Science

A.I. Predicts the 2018 Rose Bowl and Sugar Bowl Winners

Seventy college football fans swarm to make a prediction

by Unanimous AI
Getty Images / Mike Comer

Researchers at the San Francisco-based startup [Unanimous AI have used our Swarm AI platform to generate weekly predictions for every single NFL game this season. Using its Swarm AI platform to form a real-time human network, these groups of fans have amplified their intelligence with sometimes remarkable results.

Now, Unanimous has used that same Swarm AI platform to create the AI Guide to the College Football Playoffs. Our Swarm AI technology works by amplifying the intelligence of any group, in this case college football fans, by combining their knowledge, wisdom, and intuition using AI algorithms. And, just for good measure, Unanimous researchers used the Swarm AI technology to consider all of the major college bowl games to be played in the coming days. Last year, a similar group of 75 college football fans beat the ESPN experts by working together as a Swarm AI system. In that experiment, the group amplified its own individual intelligence by 15 percent and outperformed ESPN experts by 11 percent in another example of swarming’s ability to turn groups of regular people into “super-experts.”

As you’ll notice, the chart above (larger view here) is sorted by “best bets,” then games where the College Football Swarm recommends you “proceed with caution” and finally, games where the Swarm AI’s insight is that there is no insight to be gained in a particular matchup. As we’ve seen over the course of the NFL and English Premier League seasons, this sort of ranking is crucial to good forecasting, as picking winners is only half of the battle. The other half is knowing when no significant insight is available and it’s best to leave your money in your pocket.

With that said, while the games are also assigned a betting confidence and a predicted spread, theoretical wagers were generated by the Swarm AI system as a numerical expression of confidence in the forecast outcome, enabling us to rank the relative strength of the predictions across games.

While there are a number of exciting match-ups on the table, the Rose and Sugar Bowls (January 1) are the most exciting games to discuss, since the winners will advance the College Football Championship. This four-team playoff is in just its fourth year of existence, but has already generated some incredible games and with any luck, this year will be no different. So, how did the College Football Swarm separate Alabama from Clemson, and Georgia from Oklahoma?

First, our researchers asked the Swarm AI system to pit the offenses and defenses of the semifinal teams head to head. The replay below shows the group of nearly 70 college football fans converging on an optimized response to the question, “Which Sugar Bowl team has the best defense?”

What may seem like a simple response actually incorporates a tremendous amount of information, as each participant must process whatever objective information he or she has in combination their own wisdom, experience, intuition and insight. Then, the Swarm AI system must enable rapid convergence from 70 different perspectives in real-time.

For example, Clemson’s defense is sixth in the nation in yards per game, while Alabama’s is second, but Clemson played the fifth-hardest schedule, while Alabama’s strength of schedule was ranked ninth. The replay above shows how a large group of people can process that information and converge on the insight that Clemson’s defense is even more formidable than Alabama’s (by a little) in this rematch of the last two national championship games. A similar question from Unanimous researchers revealed the perception that Clemson’s offense is also better than Alabama’s. With those two insights in hand, it’s little surprise that the Swarm AI system picked Clemson to beat Alabama.

Who Will Win the 2018 Rose Bowl?

As reminder, the Clemson pick was slotted into the “Proceed With Caution” category, which is almost certainly good advice where betting against Nick Saban and Alabama are concerned, especially when he has extra time to prepare his team. Still, the group was far less decisive when asked to consider the Rose Bowl matchup between Oklahoma and Georgia. In a question that naturally pit the Sooners’ Heisman trophy winning quarterback vs the Bulldogs’ record-setting freshman, the group quickly converged on the response that Oklahoma’s offense outpaced Georgia’s (by a little). On the other hand, Georgia’s defense is widely considered to be one of the best in the nation, while Oklahoma plays in the Big 12 where defense is more of an afterthought to their high-scoring offenses. It follows then, that the Swarm AI technology perceives the Bulldogs’ D to be better than the Sooners’.

Thus the group of college football fans had to weigh Oklahoma’s perceived advantages on offense against Georgia’s punishing defense. This is the type classic “unstoppable object vs immovable wall” matchup that fans love, but forecasters hate. So, while the Swarm AI system was able to converge on a prediction, it’s worth a quick reminder that this pick was filed under “High Risk” or inconclusive. So, really, the insight from the Swarm AI technology is that you should just sit back and enjoy what should hopefully be a great game.

Finally, since Clemson and Alabama are destined (doomed?) to meet in the semifinal, the national championship game will not include both those teams for the first time since Oregon played Ohio State in the first ever College Football Playoff Championship. So, if the predicted teams both advance, Clemson and Georgia will face off for the national title. While our researchers will be doing a deep dive on the actual matchup next week, if these two teams do meet for the championship, at this point the Swarm AI system gives the edge to Clemson.

Related Tags