Automated Valuation Model (AVM) info

Select the type of valuation model you want to view.

How our models work

When estimating a property value, we use the most specific model available. If a city-specific model exists, it will be used. If not, we fall back to the provincial model. If no provincial model is available, the country-wide model provides the estimate. This hierarchical approach ensures you always get a prediction while maximizing accuracy for areas with sufficient data.

Understanding the metrics

Median MAPE (Median of the Mean Absolute Percentage Error):

This tells us how accurate our predictions typically are. If the Median MAPE is 5%, it means that half of our estimates are within 5% of the actual value. So for a $500,000 home, we'd typically be off by $25,000 or less in either direction.

This metric focuses on the 'typical' error rather than the average, which gives a more realistic picture of accuracy since it's not skewed by outliers. Lower numbers mean more accurate predictions.

R² Score:

This measures how much of the variation in property prices our model can explain. An R² of 0.850 means our model accounts for 85% of the differences in property values.

To put this in context: imagine trying to predict home prices using only the city name (maybe R² = 0.30), versus using a few features like square footage, bedrooms, and neighborhood (maybe R² = 0.65). The higher R² shows our model uses relevant information effectively.

Within 10%:

This shows how often our estimates are close to the actual value. If this number is 75%, it means that 3 out of 4 times, when we estimate a property is worth $500,000, the actual value will be between $450,000 and $550,000. The higher this percentage, the more reliable our estimates are.

Details:

Each model has a details page with more in-depth metrics to help you understand its performance better.