交易系统实验室

28 Studies, One Survivor

Since September we have written up 28 studies, and exactly one mechanism has held up. Here is what failed, why most of it failed the same way, and what we now do with strategies that work for a while but cannot be proven.

research log regime forward testing Tokyo fix stop rules backtesting MT5 EA development

Three Ideas We Can't Prove Yet, So We're Watching Them

Three ideas looked good in our data, but only because we went looking. So we froze the rules, wrote down in advance what would count as success or failure, and are now watching them on data that did not exist when the rules were written. Why that takes years, and what we will publish either way.

forward testing out-of-sample backtesting FOMC US100 EURGBP mean reversion MT5 EA development