The predictions come from a model by Bing Predicts that takes into account 2013 results, offensive and defensive stats, margin of victory, location, weather, and public sentiment (which is meant to help identify real-time factors, like injuries).
Since the model just picks a winner and doesn’t predict a score against the spread, the hit rate should be well over 50%.
Cortana did well at the World Cup, and even picked Germany over Brazil in the semifinal when other statistical models had Brazil as the heavy favourite. So if you’re the gambling sort, it might be wise to take these picks into account.
Here are the Bing Predicts odds for Week 1:
- Green Bay Packers at Seattle Seahawks — Seahawks win (74.2% chance)
- New Orleans Saints at Atlanta Falcons — Falcons win (61.4% chance)
- Minnesota Vikings at St. Louis Rams — Rams win (67.4% chance)
- Cleveland Browns at Pittsburgh Steelers — Steelers win (70.2% chance)
- Jacksonville Jaguars vs. Philadelphia Eagles — Eagles win (75.4% chance)
- Oakland Raiders at New York Jets — Jets win (62.9% chance)
- Cincinnati Bengals at Baltimore Ravens — Ravens win (59.8% chance)
- Buffalo Bills at Chicago Bears — Bears win (64.4% chance)
- Washington Redskins at Houston Texans — Texans win (51% chance)
- Tennessee Titans at Kansas City Chiefs — Chiefs win (64.4% chance)
- New England Patriots at Miami Dolphins — Patriots win (61.4% chance)
- Carolina Panthers at Tampa Bay Buccaneers — Buccaneers win (56.6% chance)
- San Francisco 49ers at Dallas Cowboys — 49ers win (68.8% chance)
- Indianapolis Colts at Denver Broncos — Broncos win (77.8% chance)
- New York Giants at Detroit Lions — Lions win (61.4%% chance)
- San Diego Chargers at Arizona Cardinals — Cardinals win (58.2% chance)
Interestingly, Cortana picked the Vegas betting favourite in 15 of 16 Week 1 games. The only disparity is in the Saints-Falcons game. Vegas has New Orleans as a three-point favourite, while Cortana has Atlanta as a pretty big 61% favourite.
Business Insider Emails & Alerts
Site highlights each day to your inbox.