How real is REALas property price prediction?

31 Oct 2017

How REAL is REALas?

I have subscribed to REALas, and so far, the majority of the predictions I have compared with actual sales prices; have been out.

In some cases, out and by as much as 9-12 percent. For an agent, this may constitute as a reprimandable offence. Our governing bodies would grill us, if our files reflected these figures.

Was the birth of these property price-predictions sites another by product of an upward property trend? Or is there a real science, behind them and will they last?

Nevertheless, how real, the REALas predictions, will continue to be in a FALLING property market, will only be determined in time. However, I’m not holding my breath.

I say, and like other DATA we have analysed with Anthony Carroll and George Rousos. We will continue to see, a separation of REAL sales figures with PREDICTED sales figures.

The data which many price-predictive platforms use, we feel is old and outdated. When I say outdated, I mean, in a rapidly changing market; three weeks is old. Some data can be up to three – six month old, and before it’s shared with the general public.

Nevertheless, the consumers (vendors), are relying more and more on these platforms. An appetite and dependence that at face value, distorts the property market.

Below is a screen shot of a prediction and an actual sale price.
REALas prediction: $1,065,000
Actual sale price: $936,000


Reliability of Data

David Crombie, CEO of the Estate Agents Co-Operative highlights and concurs with my old-data premise. “The issues with the algorithms used by these sites are twofold. On the one hand, they capture outdated figures and therefore are inaccurate in real time, and they are to slow, in keeping up with a rapidly falling market…”

As I said before, three week old figures, can be outdated and old in this arena. Not to mention, rapid changes within the property market.

Industry Concerns

When a distortion is created in the pricing of a property; no one comes out the winner.

Vendors that become overly reliant, on predictions of what their properties are worth: sit on the market with false expectations and for a very long time.

Agents, on the other hand can and often, lose the buyers that are offering fair-market-price. Buyers that may attend, early in the marketing campaigns.

Further Industry Questions

  • Are these predictions, a form of over-quoting?
  • Who is governing these websites?
  • Should NSW Fair Trading and all other Fair Trading bodies, begin to question the data. The impact on consumers and to an extent; misrepresentation?
  • How can a virtual, valuation/prediction be accurate or close to real, if a formal and internal physical inspection wasn’t conducted?

I raised the above questions with David Crombie. The unfortunate thing is; none of these sites fall under any Property, Stock and Business Agents Act. Nonetheless, both consumers and the industry appear to be led down a dangerous path, by outside powers and influence.

Industry Call and Solution

More so now, than ever. I will urge potential property-sellers, to obtain no less than three Comparative Market Appraisals, before relying too much on price-predictive sites.

A well versed real estate agent, has grass-root knowledge of the market place and LIVE – DATA to back themselves with, and provide more of an accurate selling price.

My caveat on what agents provide is: one thing is what some agents tell vendors and place on their Agency Agreements, and this may differ from what they share with the general public.

Edwin Almeida
Licensed Real Estate Agent
Raine & Horne Blacktown


  • getREALas says:

    We are transparent about our performance. The REALas accuracy is here: Furthermore REALas predictions and sold results are available on our site.

  • George says:

    Edwin, thanks for your insights. Your concerns are well founded. Lenders are greatly increasing their use of Automated Valuation Models to confirm property value and to streamline the lending process. For risk management, lenders must understand the statistical performance of the AVM they use. That also means understanding Forecast Standard Deviation (an accurate forecast of the price range). I have found no AVM’s on the market in Australia that publish an accurate FSD, indeed many are fixed! Moreover, many do not tie their confidence scores to FSD, instead basing scores on unreliable factors such as number of local properties used in the model’s estimate, neighborhood range of values, or some other measure that does not correspond precisely to expected AVM performance.

    As for the above-mentioned company, my understanding is that this is not an proper AVM because it only attempts to predict the price of properties on the market and moreover with an advertised range, not properties without a real estate agent’s price prediction, on or off the market. This is akin a walking up to a target with a bow and arrow and hitting a bulls eye at close range! I don’t think there is much science involved in that? And if you can’t even get that right, well…

    For there to be scrutiny and proper oversight o the AVM’s on the market, as suggested, I believe that a governing body of sorts should insist that all AVM’s on the market publish a Forecast Standard Deviation that is unique to each property. Furthermore, Lenders should verify that a reported model standard deviation is truly an FSD and that the reported FSDs match standard deviations in their tests — to insist FSDs are reported accurately by the model. This includes both ‘precision’ and ‘accuracy’ metrics, two different measures.

    The best way to validate confidence score or FSD is by an independent body (not by a sponsor company) performing ‘out-of-sample test’ of property valuations. Then all AVM platforms can be verified against selected benchmarks.The tests I have seen on-line are not statistically valid, and many appear to be cherry picking…

    I hope that helps

    Regards George

Leave a Reply