HOUSE PRICER

Christchurch City, New Zealand

Please note: only Christchurch city houses are supported at this time.
Model Accuracy
A key aim of House Pricer is to provide price estimates that are accurate, yet transparent. This means making it clear to users how the ML model powering these predictions is performing, even when it's performing badly! This page presents metrics that show how the model has performed on recent data. When training the model on historical sales, a certain portion of the historical sale data is kept separate until after the training is complete. As the model has not "seen" this data yet, it's a good way to gauge its accuracy. The predictions made by this "unseen" data are presented in several forms below.
The median accuracy (the middle accuracy value when the results are sorted) is the most reliable figure as it's easy for a bad prediction to throw the average off. For example, say the model makes a prediction on a property with a house that has recently been demolished. It may think based on the not yet updated council data that the house and land is worth $900k, however the bare land only actually sold for $300k. This would show up as a poor prediction by the model and skew the average.
That being said, the test results here are only a rough indication of the model's performance, and as always, feedback on poor predictions is welcome!
Model Performance Figures
  • The average prediction difference is 6.55%.
  • The median prediction difference is 5.16%.
  • 95.00% predictions are within 20%.
  • 76.00% predictions are within 10%.
  • Tested using data from 07/05/2025 until 06/06/2025.
    Model Predictions on Recent Sales
    Note: Difference is measured between the medium predicted price estimate and the actual sale price.
    Street Address
    Predicted Price Low
    Predicted Price Med
    Predicted Price High
    Sale Price
    Sale Date
    Difference
    23 Stillwater Avenue, Burwood$747,600$765,600$783,600$765,0005/06/2025
    0.08%
    54c Aorangi Road, Bryndwr$825,700$832,650$839,600$830,0008/05/2025
    0.32%
    96a Brookside Terrace, Bryndwr$1,077,900$1,107,250$1,136,600$1,111,00023/05/2025
    0.34%
    18a Hay Street, Bromley$509,700$530,850$552,000$533,00019/05/2025
    0.40%
    1 Malachy Grove, Aidanfield$970,200$976,500$982,800$981,00021/05/2025
    0.46%
    1/4 Ruddenklau Lane, Bishopdale$515,600$517,000$518,400$514,0007/05/2025
    0.58%
    24 Teesdale Street, Burnside$708,400$710,450$712,500$715,00016/05/2025
    0.64%
    10 Eastling Street, Bishopdale$666,800$691,500$716,200$696,00021/05/2025
    0.65%
    2/106 Bealey Avenue, Christchurch Central$474,900$480,350$485,800$477,00023/05/2025
    0.70%
    81 Cotswold Avenue, Bishopdale$657,300$678,300$699,300$672,00023/05/2025
    0.94%
    56 Colombo Street, Cashmere$911,900$969,650$1,027,400$960,00022/05/2025
    1.01%
    20 Robert Duncan Road, Belfast$715,700$722,600$729,500$730,0009/05/2025
    1.01%
    40 Damien Place, Bromley$524,100$527,500$530,900$522,0008/05/2025
    1.05%
    2 Reginald Street, Burwood$470,500$485,700$500,900$491,0008/05/2025
    1.08%
    30 Borderdale Street, Belfast$743,900$745,700$747,500$737,00012/05/2025
    1.18%
    54 Colwyn Street, Bryndwr$780,600$797,000$813,400$787,0003/06/2025
    1.27%
    5 Armitage Street, Bishopdale$659,100$674,800$690,500$684,00030/05/2025
    1.35%
    16 Augustine Drive, Aidanfield$1,200,200$1,243,000$1,285,800$1,260,0008/05/2025
    1.35%
    2/388a Pages Road, Aranui$474,200$490,000$505,800$497,0004/06/2025
    1.41%
    35 Collies Road, Casebrook$894,300$906,700$919,100$920,0009/05/2025
    1.45%
    Frequently Asked Questions
    How up to date is this?
    The model is trained weekly on a Saturday using data up to a month prior to the training date. The model is then validated on the most recent month of data, providing the performance metrics in the section above. This does mean the model is behind by... Show more
    So can I really trust these predictions?
    The short answer is no. The model is only as good as the data it is trained on, and there are many factors that can affect the price of a property that the model doesn't take into account. For example, the model doesn't know if the property is in a g... Show more
    I searched for a property and the predicted price was way off. Why?
    The model is very much a work in progress and there will certainly be cases where it is completely wrong. There could still be problems with how the data is being prepared for the model. If you find a property with a predicted price that you think is... Show more
    The average difference is high, but the median difference is low. Why?
    The average difference is the average of all the differences between the predicted price and the actual price. The median difference is the middle value of all the differences. If there are properties for which the model is completely wrong (say 50% ... Show more
    How does the model work?
    The model takes in a number of features about a property and outputs a predicted price. The features are things like the number of bedrooms, the property's coordinates, the size of the property, the year it was built, etc. The model is trained on a d... Show more
    What do the model performance metrics mean?
    After the model is trained on data up to a month prior to the training date, it's tested on the remaining month's worth of data. The model performance metrics are then calculated by comparing the predicted price to the actual price for each property ... Show more
    I can't find my house or a house I'm interested in seeing the price for?
    At this point in time, only Christchurch city houses are supported. This means no Selwyn or Waimakariri houses because they are separate to Christchurch city council and have their own council data. If you're searching for a house and it's in Christc... Show more