What kind of demo is considered good?

Hi everyone,

I’ve generated several strategies by SQX. They passed Monte Carlo testing and WFM, and showed decent backtesting results on MT5, but after deploying them to an MT5 demo account for two weeks, the performance was not so good, in fact, most strategies are losing money. I understand this is part of trading. Drawdowns are inevitable, and I’m only targeting a ​​30% win rate​​.

My question is:

But I still can’t tell if my demo performance is good or bad. I don’t know how to analyze it.​

​​Do you have systematic methods to analyze demo performance?​​ When a demo account is running, ​​do you follow specific procedures or workflows to evaluate it?​​ What qualifies as a “good” demo? And at what point should a strategy transition from demo to live trading? The only method I know comes from Zdenek’s ebook: let the demo run for a period, then compare its performance with backtesting results.

Thank you.

You need large enough sample size on demo. At least 50 – 100 trades. If the performance deviates a lot compared to the backtest it is a sign of overfitting. But there is another way how to approach it – frequent re-optimization such as weekly, or every other week etc. This could be an alternative way.

Thank you Tomas, re-optimization seems controversial, could it lead to overfitting?