HtAG predictions

Discussion in 'Property Market Economics' started by Terry2020, 18th Oct, 2018.

Join Australia's most dynamic and respected property investment community
  1. Terry2020

    Terry2020 Well-Known Member

    Joined:
    18th Oct, 2018
    Posts:
    47
    Location:
    Sydney
    Edit: Split off from the Introductions forum, as this is information more generally useful as a topic in its own right.
    The quote is from @jprops, but it's been removed to ensure that @HtAG is shown with the originating post.

    Hi jprops, there are multiple markets in the Greater Sydney region with their own localised property cycles. If you can narrow your question down to an area, then we can try and answer your question.
     
    Last edited by a moderator: 22nd Oct, 2018
  2. jprops

    jprops Well-Known Member

    Joined:
    24th Sep, 2015
    Posts:
    885
    Location:
    Sydney
    Let's try Hornsby 2077
     
  3. Terry2020

    Terry2020 Well-Known Member

    Joined:
    18th Oct, 2018
    Posts:
    47
    Location:
    Sydney
    Recovery by the end of 2019. See chart below for houses. Units follow a similar pattern in this area. Capture.JPG
     
    Last edited by a moderator: 19th Oct, 2018
  4. datto

    datto Well-Known Member

    Joined:
    23rd Jun, 2015
    Posts:
    6,675
    Location:
    Mt Druuiitt
    Mt Druitt? Or just type in God's country.
     
    TheSackedWiggle and Dean Collins like this.
  5. Terry2020

    Terry2020 Well-Known Member

    Joined:
    18th Oct, 2018
    Posts:
    47
    Location:
    Sydney
    3% growth next year in Mount Druitt.
     
    datto likes this.
  6. Terry2020

    Terry2020 Well-Known Member

    Joined:
    18th Oct, 2018
    Posts:
    47
    Location:
    Sydney
    Jprops, Datto you both asked questions relating to specific areas. What are your reasons for interest in Sydney(Hornsby) and Mount Druitt?
     
  7. datto

    datto Well-Known Member

    Joined:
    23rd Jun, 2015
    Posts:
    6,675
    Location:
    Mt Druuiitt
    Better than bank interest!
     
  8. datto

    datto Well-Known Member

    Joined:
    23rd Jun, 2015
    Posts:
    6,675
    Location:
    Mt Druuiitt
    Well, I hold property in the area. There are views out there that house prices will go sub 400K and as a consequence my sleeping patterns have not been quite the same. But 3% growth can make me sleep a little easier.
     
  9. Foxdan

    Foxdan Well-Known Member

    Joined:
    22nd Jun, 2015
    Posts:
    457
    Location:
    Hills district, sydney
    @HtAG what factors / assumptions have you built in to make these predictions?
    Given APRA restrictions are a “new” factor, I would like to know whether future borrowing capacity is built into your predictions?
    Do you factor the amount of new developments be their completion dates into your calculations?
    What about predictions of future interest rates and it’s affect on prices?
     
  10. Terry2020

    Terry2020 Well-Known Member

    Joined:
    18th Oct, 2018
    Posts:
    47
    Location:
    Sydney
    Foxdan, At HtAG, we divorce the analysis of causality between fundamental market variables such as population, unemployment, interest rates, new developments and property price changes. By recording real time movements in the property market, all of these causes that impact on fluctuations in property prices have already been accounted for within the collected Big Data that serves as an input into the forecast algorithm.
     
  11. Morgs

    Morgs Well-Known Member Business Member

    Joined:
    7th Dec, 2017
    Posts:
    1,791
    Location:
    Sydney NSW
    How does it take into account the one off change in credit availability?
     
