Sample size

A win rate of 70% sounds strong. Over ten trades it means seven went your way, which is roughly what you’d get from flipping a coin ten times and having a mildly lucky afternoon. Over a thousand trades it would be remarkable.

Same percentage, completely different claim. The difference is the sample size, and it’s why BotStock prints it next to figures that are being used as evidence — “±2.4% over 4 reports”, “n=120”.

The rough shape of it: a small sample doesn’t just mean “less certain”, it means the number you’re looking at could easily be several points away from the truth in either direction. Thirty observations is usually where a figure starts being worth printing at all; a few hundred is where it starts being worth acting on. Below that, what you have is a hint about what to measure next.

There’s no threshold where a number becomes true. There’s only a point where it stops being mostly noise.