交易系统实验室

关于 EA 开发、MT5 回测、风控与入场时机的实战笔记。不吹嘘,只谈机制。

提示:用“伦敦”“ATR”“点差”“日元”这类短词更容易搜到。

Four Techniques Tested, One Survived: the Tokyo Fix

We promised to test four techniques from our map and file the results, pass or fail. Three failed. One, the flow after the 09:55 Tokyo fix, held in every year since 2019, became an engine, and is now running live.

tokyo fix gotobi USDJPY EURJPY backtesting calendar effects momentum liquidity sweeps MT5 EA development
2026/09/11

A Map of FX Techniques, and Where Ours Have Been Tested

Every family of FX technique we know of, in one table, with what our own backtests say about each: proven, rejected, or not yet tested. Two patterns stand out, and they decide what we test next.

strategy backtesting mean reversion trend following breakout sessions MT5 EA development
2026/09/10
Why Combining Too Many Indicators Makes EAs Worse

Why Combining Too Many Indicators Makes EAs Worse

Adding more indicators often feels safer, but it usually makes Expert Advisors worse. Learn why indicator overload leads to signal conflict, overfitting, and fragile performance, and how simpler logic creates more robust EAs.

indicator overload signal conflict overfitting trading indicators EA development indicators
2026/02/20
这个博客怎么看
  • 只选一个时段(东京/伦敦/纽约),一次只验证一个想法
  • 先固定每笔的风险跑一段时间,等成绩稳定了再谈优化
  • 把有效的参数记下来,把结果一笔笔攒成记录