Base rates: ask "out of how many chances?"

One habit separates counted evidence from a good story: ask "out of how many chances?"

A count needs its denominator and a relevant comparison.

A scary number, on its own

In this constructed example, imagine a post that says: “this pattern appeared right before 89 big drops.” It sounds like a discovery. It is half a sentence, because it never says how many chances the pattern had to appear at all.

The same fact, completed

Now suppose you add the missing number: the pattern appeared 89 times, out of 383,490 traded minutes in the same stretch. Suddenly you can ask the only question that matters: does a big drop follow those 89 minutes more often than it follows any ordinary minute? That ordinary-minute frequency is the base rate, and it is the yardstick every claim has to beat. If drops follow the pattern about as often as they follow any random minute, the pattern told you nothing, however dramatic the 89 stories sound.

Survivorship: counting only the winners

Most trading content is built the other way around. Someone finds the times the pattern worked, posts those charts, and quietly never counts the times it appeared and nothing happened. That is survivorship, and it can make pure noise look like an edge.

EdgeDepth counts the other way: every match, measured forward from the moment the condition was true, with no filter on how the story ended. The boring matches count exactly as much as the spectacular ones.

A coin flip is a finding

When a scan comes back “about half went up, half went down”, that is not a failure of the tool. That split alone does not establish a directional difference. Compare outcome sizes, uncertainty and the baseline before judging whether the condition adds information. Knowing which distributions are coin flips is most of the value of keeping an honest record.

We count every match, measured forward, including the ones that went nowhere. If the honest answer is a coin flip, we print the coin flip.

See it live

Run a scan and read the verdict sentence, count included.

Or read a finished one. Every published study prints its matches, the denominator it scanned and the unconditional rate over the same markets and window, so the comparison this page describes is already on the page.