Ziprealty can kiss my price predictor

I've written before about the heinous algorithm used in Ziprealty's "Predict It" price "game." It still sucks.

There is a neighborhood that I track. I know this neighborhood pretty well. A house came on the market, bank-owned REO. Before it hit MLS, we spoke with the bank's agent about buying the house - but their price was crack-smoking high. When the house hit MLS, I "predicted" the selling price on ziprealty. The house has languished on the market, unsold at its unrealistically-high asking price. The bank finally lowered the price, but the price is still unrealistic. Another house on the same street came on the market last week, a smaller house on a larger lot, but in much better condition. The second house is priced realistically - although not low enough to drive much traffic, let alone a bidding war.

The listing agent decided to relist the property when he lowered the price. So although the asking price is now within spitting distance of my price prediction, my "Property IQ" for the property is calculated by comparing it to the asking price in the original listing. That's a bogus programming choice, and one designed to support the asking (wishing) price as the realistic value - in a market where the State budget crisis, continuing economic uncertainty, and the all-important employment rate are all dragging home prices down. The major local employers have announced cutbacks - or investment anywhere other than California - and the State of California keeps threatening to layoff State workers. These are not conditions that bode well for rising home prices anytime soon, so why design a price algorithm that is heavily weighted towards asking price, unless the intent is to subtly manipulate buyer sentiment towards the asking price in the complete absence of any validation of asking price?

Ziprealty has an "Offer Evaluator" tool that compares the offer price you enter to recently sold properties in the same area. I simply enter $1 as the offer price, and the Offer Evaluator tells me that $1 is unlikely to be accepted because most comparable properties sold within ___% of asking price. But that's the key piece of information - it tells you what percent of asking price (typically 95%-105% of asking) most similar properties sold for. And yet, the Price Predictor skews towards asking price, even when the Offer Evaluator shows that most similar properties sold well under asking price.

Another flaw in the price predictor is that it basically ignores user input. On several properties I've "played," all players entered prices well below asking price, yet the algorithm claims that the Community prediction is merely hundreds to a couple thousand below asking price. Which begs the question - how many users does it take to make the "Community Prediction" actually match the predictions entered by the Community? A thousand? A million? 4.6 Trillion?

If this was meant to be a useful tool, it would have been designed without so many major flaws in its algorithm. If it was meant to be a subtle marketing manipulation, well, it's perfect. Bankers weren't the only sleazebags that created the bubble, they're just the only ones who've had to stop being so overtly sleazy.