Trading Strategy Optimization Strategy Optimization
A systematic strategy optimization process, from diagnosing problems to achieving goals
Applicable Scenarios
- Win rate too low (< 40%)
- Returns below target
- Strategy performs poorly in specific market states
- Need to achieve a specific return target (e.g., 15%+)
Core Process
┌─────────────────────────────────────────────────────────────────┐
│ │
│ 策略優化迭代循環 │
│ │
│ ┌─────────┐ ┌─────────┐ ┌─────────┐ ┌─────────┐ │
│ │ ANALYZE │ → │ DIAGNOSE│ → │ RESEARCH│ → │IMPLEMENT│ │
│ │ 分析 │ │ 診斷 │ │ 研究 │ │ 實作 │ │
│ └────┬────┘ └────┬────┘ └────┬────┘ └────┬────┘ │
│ │ │ │ │ │
│ ▼ ▼ ▼ ▼ │
│ 收集指標 找根因 搜尋方案 實作變更 │
│ │
│ ┌─────────────────────────────────────────┐ │
│ │ │ │
│ ▼ │ │
│ ┌─────────┐ ┌─────────┐ │ │
│ │VALIDATE │ → │ ITERATE │ ────────────────────┘ │
│ │ 驗證 │ │ 迭代 │ 未達標 → 回到 ANALYZE │
│ └────┬────┘ └─────────┘ │
│ │ │
│ ▼ │
│ 達標 → 文檔化部署 │
│ │
└─────────────────────────────────────────────────────────────────┘
Phase 1: ANALYZE
Collect current performance metrics:
| Metric |
Description |
Healthy Standard |
| Win Rate |
Proportion of profitable trades |
> 40% |
| Return |
Total profit and loss |
Positive |
| Drawdown |
Maximum loss magnitude |
< 20% |
| Number of Trades |
Statistical significance |
> 100 trades |
| Expected Value |
Expected profit per trade |
Positive |
Expected Value Formula
Expectancy = (WinRate × AvgWin) - ((1-WinRate) × AvgLoss)
範例：
勝率: 54.5%, 平均贏: $7.42, 平均輸: $8.26
期望值: (0.545 × 7.42) - (0.455 × 8.26) = +$0.29/筆
正期望值 × 大量交易 = 穩定獲利
Phase 2: DIAGNOSE
Key: find the root cause, not the symptom
Common Root Causes
| Symptom |
Possible Root Cause |
Diagnostic Method |
| Low win rate |
Strategy not suited to market state |
Check ADX/market regime |
| Low return |
Poor SL/TP ratio |
Analyze profit/loss distribution |
| Large drawdown |
Leverage too high |
Calculate risk exposure |
| Few trades |
Conditions too strict |
Loosen entry conditions |
Market State Analysis
┌─────────────────────────────────────────────────────────────────┐
│ 市場狀態 vs 策略匹配 │
│ │
│ ADX < 25 (盤整) → Grid Trading, Mean Reversion │
│ ADX > 25 (趨勢) → Trend Following, Momentum │
│ 高波動 → Scalping, 減少部位 │
│ 低波動 → Grid Trading, Breakout 準備 │
│ │
│ ⚠️ 90% 時間市場在盤整，趨勢策略會持續虧損！ │
└─────────────────────────────────────────────────────────────────┘
Diagnosis Example
症狀: 勝率只有 25%
↓
分析: 使用 RSI+MACD 趨勢策略
↓
檢查: ADX 平均 < 25（90% 時間在盤整）
↓
根因: 趨勢策略用在盤整市場
↓
方向: 需要盤整市場策略（Grid Trading）
Phase 3: RESEARCH
Strategy-Market State Mapping Table
| Market State |
Recommended Strategy |
Reason |
| Consolidation (ADX < 25) |
Grid Trading |
Exploit price oscillation |
|
Mean Reversion |
Mean reversion |
|
Scalping |
Quick in, quick out |
| Trend (ADX > 25) |
Trend Following |
Follow the trend |
|
Momentum |
Momentum continuation |
|
Breakout |
Breakout follow-through |
| High Volatility |
Scalping |
Quick profit taking |
|
Reduce position size |
Control risk |
