QUANT RESEARCH · MT5 · WALK-FORWARD VALIDATED

AI Trading Lab

AI 自動交易研究所

Let the strategy speak. Let the system execute.
No gurus, no screenshots, no hype — only rules, data, and out-of-sample evidence.

讓策略說話,讓系統執行。沒有大師、沒有截圖、沒有炒作 — 只有規則、數據,與樣本外的證據。
4,961
Daily Bars Tested
日線數據
97
Walk-Forward Folds
滾動驗證次數
18
Years OOS Data
樣本外年數
0
Curve Fitting
零過度擬合

Our Research Principles

研究原則

Three non-negotiables that separate a backtest from a trading system.
三條底線 — 把「回測曲線」變成「可交易系統」。

🔎

Walk-Forward Or Nothing

非滾動驗證,不算數

Every parameter we publish was selected on in-sample data and then tested on data it never saw. If a system cannot survive 90+ walk-forward folds, it does not ship.

所有參數必須經過滾動前進測試 — 在樣本外盲測存活 90 次以上,才有資格進入實盤候選。
📊

Real Broker Costs

真實成本測試

Spread, commission, swap and estimated slippage are modelled from live Vantage tick data. A strategy that only works with zero-cost assumptions is a fantasy.

點差、佣金、隔夜息與滑點全部用 Vantage 真實 tick 建模 — 脫離成本的策略只是幻想。
🛡

Robust, Not Optimised

穩健,而非最佳化

We reject parameter sets that win big on one configuration. We publish the average performance across a parameter neighbourhood, not the peak.

拒絕「單點最佳解」— 公佈的是參數鄰域的平均表現,而不是巔峰數字。

What You Get Inside The Lab

研究所內容

A weekly research drop, a transparent trade journal, and an open methodology.
每週研究報告 · 透明交易日誌 · 公開方法論。

Weekly Research Drop 每週研究

Strategy breakdowns 策略拆解WEEKLY
Walk-forward reports 滾動驗證報告WEEKLY
Parameter-neighbourhood plots 參數穩健圖INCLUDED
Failed experiments 失敗紀錄YES, HONESTLY

Live Track Record 實盤紀錄

Trade journal 交易日誌COMING Q2
Entry / exit timestamps 進出場時間戳Full disclosure
R-multiples, not % · 用 R 值,不用%Always
Drawdowns shown 回撤全部公開Yes
Why R-multiples matter · 為何用 R 值

A +5% return on a 20x levered FX account is not the same as +5% on a cash equity book. R-multiples normalise every trade to risk taken, which is the only way to compare strategies honestly.
20 倍槓桿 FX 帳戶的 +5%,同現金股票戶口的 +5% 完全唔同概念。用 R 值(每筆風險倍數)係唯一可以誠實比較策略嘅方法。

The Research Stack

技術堆疊

Boring, reliable, and reproducible by anyone with the same data.
無趣、穩定、任何人拿到相同數據都可以重現結果。

MT5
Execution Engine
執行引擎
Python
Research & Stats
研究 & 統計
Vantage
Live Tick Source
真實 Tick 來源
Bootstrap
Confidence CI
信心區間

Our 2026 Roadmap

2026 年度路線圖

Where we are and where we're heading.
現在的位置,與下一步的方向。

Q1 2026 · Methodology Published

Walk-forward framework, bootstrap CI, parameter-neighbourhood testing

Opened the research notebooks. Every future claim on this site will be reproducible from this baseline.
公開方法論與 Jupyter notebook — 所有未來宣稱都可以用呢套框架重現。

Q2 2026 · Paper Trading Phase

Donchian XAUUSD · RSI-Fade EURUSD · Regime-switch Indices

Three validated strategies entering 90-day live paper testing on Vantage RAW accounts. Results posted weekly.
三隻通過驗證嘅策略進入 90 日真實 paper test,每週公開結果。

Q3 2026 · Live Capital Deployment

First live trade journal goes public

Strategies that survive paper testing will be deployed with live capital and tracked on the Live page, trade-by-trade.
通過 paper test 嘅策略將會用真金白銀上線,每筆交易公開追蹤。

Q4 2026 · Portfolio Construction

Multi-strategy allocation · correlation-weighted risk budgeting

Once we have 3+ live strategies with uncorrelated return streams, we move from single-system to portfolio research.
當實盤策略超過 3 隻且相關性夠低,就會進入組合層研究。

Follow The Research, Not The Noise

跟隨研究,而非市場雜音

Join the Telegram channel and Discord server to get weekly drops, watch the walk-forward tests run, and see the paper trades land in real time. Free, forever.
加入 Telegram 頻道同 Discord — 每週研究更新、滾動驗證實況、paper trade 即時推送。永久免費。

Join The Lab · 加入研究所