How long before you'd know?
Someone claims a strategy works. Before you give it a month of your life, or your money, there is a question with an arithmetic answer: how many trades would it take to tell that claim apart from luck?
Usually far more than anyone has. And most claims never reach the question, because the edge is smaller than the cost of trading it, which is a loss rather than a small win.
About 3,974 events, which is 13 months at this rate. Testable, but know that going in: a shorter run cannot tell you whether this works.
Why two numbers. The smaller one is when the result would just cross the line, assuming the claim is exactly true. The larger one is what you actually need to be confident, because an estimate from a short run is usually flattering. Designing a test to the smaller number is how people end up running something that was never capable of answering its own question. We use the larger one.
What this is not.It takes the claim's own numbers at face value and tells you what would follow if they were true. It cannot tell you whether they are. It also assumes every trade is independent, which real strategies frequently are not, and that makes these figures optimistic rather than cautious. If a claim fails here, it fails on its own arithmetic before anyone argues about the data. Every strategy we ran through this is on the ledger.
Where this came from
We spent five weeks collecting data on a signal that looked promising, then worked out it would take about 91,900 trades to know whether it was real. At the rate those trades actually appeared, that is roughly nine years. Not “collect more data”, then: a question that cannot be settled, which is a different and much more useful answer.
Doing that arithmetic first would have saved the five weeks. That is the whole reason this page exists, and the numbers from that case are one of the presets above.
One honest caveat about the maths. It assumes each trade is independent of the last. Real strategies often are not: they win together and lose together, which means the true number of trades needed is higher than this shows, sometimes much higher. Treat every figure here as the best case.