Robustness should eventually include:
+
+* different historical periods
+* different market regimes
+* nearby ATR thresholds
+* nearby range lengths
+* different exit parameters
+* BUY versus SELL separation
+* trading-session analysis
+
+The purpose is not to find the highest historical profit.
+
+The purpose is to determine whether the strategy's underlying behavior remains stable.
+
+---
+
+## 12. Production Readiness
+
+EA-096 should only be considered production-ready after:
+
+* sufficient historical trade sample
+* stable baseline comparison
+* parameter robustness
+* out-of-sample validation
+* walk-forward validation
+* forward testing
+* acceptable drawdown
+* acceptable execution behavior
+
+The current supplied baseline does **not** satisfy all of these requirements.
+
+Current status:
+
+`RESEARCH — NOT PRODUCTION READY`
+
+---
+
+## 13. Current Research Reference
+
+The official baseline for EA-096 is:
+
+`EA096-M1-BASELINE-001`
+
+Current conclusion:
+
+**The supplied test is positive and worth further research, but the sample size of 86 trades is insufficient to establish production-level robustness.**