How is the AI score calculated

Side Street displays two separate but related scores for each property: a prediction confidence score and a neighbourhood reliability tier. Together they form a final rating (shown as low, medium, or high) that tells you how much to trust a given prediction.

Prediction confidence is derived from the model's own uncertainty — specifically how wide or narrow the 95% confidence interval is around the predicted price. A narrow interval means the model is internally consistent and the prediction is tight. A wide interval means there's more spread in the possible outcome.

Neighbourhood reliability is a separate, historically-grounded score. It comes from back-testing the model against actual prices in each neighbourhood — measuring how well predictions matched reality in the recent past. A neighbourhood with a lot of clean historical data and consistent pricing behavior will score higher than one with sparse or erratic history.

The final score combines both: if the model is confident but the neighbourhood has historically been unreliable, the score is moderated downward. If the neighbourhood has a strong track record and the model is also confident, the score is elevated. This means a "low" score doesn't necessarily mean the prediction is wrong — it means you should treat it with more caution than a "high" rated one.