Swarm Intelligence Delivers 540:1 Kentucky Derby Superfecta

Yes that’s right. 540 to 1 odds. $1 would pay you $540, a $100 bet would have seen you filling IRS paperwork at the track to take home your $54,000 and $1,000 bet on the computer’s call would have gotten you more than half way to millionaire status for spending a sunny Saturday afternoon watching some of the Earth’s most majestic creatures turn left in a manner considerably more exciting than NASCAR.

Swarm Intelligence Delivers 540:1 Kentucky Derby Superfecta

Swarm intelligence picks the Superfecta in Kentucky Derby

If unaware, picking the first and second place horses in a race in correct order is an exacta. 1-2-3 will allow you to claim the trifecta, but getting 1-2-3-4 (again, in correct order) is the ever elusive and big payout…SUPERFECTA and an AI program just pulled it off last weekend.

Before talking about this accomplishment by a software program called UNU, let’s be clear this is one of the lowest paying superfecta payouts in the history of the Kentucky Derby. For example in 2005, a $1 bet would have seen you win nearly $865,000, which is a bit more than $540. The Superfecta is paid out based on the multiplication of the odds of each horse that finishes 1-2-3-4. Clearly, this year’s superfecta featured a number of the favorites in the field.

The software program (UNU) essentially takes the swarm of data and ideas from human experts and goes about its own business. In this case, UNU used the predictions of 20 leading horse race experts, none of which correctly predicted the year’s Derby superfecta, and made the correct pick using its artificial intelligence.

TechRepublic journo Hope Reese was the one who suggested to the makers of UNU, Unanimous A.I., a Silicon Valley startup, that UNU tackle the Kentucky Derby and one can only hope she took the program’s advice and saw a big payday herself.

Horse racing is notoriously difficult to pick with a number of factors coming into play. Firstly, you have a human on a horse. Beyond that you have the randomly selected position that the horse and jockey are put into the gate. A horse that runs well in traffic is hardly suited to the outside, any more than a horse that prefers the rail is. Additionally, you could make your picks as an expert and a bit of unexpected rain would throw them out the window. Then their is organized crime…point is there are a number of variables.

“We were reluctant to take this challenge,” says David Baltaxe, Chief Information Officer at Unanimous. “Nobody here knows anything about horse racing, and it’s notorious for being unpredictable.”

While the program had predicted the winners of the Oscars, NCAA championships and the Super Bowl in the past, horse racing is a different animal (oops) and you can understand the company’s reluctance to take up the challenge.

20 experts for 20 minutes

The program asks 20 experts on horses, Oscars etc. a series of questions knowing that “various units of the group will influence each other in order to arrive at a correct decision — one that is more accurate than any individual prediction.”

Each of the 20 horse experts answered 20 minutes of questions asked by UNU. The field was then narrowed down to the four potential horses for a first, second, third, and fourth place finish. Once that was done, each expert was asked to pick those four horses in order: Win, Place, Show, and Fourth. This whole process took about 10 minutes.

A week before the race, the aforementioned, post positions were announced and the experts were asked questions again based on these positions.

“Personally, I was speechless,” Rosenberg told Market Wired when on race day UNU’s decision to bet Nyquist to win, Exaggerator to come in second, Gun Runner to show and Mohaymen  in fourth.

“We’ve been blown away by how smart UNU has been in prior predictions, but when the horses crossed the line I almost didn’t believe it, especially since we put ourselves out there by publishing the picks. And here’s the amazing thing — while the Swarm A.I. got the picks perfect, not a single individual who participated in the swarm got the picks right on their own — not one,” continued Rosenberg.

As sports gambling is by definition quite unpredictable with a trick play, weather, luck, a fluke injury (or in the case of the NFL/NBA arrest of a player before game day) there are so many variables that it’s difficult to see swarm intelligence reshape gambling, most sports gamblers are losers and using UNU to make your decisions couldn’t hurt?

The experts consulted by UNU have shown a track record (wow, two puns) of success, so who knows perhaps UNU can pull it off consistently.

CORRECTION: I mistakenly referred to Ms. Reese as a reporter from Tech Insider rather than a writer for TechRepublic. Additionally, I’ve learned that Ms. Reese boxed the UNU’s four horses for $1 ($24 bet that covers all 24 possible 1-2-3-4 finishes for the horses selected).