It works on one instrument. Does that mean anything?

It means something, but less than it feels like, and there is a specific way to find out how much.

One market is one observation

If you tested nine instruments and are describing the one that worked, you have run a search and the winner is the expected outcome of searching. The result is about your selection, not about the strategy, and the only thing that distinguishes the two cases is whether you decided the instrument before or after you looked.

This is why the brief matters more than it sounds: writing down which market you intend to test, before testing, converts a search back into an experiment.

What a cross-instrument test actually tells you

If the idea is real, some trace of it should appear on related markets, even weakly. A mechanism that exists on one pair and nowhere else is either extremely specific to that market's structure, which is possible and worth stating explicitly, or it is not there at all.

Do not require it to be profitable everywhere. Require the behaviour to be visible, and be suspicious when it is visible nowhere else.

Correlated markets are not independent tests

Testing the major dollar pairs and finding the same result four times is close to testing the dollar once. The same is true of index CFDs that track the same underlying, and of two crypto majors in a trending week.

For the test to add evidence the markets have to be able to disagree. Pick instruments that are driven by different things, and treat a family of correlated results as one observation rather than four.

What to do with a result that survives

Write down where it worked, where it did not, and what you think the difference is. That last sentence is the valuable one: a strategy with an explanation for its own failures is testable, and one that works everywhere for no stated reason usually stops working for no stated reason too.

Questions

Should a good strategy work on every instrument?

No, and expecting that leads to weaker strategies. Many real effects are specific to a market's structure or participants. What matters is whether you can say why it should be specific, before checking whether it is.

How many instruments should I test?

Enough uncorrelated ones that a coincidence would have to repeat itself. Two markets driven by genuinely different things are worth more than six that move together.

Is it cheating to pick the instrument that works best?

It is only a problem if you then report that result as though the instrument had been chosen in advance. Selecting is a legitimate part of research; the error is forgetting how many you rejected when judging what you kept.

Sonar Sciences is a research and validation platform, not an investment adviser. Nothing on this page is investment advice or a recommendation to trade any instrument or strategy. Platform names are trademarks of their respective owners and we are not affiliated with them. Trading leveraged instruments involves substantial risk of loss.