Hacker News Reader

Comment by jrochkind1 | Hacker News Reader

I don't think they lost $420 million on it because they knew what they were doing and were good at forecasting home prices and know things others don't about them. I mean, it's an interesting story you tell, but doesn't seem like the most obvious explanation to fit the data -- which would be, actually, they were bad at forecasting home prices, is why they lost $420 million trying to predict home prices.

View on HN

As someone who's been on the buyer side of the market for the past few months, I think it's even worse for their objective: they might be good at forecasting home prices on average conditional on their statistics, but bad at forecasting home prices of the subset that would be willing to sell to Zillow. Every zillow-owned home I've seen, bar none, has had either a bad micro-location (eg: next to a railroad in an otherwise good location), or a bad interior (eg: a smoker home). The same was true for Opendoor, who are in the same business. IE conditional on Sq Footage, location, acreage, etc. they might be incredibly accurate, but they're only able to buy the 50% below that average (and the gap is large because the missing/intangible variables matter a lot).

I agree that OpenDoor and Zillow bought really weird houses. My neighborhood is completely brick houses except for one that has some plastic looking siding and was in slight disrepair, guess which one Zillow bought and it sat for 6 months? There was another house OpenDoor bought down the street and at one point the roof was caving in, I dove by it for a decade wondering who lived there. A house flipper came along and repaired the cosmetics then sold it to OpenDoor for a premium. It sat for a year. There were also other weird houses like a fake barn and one’s, as you said, with bad micro-locations. Every single one within 5 miles of my house was a house with something weird.

The brilliant Zillow AI got fooled by old fashioned American DIY home improvement catastrophes.

This reminds me of this: https://en.wikipedia.org/wiki/The_Market_for_Lemons

The Indicator from Planet Money podcast just did an episode on Zillow with the same thesis: https://www.npr.org/2021/11/08/1053689886/ibuyers-zillow-and...

That's a really good point. The people most likely to sell to zillow, after all, are going to be the people who think zillow's price is way more than it should be!

En masse buying of buildings based on algorithmic pricing may be doomed because of this. Even if you're on the mark 80% of the time, under-estimate 10% and over-estimate 10% -- the 10% where you over-estimated will be the people most eager to sell to you!

Which is true for anyone in the property business, but if they didn't have local people who understood the market actually looking at the individual property and being like "uh, that one is next to a railroad track which the algorithm doesn't seem to be accounting for..." those people realized they hit the jackpot when zillow offered them a price as if they were NOT next to a railroad track!

Reply on news.ycombinator.com