Sharp Edge: Research Data Validation
┌─────────────────────────────────────────────────────────────────┐
│ ⚠️ SE-1: 研究數據必須實測驗證 │
│ │
│ 嚴重度: critical │
│ │
│ 案例： │
│ - 研究聲稱圖表模式有 80-84% 成功率 │
│ - 實際回測：讓策略表現更差 │
│ - 原因：股市數據不適用於加密貨幣 │
│ │
│ 教訓： │
│ ❌ 看到研究數據就實作 │
│ ✅ 先用真實數據回測驗證 │
└─────────────────────────────────────────────────────────────────┘
Phase 4: IMPLEMENT
Parameter Adjustment Strategy
| Parameter |
Adjustment Direction |
Impact |
| SL/TP Ratio |
Symmetric (1:1) vs asymmetric |
Win rate vs profit/loss ratio |
| Leverage |
1x → 10x |
Return vs risk |
| Entry Threshold |
Loose vs strict |
Number of trades vs quality |
Key Findings
┌─────────────────────────────────────────────────────────────────┐
│ 優化發現 │
│ │
│ 1. 對稱 SL/TP 更好 │
│ - 1.8%/1.8% 產生 ~55% 勝率 │
│ - 比 1%/3% 的表現更好 │
│ │
│ 2. Grid Trading 在盤整市場 │
│ - 勝率從 25% 提升到 55% │
│ - 關鍵：ADX < 25 時啟用 │
│ │
│ 3. 更多交易 + 正期望值 │
│ - 小獲利累積成大報酬 │
│ - 191 筆交易 × $0.29/筆 = +$55 │
└─────────────────────────────────────────────────────────────────┘
Phase 5: VALIDATE
Backtest Requirements
| Item |
Minimum Requirement |
Recommended |
| Backtest Period |
60 days |
180 days |
| Number of Trades |
50 trades |
100+ trades |
| Out-of-sample testing |
Yes |
Required |
| Transaction costs |
Included |
Including slippage |
Validation Checklist
□ 勝率是否改善？
□ 報酬是否達標？
□ 回撤是否可接受？
□ 期望值是否為正？
□ 樣本外表現如何？
Phase 6: ITERATE
未達標 → 回到 ANALYZE
│
├── 分析新結果
├── 調整參數或策略
├── 重新驗證
└── 重複直到達標
Best Practices
- Diagnose before acting - Do not blindly tune parameters
- Change one variable at a time - Only then can you know what works
- Validate research conclusions - Theoretical data is not trustworthy
- Sufficient number of trades - Statistical significance
- Document successful experience - Write it as an ADR
Common Mistakes
| Mistake |
Consequence |
Fix |
| ❌ Only tuning parameters without changing the strategy |
Cannot solve the root problem |
Diagnose the root cause first |
| ❌ Backtest period too short |
Overfitting |
180+ days |
| ❌ Ignoring market regimes |
Strategy-market mismatch |
Use Regime Detection |
| ❌ Pursuing a high win rate |
Sacrificing the profit-loss ratio |
Focus on expected value |
Success Cases
優化前：
- 報酬: 0.46%
- 勝率: 23.5%
- 策略: RSI+MACD（趨勢策略用在盤整市場）
診斷：
- 90% 時間 ADX < 25（盤整）
- 趨勢策略不適合
優化後：
- 報酬: +19.93%
- 勝率: 54.5%
- 策略: Grid Trading + Regime Detection
關鍵改變：
1. 新增 Grid Trading 策略
2. 使用 Regime Detection 動態切換
3. 對稱 SL/TP (1.8%/1.8%)
4. 10x 槓桿 + 191 筆交易
Risk Warnings
⚠️ 高報酬伴隨高風險
- 上述案例最大回撤 48.84%
- 考慮降低槓桿 (5x) 換取更低風險
- 持續監控市場狀態變化
- 設定停損機制
Reference Configuration
# 經過驗證的配置
config = {
"leverage": 10.0, # 或 5.0 降低風險
"stop_loss_pct": 0.018, # 1.8%
"take_profit_pct": 0.018, # 1.8%
"top_n_signals": 3,
"min_confidence": 0.5,
"use_regime_selector": True,
}
Related Skills
finance/quant-trading - Quantitative Trading Fundamentalsfinance/investment-analysis - Investment Analysismethodology/knowledge-acquisition-4c - Learning New Domains
