Predictions
Forecasts made from historical analogs before the outcome was known, so accuracy can be judged honestly against what actually happened.
Each forecast asks an empirical question: when the signal profile looked like this before, what did the stock actually do next? A forecast backed by 12 analogs is not the same confidence as one backed by 200; the analog count is shown everywhere for that reason. Fewer than 12 similar setups and the model declines to forecast rather than guess.