# Predicting the Super Bowl Winner is Like Predicting the Weather

Will it snow at Super Bowl XLVIII? Will the more defensive-minded Seattle Seahawks prevail over the hard-charging Denver Broncos in the cold weather? Seems like every football fan is talking about the weather and how it will play a big factor in this year’s Super Bowl – the first outdoor Super Bowl ever to be held in a cold-weather city.

Taking some of the guess work out of game day predictions

As fantasy football players know, predicting the outcome of a football game is never easy because many factors come into play – and all the factors are intertwined in complex ways. Earlier this year, I was fortunate enough to be part of an SAP team that put together a tool for the NFL called the Player Comparison Tool. And the tool tries to do just that – predict how well individual football players will do in upcoming games. The challenge we faced was applying an advanced statistical model to a tool that every fantasy football player can use.

The intent of the tool is very straightforward – it is designed to help the fantasy football player make decisions, such as whether to sit or start a player in a certain week, whether to add or drop a player from his roster, or how to make fantasy football player trade decisions. The tool needs to be powerful enough to appeal to the seasoned fantasy football player, but user-friendly and intuitive for the beginner player.

Analyzing the factors that come into play on game day

We have a data scientist on the SAP team who is an expert in discovering patterns among the myriad of statistics kept on individual football players, plays within a game, and football matches. He crunches the numbers and discovers trends and correlations hidden within. What makes this interesting is the breadth and the depth of how these factors end up being used in our predictive statistical model.

To make the tool effective, we keep track of close to 100 contributing factors, including:

• How jet lagged the players might be
• How good a team is on rushing
• How the weather is predicted to be on game day (we query a national weather forecast service to get up to date weather data)
• How well teams perform on different types of turf

The data scientist determines how these individual factors need to be weighed using statistical methods. We then take the statistical model and implement it in SAP HANA.

Creating a customized experience

At runtime, we allow the fantasy football player to tailor the projected player performance to his or her own liking. For example, if a player “Bob” feels consistency is the most important thing when he forms his roster, he can slide the consistency slider up to make it a more important factor. When Bob runs the Player Comparison Tool, available through the NFL.com portal, a live query is sent to one of the SAP HANA servers running in the cloud. The tool crunches the statistical model in real time – taking into account the custom user slider values, the latest weather info, game scores, and player statistics – and projects the players’ performances.