A $1m competition Zillow is running to improve its home valuation tool – Zestimates – is tracking to become one of the most high-profile machine-learning competitions ever hosted, according to the Seattle-based U.S. real estate marketplace.
The competition launched in May 2017 and delivered a callout for data scientists, engineers and visionaries to compete in a bid to improve automated home valuations of 110 million homes across the U.S. offering a $1m prize.
Zillow Prize, designed to inspire the brightest scientific minds to compete for a $1 million grand prize, has already attracted more than 15,500 individuals who have downloaded the competition dataset. Since the contest launch on May 24, more than 2,500 competitors from 76 countries around the world have submitted an average of 350 entries a day into the contest.
The contest will be administered by Kaggle, a platform designed to connect data scientists with complex machine learning problems and is staggered into two rounds; the public qualifying round which is now open and concludes Jan. 17, 2018i and a private final round that kicks off Feb. 1, 2018 and ends Jan. 15, 2019ii.
“From the onset, we expected that Zillow Prize would be hugely popular. It’s a chance for our community to impact the Zestimate, an algorithm regarded by many as one of most high-profile examples of machine learning,” says Anthony Goldbloom, Kaggle CEO.
“Since the contest launch, we’ve been impressed by the response from Kaggle’s scientific community. As the qualifying round deadline nears and teams compete to earn their invitation into the final round, we expect the number of competing teams and daily contest entries to increase significantly – and Zillow Prize is on pace to become one of the most popular contests Kaggle has ever hosted.”
Zillow releases new version of Zestimate
While Zillow Prize competitors around the world are working to improve the Zestimate accuracy, the Zillow team is also making new gains in home valuation accuracy. Zillow has released a major update and most accurate version of the Zestimate, bringing the algorithm’s accuracy to 4.3 percent nationwide, down from 5 percent. With valuations on nearly 100 million homes across the U.S., ensuring the Zestimate is as accurate as possible is a top priority for Zillow as it helps homeowners understand the value of what is likely their largest asset, their home.
To establish these new gains in home valuation accuracy, Zillow transitioned all its data to the cloud and can now compute the Zestimate in near-real time. Now, Zillow can process three times as much data as before, which allows its data scientists to experiment and iterate faster than ever, creating more accurate valuations.
“We are excited by the global response we’ve gotten to date for Zillow Prize,” said Dr. Stan Humphries, Zillow Group chief analytics officer.
“The Zillow Prize competitors have responded with the same passion we see everyday within our internal team of data scientists and machine learning engineers – we never rest in striving to make the algorithm more precise. The Zestimate is trying to answer an incredibly complex and important question, and with the strong contest submissions we’re already seeing, we are on pace to reach our goal of becoming one of the world’s most impactful machine learning competitions. In the meantime, we think homeowners will be pleased with the new enhancements we’ve made to ensure they have a trusted starting point when monitoring the value of what is often the largest purchase of their lifetime.”
Zillow Prize participants have until Oct. 16, 2017 to register for the qualifying round, download and explore the competition data seti, and submit their own Zestimate residual error modelii. The top 100 teams from the qualifying round will earn prize money and be invited to participate in the final round, with a chance to walk away with the $1-million-dollar prize.
The company recently won the dismissal of a federal lawsuit in Chicago challenging the accuracy of the online real estate website’s ‘Zestimate’ tool for estimating U.S. home values.