Add EA-095 research methodology document

This document outlines the research methodology for EA-095_ATR_Stretch_2_0, including principles for controlled experiments, research order, performance metrics, and testing protocols.
2026-10-05 08:05:17 +07:00
parent 49089c3f15
commit ac22093e16
1 changed files with 276 additions and 0 deletions
@@ -0,0 +1,276 @@
+# EA-095 Research Methodology
+
+## 1. Purpose
+
+This document defines the methodology used to research and evaluate EA-095_ATR_Stretch_2_0.
+
+The purpose is to determine whether the strategy contains a repeatable trading edge while minimizing overfitting and uncontrolled parameter optimization.
+
+---
+
+## 2. Baseline Principle
+
+The first experiment is the unchanged baseline:
+
+`EA095-M1-BASELINE-001`
+
+The baseline is the permanent reference point for subsequent research.
+
+No future modification should overwrite or replace the baseline.
+
+---
+
+## 3. Controlled Experiment Principle
+
+Each research experiment should change only one major strategy component.
+
+Example:
+
+```text
+Baseline
+   ↓
+Change Timeframe only
+   ↓
+Compare Results
+```
+
+The following experiment should not simultaneously change:
+
+* timeframe;
+* ATR multiplier;
+* SL;
+* TP;
+* trailing stop;
+* spread;
+* entry rules.
+
+Changing multiple components simultaneously makes it difficult to determine which component caused the performance change.
+
+---
+
+## 4. Research Order
+
+Research should follow this sequence:
+
+### Stage 1 — Timeframe
+
+Evaluate the strategy across suitable timeframes while keeping the entry logic and risk model unchanged.
+
+Objective:
+
+Determine whether the ATR Stretch hypothesis behaves differently across market sampling frequencies.
+
+---
+
+### Stage 2 — ATR Multiplier
+
+Evaluate the ATR stretch threshold.
+
+The baseline is:
+
+`ATR(14) × 2.0`
+
+The purpose is to determine whether the strategy requires a different definition of an extreme price extension.
+
+Only the ATR multiplier should be changed during this stage.
+
+---
+
+### Stage 3 — Entry Confirmation
+
+Evaluate whether the current recovery-candle confirmation contributes positively to the strategy.
+
+The baseline confirmation consists of:
+
+### BUY
+
+* Previous candle closes below EMA - ATR × Multiplier.
+* Current closed candle is bullish.
+* Current close is above the previous close.
+* Current close remains below EMA.
+
+### SELL
+
+* Previous candle closes above EMA + ATR × Multiplier.
+* Current closed candle is bearish.
+* Current close is below the previous close.
+* Current close remains above EMA.
+
+Potential future tests should isolate individual confirmation components rather than changing the entire entry system simultaneously.
+
+---
+
+### Stage 4 — Trading Session
+
+Evaluate whether performance is concentrated in specific market sessions or hours.
+
+The objective is to determine whether the strategy's edge, if any, is dependent on a particular trading period.
+
+---
+
+### Stage 5 — Directional Analysis
+
+Evaluate BUY and SELL trades independently.
+
+The baseline report contains separate BUY and SELL trade statistics.
+
+The research should determine whether one direction contributes disproportionately to the result.
+
+---
+
+### Stage 6 — Exit Management
+
+Evaluate the effect of:
+
+* Stop Loss;
+* Take Profit;
+* Break Even;
+* Trailing Stop.
+
+Exit parameters should be tested separately from entry logic whenever possible.
+
+---
+
+## 5. Performance Metrics
+
+Every experiment should record at minimum:
+
+* Net Profit
+* Profit Factor
+* Expected Payoff
+* Maximum Equity Drawdown
+* Total Trades
+* Profit Trades %
+* Loss Trades %
+* Average Profit Trade
+* Average Loss Trade
+* Maximum Consecutive Wins
+* Maximum Consecutive Losses
+* Sharpe Ratio
+* Recovery Factor
+
+The baseline values should remain visible for comparison.
+
+---
+
+## 6. Statistical Discipline
+
+A profitable result from a single historical period should not automatically be treated as proof of a trading edge.
+
+Research should distinguish between:
+
+```text
+In-Sample Discovery
+        ↓
+Independent Validation
+        ↓
+Robustness Testing
+        ↓
+Forward Testing
+```
+
+The objective is to determine whether performance survives changes in historical period, market conditions, and reasonable parameter variation.
+
+---
+
+## 7. Optimization Policy
+
+Broad optimization should remain blocked during the early research stage.
+
+Optimization should only be considered after controlled experiments identify which strategy components appear to have meaningful influence on performance.
+
+The objective is not to find the single best historical parameter combination.
+
+The objective is to identify a robust parameter region that behaves consistently.
+
+---
+
+## 8. Out-of-Sample Testing
+
+Once a candidate configuration is identified, it should be tested on historical data that was not used to discover the configuration.
+
+The out-of-sample test must use the finalized candidate parameters without further adjustment.
+
+If the strategy fails out-of-sample, the candidate should not be considered validated.
+
+---
+
+## 9. Robustness Testing
+
+Robustness testing should examine whether the strategy remains viable when reasonable changes are applied.
+
+Examples include:
+
+* small changes to ATR multiplier;
+* small changes to EMA period;
+* different historical periods;
+* different execution conditions;
+* reasonable spread variations;
+* BUY/SELL asymmetry.
+
+A strategy that only works at one exact parameter value should be treated with caution.
+
+---
+
+## 10. Forward Testing
+
+Forward testing should only begin after:
+
+1. baseline research is completed;
+2. a candidate configuration has been identified;
+3. out-of-sample testing has been completed;
+4. robustness testing provides acceptable evidence.
+
+Forward testing should be treated as a separate validation stage.
+
+---
+
+## 11. Production Readiness
+
+EA-095_ATR_Stretch_2_0 should only be considered for production after the research process provides evidence of:
+
+* positive expectancy;
+* acceptable drawdown;
+* stable performance;
+* sufficient trade sample;
+* out-of-sample consistency;
+* robustness;
+* and successful forward testing.
+
+The current baseline does **not** meet this standard.
+
+---
+
+## 12. Current Baseline Reference
+
+```text
+Experiment ID : EA095-M1-BASELINE-001
+Symbol        : XAUUSD.PRO
+Timeframe     : M1
+Period        : 2026.01.02 – 2026.03.31
+Initial Fund  : $1,000
+Lot           : 0.01
+EMA           : 20
+ATR           : 14
+ATR Multiplier: 2.0
+SL            : 300 points
+TP            : 600 points
+BE            : 150 points
+Trailing      : Start 200 / Distance 100 / Step 10
+Max Spread    : 30 points
+```
+
+Baseline result:
+
+```text
+Net Profit            : -$330.84
+Profit Factor         : 0.95
+Expected Payoff       : -$0.07
+Maximum Equity DD     : 48.57%
+Total Trades          : 4,971
+Win Rate              : 50.65%
+Classification        : FAIL
+Production Ready      : NO
+```
+
+The baseline must remain unchanged and serve as the reference for all future EA-095 research.