New England vs. Atlanta: AI’s Predicted Winner May Surprise You
With Sunday’s big game around the corner, you may be wondering how to outsmart your co-workers to win the office pool.
You could stick with the Las Vegas line, which has New England favored to win by three. You could go with the theory that says that Atlanta lost its chance when it opted for red uniforms, as teams wearing white, like New England on Sunday, have won 11 of the last 12 championships.
If you hate Tom Brady, you’ll like what it says for Sunday’s game. Although most expect Brady and New England to win a sixth championship, as of Feb. 2, Swish’s AI said the star quarterback will be disappointed.
“This is going to be a close one, but our models give the Falcons a 5 percent advantage,” said Corey Beaumont, Swish co-founder and chief operating officer.
Go Crazy: Bet on Every Play
But forecasting basics like the game winner, the spread or the total points is just a warm-up for Swish’s GPU-accelerated predictions. Its machine learning algorithms offer bettors and fantasy players live play-by-play odds as the action unfolds. If you’re game, you can bet on every single play.
“We’re collecting thousands of data points before and during the game so we can make real-time projections,” Beaumont said.
That data includes media reports, social media, league data and information from statistical services.
Goodbye Credit Cards, Hello Hoops
Swish’s founders were working at a large credit card company when they realized they could apply some of the analytical tools from the world of finance to the $1 trillion sports betting market. Although sports is heavy with statistics, Beaumont said stats focus on how a player or team has performed in the past.
The founders, a trio of what Beaumont describes as “massive” basketball fans, saw an opportunity to offer predictive data, with help from AI computing. As an experiment, they tried predicting the outcome of NBA basketball games in their spare time.
“It turned out we were pretty good at it,” Beaumont said.
They launched Swish in 2014, offering machine learning predictions for NBA games for the first year. They’ve since added professional baseball and football, and are working on ice hockey.
“The (new) GPUs will allow us to iterate faster and handle more data. This will allow us to provide a more accurate product faster,” Beaumont said.
Checking the Stats on Swish
Swish provides subscribers data like a win confidence percentage, suggested betting amounts, up-to-the-minute injury lists, player and referee analysis, and total projected wins per team for the season. Subscribers looking for betting analysis pay $99 a month per sport. Daily fantasy tools that help players build winning teams costs $20 a month per sport.
Swish also has a free NFL Live app that predicts plays during the game.
So how good are Swish’s AI sports predictions? Bettors who followed Swish’s recommendations for every game of the latest NFL season saw returns as high as 28 percent, Beaumont said.
Beaumont said users find Swish a refreshing change from sports commentators and columnists among all the other biased perspectives in sports. “We take out all the biases and just look at the data,” he said.
And if using AI is a leap too far for your Big Game picks, you can always rely on the frog forecast from a Texas conservation center. The frogs had the choice of hopping to a fruit fly-infused football on the Atlanta side or the New England side of a mini football field. They chose Atlanta, by a small margin.