How 60 random sports fans on Reddit, blew away the forecasters at FiveThirtyEight
ESPN’s FiveThirtyEight website has built a reputation for deep analysis and forecasting confidence. Nowhere is this more evident than the prediction they published before yesterday’s Orange Bowl. The in-depth preview stretches out over 1,000 words, contains two statistical tables, a plot chart and a host of fancy acronyms (QBR, FPI, ELO, etc.) to support the claim that #4 Oklahoma would beat top-ranked, undefeated Clemson.
We now know how that turned out: OUCH!
Of course, one incorrect prediction isn’t particularly noteworthy. What is noteworthy is the arrogance exuded by FiveThirtyEight in naming Oklahoma as the team most likely to win the national championship. In saying that an Oklahoma win would “reconfirm FiveThirtyEight’s soothsaying abilities,” the folks at FiveThirtyEight have cast themselves as the super-experts that average fans could never match.
It turns out, average fans have a lot of wisdom to offer.
Researchers used a “human swarming” platform called UNU to tap the collective knowledge, insights, and intuition of 60 average sports fans making predictions around the College Football Playoff. A merger of artificial intelligence and collective wisdom, UNU swarms quickly converged on optimal solutions. Below is a replay of the 60 person swarm predicting the winner of Orange Bowl between Oklahoma and Clemson:
Clemson’s victory marked the second time in two days that average fans using UNU have defied both Vegas and the so-called experts. The story was much the same when UNU correctly predicted that Baylor would beat UNC, while ESPN, Sports Illustrated and Bleacher Report all predicted an easy win for the Tar Heels. Below is a replay of 68 people, working together to predict the Baylor win:
Of course, a few accurate predictions can be written off just as easily as a few inaccurate ones. That’s why researchers have been testing UNU across a range of venues, from predicting the Super Bowl and Stanley Cup, to predicting the winners of the 2015 Oscars, in all cases outperforming FiveThirtyEight. This research demonstrates that average people, thinking together, can consistently beat the so-called experts at their own game.
Stay tuned to this blog throughout the College Football Playoff, as we’ll be publishing the results of all ten major bowl games that fans used UNU to predict. You can see the full rundown of those predictions HERE, as well as replays of all of the swarms in action.
Want to be part of a swarm? Great, researcher are looking for sports fans to help make new predictions. Just drop us a line in the link below and we’ll include you in the Beta Program. And it’s not just sports that human swarms are predicting – also politics, finance, movies, music…