Microsoft's prediction system, known as "Bing Predicts," already established a name for itself prior to the start of the 2014 NFL season. The prediction model, which is popularly spread by the company's virtual assistant voice, Cortana, fired off an electric 15 of 16 correct picks during the World Cup knockout stages back in July. The only game it missed was the meaningless third place game (which saw the Netherlands defeat Brazil). Predicting results in the NFL is a little harder, but the company's method is both unique and potentially groundbreaking.
Bing Predicts harnesses a number of classic statistically-based approaches to building an accurate pick, such as a team's previous results, and specific measurements of talent on both offense and defense. Where it gets interesting, however, is in the next level of categories that were included in the model's special brew.
Fans' opinions, as expressed on Twitter and Facebook, are factored in (and Microsoft is the only company that can get access to the enormous amount of raw data from both social media giants). This allows the model to anticipate things that purely statistically-based models don't: Injuries and player news being prime examples.
In this way, it means that Bing Predicts could harness the one variable that swings hot and cold more than any other pre game measurement. Bing can shrewdly employ smart fan opinion one week, while going entirely overboard in another situation (as fans, by their very nature, can occasionally do). It's essentially both the best and worst feature in the prediction system, but undeniably makes their capability a unique one.
For those curious: Bing Predicts is 19-13 through all games this season, which is actually better than the 17-15 mark set by Vegas. Take a bow fans, but good luck keeping that going.
Screenshot via Bing
↧
