← Back to all models

AVM model details: San Francisco, California, United States

Performance metrics for AVMs in this location.

Model type

Select the type of valuation model you want to view.

List price estimation model

Estimates list prices in this location from historical data and market trends.

Key performance metrics

Median Absolute Percentage Error (M. MAPE)11.3%Half of our predictions are within this percentage of the actual value
Testing R² Score0.850How much of the price variation our model explains
Root Mean Square Error$831,400Average prediction error in the local currency
Mean Absolute Error$342,427Average absolute difference between predicted and actual

Accuracy distribution

Percentage of predictions within various error ranges

Within 5%23.7%
Within 10%45.4%
Within 15%61.5%
Within 20%73.5%

Prediction bias analysis

Does the model tend to overestimate or underestimate?

Mean Bias-$81,888Average prediction bias (+ means overestimate)
Median Bias-$3,219Typical prediction bias
Over-prediction Rate49.0%Percentage of predictions that are above the actual value
Under-prediction Rate51.0%Percentage of predictions that are below actual value

Error distribution percentiles

The absolute error amount at different percentiles - for example, '90th percentile: $50,000' means 90% of predictions have errors less than $50,000

Percentile10th25th50th75th90th95th
Error Amount$22,975$61,620$145,008$333,500$732,905$1,146,369

Model configuration & training

Training Samples9818
Testing Samples2151
Training R²0.970
Testing R²0.850
Last Updated2026-10-12

Additional performance metrics

Mean Absolute Percentage Error16.9%Average percentage error across all predictions
Forecast Standard Deviation$827,358Spread of prediction errors
Coefficient of Variation48.8%Relative variability of predictions
10th Percentile Error %2.0%10% of predictions have error less than this
90th Percentile Error %35.8%90% of predictions have error less than this