Every signal is a feature vector — RSI, MACD, volume Z-score, momentum, etc. The brain computes Euclidean distance between the current signal's feature vector and every historical labeled trade.

The 10 closest matches tell you: "When I've seen something this similar in the past, what happened?"

If 8 of 10 most-similar trades were wins, the current signal is in a high-quality cluster. If 8 of 10 were losses, even if the model says 60% confident, you should be skeptical — your own history disagrees.