At what point do you stop optimizing an EA?

New comment
 

This is something I genuinely struggle with.

Whenever I improve a strategy, I can usually find another parameter to tweak, another filter to add, or another optimization run to perform.

At some point, though, it feels like I'm fitting the system to the past rather than building something robust for the future.

How do you personally decide that an EA is "good enough" and ready for live trading?

Are there specific metrics or rules you follow before you stop optimizing and start trading?

 
Eloise Quinn:

This is something I genuinely struggle with.

Whenever I improve a strategy, I can usually find another parameter to tweak, another filter to add, or another optimization run to perform.

At some point, though, it feels like I'm fitting the system to the past rather than building something robust for the future.

How do you personally decide that an EA is "good enough" and ready for live trading?

Are there specific metrics or rules you follow before you stop optimizing and start trading?

Stop when your out-of-sample results stop improving.

That's the only metric that actually matters.

If tweaking parameters keeps making your backtest prettier but your unseen data stays flat or gets worse, you're curve-fitting, not improving.

A good rule: if you can't explain why a setting works with real market logic, remove it.

Fewer inputs usually means more robust. Go live when it's consistently decent on data it's never seen, not when it looks perfect on data it already knows.

 
Eloise Quinn:

This is something I genuinely struggle with.

Whenever I improve a strategy, I can usually find another parameter to tweak, another filter to add, or another optimization run to perform.

At some point, though, it feels like I'm fitting the system to the past rather than building something robust for the future.

How do you personally decide that an EA is "good enough" and ready for live trading?

Are there specific metrics or rules you follow before you stop optimizing and start trading?

I stopped when I hadn't found any new ideas for how to enter, how to exit and so on. Sometimes this idea gives rise to a new EA that is very different from the previous EA or just a version change due to additional features. New ideas often emerge for creating EAs, after they are done and backtested several times the results are disappointing and in the end I delete the EA. Maybe it is human nature to seek better achievements. I think the same thing applies to any development, for example OS, AI etc. which also continue to develop. EA is also like that, if there is a new idea, it will be developed, if there isn't, look at the trading carried out by the EA that has been created and maybe you will find a new idea. This is the same as the Software Development Life Cycle (SDLC).
 
Great perspective on treating EA development like an ongoing SDLC. It makes a lot of sense that we don't necessarily 'stop' optimizing, we just shift our focus to whether the current logic actually holds up. For those who do 'go live' with a version, how long do you usually run it on a demo before you feel confident enough to push it to a live account?
 
Eloise Quinn:

This is something I genuinely struggle with.

Whenever I improve a strategy, I can usually find another parameter to tweak, another filter to add, or another optimization run to perform.

At some point, though, it feels like I'm fitting the system to the past rather than building something robust for the future.

How do you personally decide that an EA is "good enough" and ready for live trading?

Are there specific metrics or rules you follow before you stop optimizing and start trading?

I am saying this from my own experience, never start optimizing an EA before your core logic is solid. Your core logic, before any optimization should give you a good practical result. Optimization is for the end product not the development version. 
 
Stop optimizing the EA at the point that the same backtest for the same time period is worse than before optimizing. The strategy just has to be executed properly. Take some losses and don't try to aim for a higher win rate if the strategy is more alpha with a lower win rate.
 
You have to stop thinking like it's a game of clean recovery. This is developer thinking and not trader thinking. The market isn't a game of recovery, it's strategy and consistency.
New comment