設計透明，預設嚴謹
Trading.hk 上的每份報告都遵循相同的透明、可重複的方法論。我們公開我們的流程，讓您可以自行評估我們研究的質量。
Transparent
Every assumption, every parameter, every test period is disclosed in the report. No black boxes.
Reproducible
The same inputs always produce the same outputs. Our engine is deterministic and version-controlled.
Rigorous
Four mandatory bias controls applied to every single backtest — no exceptions, no shortcuts.
Methodology Chapters
Select a chapter to explore each component in full detail.
Data Sources & Coverage
Where our data comes from, how it is validated, which exchanges and asset classes are covered, and how we handle corporate actions.
Backtesting Engine
Event-driven architecture, execution simulation, realistic transaction costs, position sizing models, and portfolio-level simulation across multiple instruments.
Bias Controls & Validation
The six sources of bias that invalidate most retail backtests — and exactly how we eliminate each one. Includes walk-forward testing and out-of-sample validation.
Performance Metrics
Complete glossary of every metric used in our reports — formulas, interpretation guides, thresholds, and why each metric matters for evaluating a strategy.
At a Glance
| Aspect | Our Approach | Why It Matters |
|---|---|---|
| Data history | Minimum 5 years, up to 20+ years | Covers multiple market regimes |
| Engine type | Event-driven, bar-by-bar | Eliminates look-ahead bias at the root |
| Transaction costs | 0.1% round-trip (HK), 0.05% (US) | Realistic, conservative assumptions |
| Fill assumption | Next-bar open after signal bar | No idealized same-bar fills |
| Universe | Includes delisted stocks | Prevents survivorship bias |
| Validation | In-sample + out-of-sample holdout | Ensures results are not curve-fitted |
| Metrics reported | 12+ standardized metrics per report | Consistent, comparable evaluation |
| Optimization | Walk-forward, not best-fit | Simulates real adaptive management |
重要聲明
Trading.hk 上發布的所有研究僅供教育及資訊參考用途，不構成投資建議、推薦或買賣證券的要約。我們的報告不是個人化投資建議 — 同一報告對所有訂閱者開放。過往表現及回測結果不保證未來回報。