The two constants of March Madness are #12 seeds upsetting #5 seeds and preview articles marveling at the astronomical odds of putting together the “perfect bracket.” But there is another long-shot story that is far more compelling – and relevant – when it comes to forecasting the most unpredictable athletic competition of the year: the comparison between a small group of college basketball fans using the Swarm platform vs. the 17-million person ESPN “People’s Bracket.”
The “Wisdom of the Crowds” is oft-cited as the quintessential example of a group pooling its intelligence, but Swarm Intelligence has been proven time and again to be a more efficient method for actually amplifying the intelligence of those groups. In 2018, 50 basketball fans using the Swarm platform correctly predicted Villanova to win the national championship and beat all but 8% of brackets in the ESPN challenge. In contrast, the ESPN People’s Bracket aggregated the predictions of 340,000 times as many basketball fans and finished in the 60th percentile of all brackets. Worse still, the People’s Bracket picked UVa to win the championship, but the Cavaliers became the first #1 seed ever to fall to a #16 seed in the first round.
How do 50 people easily outperform 17,000,000? The answer is the same reason birds flock and fish school in nature. Swarming allows them to be far more intelligent together than they would be on their own. At the 2018 March Madness tournament, the Swarm platform enabled those 50 people to combine their wisdom, intelligence, experience and intuition with AI algorithms into an optimized bracket forecast. The People’s Bracket merely averaged the group’s intelligence, and got average results.
For 2019, another group of basketball fans has gathered inside the Swarm platform to try to make sense of March Madness. In addition to the complete bracket which is available below, the group’s insights also produced a forecast that ranks the likelihood of each of the eight highest seeds to cut down the nets.
That predicted matchup between rivals Duke and UNC would be a rematch of last week’s ACC semifinal, won by Duke by a single point. Zion Williamson, the transcendent Duke star, recently returned to the lineup after injuring his knee in the first minute of the earlier clash between the two top-five teams. With Williamson back in action, the AI expressed high confidence in Duke’s ability to prevail once more in what would be a historic final for these longtime rivals.
Naturally, adding a unique talent like Zion back onto Duke’s already-loaded roster means that many experts and pundits will favor the Blue Devils to be the last team left standing, so the complete bracket will be required to separate the wheat from the chaff. To check out the AI’s complete forecast for each of the 63 games required to crown a champion, follow the link below.
In addition to having the swarm generate raw picks, as shown above. Unanimous AI also post-processed these picks using the Brainpower values generated by the Swarm platform for each selection. Brainpower values give an indication of the strength of confidence expressed by the swarm when making a pick. By comparing the Brainpower confidence values for the confidence levels we would expect to see based just on the seed numbers (i.e. a #6 playing a #11), we were able to identify picks that were likely going to go the other way. This enabled us to generate a final bracket, which we entered into the ESPN contest. You can see how that Bracket is doing at the link below:
Want to learn more about our Swarm AI platform? Check out our TED talk…