Image: Jason Hiner/TechRepublic
At the 143rd Kentucky Derby on May 6, viewers of America’s most famous horse race will have a new—and, arguably, smarter—strategy for betting. Through a partnership with San Francisco startup Unanimous A.I., Churchill Downs and Twin Spires will be offering an option to bet based on predictions made through an AI-platform.
The swarm included a group of handicappers and racing experts from across the US—from Louisiana to Nevada to New York—and took place at 1:00 pm Eastern Time on Wednesday. TechRepublic will publish the results, with some analysis, on Thursday.
How was this unlikely partnership born? In 2016, I met the founder of Unanimous A.I., Louis Rosenberg, at an AI conference. Before the Kentucky Derby last year, I asked Unanimous A.I. if they could use the platform to predict the Derby. Despite the unpredictability of horse races, the company took on the challenge—and TechRepublic published the results the day before the race. The swarm nailed the superfecta—top four horses in order—beating 540-1 odds. (I had placed a $1 bet on the race, returning a $541.10 ticket).
After the major victory of the swarm in 2016, Churchill Downs took notice. “I had read a few articles about the superfecta picks last year after the fact, and thought it was kind of funny how the expert handicappers would point out that it wasn’t a particularly interesting pick of four horses,” said Drew Zipp, UX lead at Churchill Downs. “Which may be true, but it’s hard to argue with success.”
“We were approached by Churchill Downs,” said Rosenberg, CEO of Unanimous A.I. “and they had an interesting idea: What if we built a swarm intelligence of expert handicappers?”
Zipp spearheaded the partnership with Unanimous A.I. to see what might happen if a group of experts at Churchill Downs could form an even smarter swarm in 2017. The result? A group of handicappers met on Wednesday to come up with a “super-expert” prediction for the Derby.
Handicappers consider the different factors that influence horse racing, such as race histories, speed ratings, posting positions, track conditions, and horse breeds. As a result, Rosenberg said handicappers are “less likely to have unjustified overconfidence in their picks.” The group will narrow down the a list of 20 horses to the few top picks, and from that, will predict the top four.
Unanimous A.I.’s software called UNU is based on the concept of swarms in nature. The online platform created a space for people from all over the world to gather and make a real-time decisions. The collective decision stands apart from tools like polling or crowdsourcing that form averages—instead, members of the swarm can influence each other in real time. (Here’s more on how swarms work.)
Through the support of TwinSpires.com, the betting arm of Churchill Downs, the swarm results will form the basis for a $10,000 Player’s Pool that gamblers can bet on, with one click, starting Thursday, for as little as $10. Think of a Player’s Pool like chipping in on a big bet with your friends. The betting strategy comes from Ed DeRosa, the top handicapper at Churchill Downs, and a member of the swarm.
SEE: How to Implement AI and machine learning (a ZDNet/TechRepublic special report)
It’s important, said Rosenberg, to temper expectations for what the “super-expert” swarm is capable of—in other words, hitting the superfecta two years in a row may be unlikely.
“It’s less about trying to replicate results,” said Rosenberg, “than to ask: What happens when you tap into experts?”
The use of AI in racing illustrates how AI is turning up in spots that were never expected.
“When you think of the Derby, you think of this old institution,” said Joe Rosenbaum, communications manager at Unanimous A.I. “But they wanted to use AI with what they’re doing. They’re very concerned with looking at what the future of the Derby can be.”
Can the swarm out-predict the experts again in 2017? Check back on TechRepublic.com on Thursday, when we’ll publish the expert swarm picks. And the final answer will be revealed around 6:30 pm ET on Saturday.