  12. Terry2020

    Terry2020 Well-Known Member

    Joined:
    18th Oct, 2018
    Posts:
    47
    Location:
    Sydney
    Thanks Datto. Forecasted inflation rate for 2019 is 2.38%. So in terms of real value, 3% growth is 0.62%. We are seeing multiple localised pockets in the markets across the country that are showing up to 5% growth in 2019. And others (many more) a decline of up to -5%. Blacktown LGA is in the "green" for now.
     
    datto likes this.
  13. Terry2020

    Terry2020 Well-Known Member

    Joined:
    18th Oct, 2018
    Posts:
    47
    Location:
    Sydney
    Morgs, see reply to Foxdan above.
     
  14. Terry2020

    Terry2020 Well-Known Member

    Joined:
    18th Oct, 2018
    Posts:
    47
    Location:
    Sydney
    P.S. This is for the 'house' segment. 'Unit' segment is a different story.
     
  15. jprops

    jprops Well-Known Member

    Joined:
    24th Sep, 2015
    Posts:
    885
    Location:
    Sydney
    Your answer is very vague/generic and doesn't speak to the specific question. It also seems to be copy from your marketing website. If you are using machine learning then I can only assume that you are including "population, unemployment, interest rates, new developments and property price changes" as features into the trained model. The question is, how are you factoring in other data that is relevant such as one off changes. For instance, none of the data you mentioned include information on credit availability. Your predictions also are unlikely to factor in the PI2IO cliff that is expected to occur 2020/2019. Or does it?
     
  16. Terry2020

    Terry2020 Well-Known Member

    Joined:
    18th Oct, 2018
    Posts:
    47
    Location:
    Sydney
    The answer was a tailored version of the statement on our website, as we do get asked this question a lot.

    I will try to answer it differently now. Population, unemployment, interest rates, new developments etc. are NOT factored in to the model as independent variables. We use a neural network that learns off a combination of historical and recent (days and weeks) sales data to forecast future prices. One off changes you describe are "picked up" by the algorithm in the more recent sales data (i.e. marginal increases or decreases in sale price as a reaction to news and announcements) and serve as signals to adjust forecasts.

    Acknowledge that there is a lag before adjustments kick in as it takes time for prices to react to new market realities, but because we collect sales data on a daily basis the delay is in the days/weeks as opposed to months and quarters.

    Trisha

    P.S. What is the PI2IO cliff you made mention of ?
     
  17. qak

    qak Well-Known Member

    Joined:
    1st Jun, 2017
    Posts:
    1,671
    Location:
    Sydney
    I'll bite - Kellyville NSW (given existing number of properties for sale)
    Paddington NSW - self interested
     
  18. Rex

    Rex Well-Known Member

    Joined:
    12th Feb, 2018
    Posts:
    1,010
    Location:
    Perth
    I'll bite too. So basically the only inputs to the model are sales data - does this include metrics like listing times, days on market, etc? I'm guessing the model has been validated through backtestin. If so what sort of accuracy do you observe?
    And finally, when will Perth turn good? I'm dying to know...
     
  19. Foxdan

    Foxdan Well-Known Member

    Joined:
    22nd Jun, 2015
    Posts:
    457
    Location:
    Hills district, sydney
    So basically you have built a “momentum” predictor and ur using historical trends to predict the future trends.
    Given credit restrictions have changed the fundamentals of property prices in Australia, your AI needs to learn that.

    Happy for you to list the predictions of 10 major suburbs and we can all revisit this in a year or two and decide if you have any substance.

    In the meantime, I think due diligence on an area using all the standard fundamentals (supply, employment rates, interest rates, immigration to an area etc etc) and a bit of speculation will trump ur AI.

    As for Hornsby and my druitt returning to 3% growth next year... doubt it.
     
  20. jprops

    jprops Well-Known Member

    Joined:
    24th Sep, 2015
    Posts:
    885
    Location:
    Sydney
    How are you sourcing accurate sales data that is not 3 month lag?

    PI2IO(sic), was meant to be IO2PI.. is when all the 5 year IO periods expire and no one can refinance due to borrowing power. IO loans peaked a few years ago, and so the peak of expiring IO loans are due to hit 19/20 Where they will be converted to PI loans with 25 years.

    What kind of back testing have you done and over what periods?