GCN LAB IMPRESSIONS
Can computers really predict the Super Bowl?
- By John Breeden II
- Jan 23, 2012
We now know that the New England Patriots will be playing the New York Giants in the Super Bowl this year. Some people might be bored at the thought of the same teams competing against each other again so quickly, but others probably see it as the rematch of the century, given that the Giants played spoiler to the Patriots getting a perfect season four years ago.
If only computers could predict who would win this time. Apparently, scientists at Los Alamos National Laboratories are thinking about that very same thing.
Trying to predict the future is nothing new. Ask Nostradamus. The History Channel, in its apparent attempt to prove that the world is getting ready to end this year, keeps running specials about how he and others think our time on Earth is drawing to a close. Forgive me on this, but I liked the History Channel better when it actually dealt with the past. I used to lovingly call it The Nazi Channel since most programs were about World War II in some way. But now it’s all about aliens and the end of the world and junk I don’t really care about.
Anyway, one interesting thing on the History Channel was a segment about how, if the actions of people could be expressed mathematically, future actions could possibly be predicted. Perhaps that is what the scientists at Los Alamos are researching. In fact, if you think about it, that may have been what Nostradamus was trying to do in his head.
But I’m not so sure it could ever work. Let’s get back to the Super Bowl for a moment. Here is an event where most of the variables are known, at least compared to the wide world of real life. We know the players, their records, their abilities, the strategies used by both coaches in the past. Even the weather can be factored in most of the time, especially for indoor stadiums.
I have a friend who tries to predict the outcome of the game every year using his Xbox 360 and a copy of Madden NFL. He makes sure that all the most recent data is part of the game and then runs the simulation a hundred times to get the most probable outcome.
Even so, his results have been less than stellar. While he is often fairly close with the final scores, within three points a lot of times, the actual winner of the game has only been correct 50 percent of the time over the past six years. That’s no better than a coin flip. I can program my computer to give me a random answer and get the winner with the same accuracy. Then again, he could probably beat the point spread using this method. Hmmm...
For what it’s worth, EA Sports, which makes the Madden NFL game, has run its own Super Bowl simulations since 2004 and has correctly predicted the winner of six of the last eight games. The two it got wrong: last year’s Green Bay Packers win over the Pittsburgh Steelers and the 2008 game between the Giants and Patriots.
But even if there is something to computer predictions of sports games, expanding this to other events in the real world would be so much more difficult. You can find out that Tom Brady completes 82 percent of his passes on second-down plays and add that into your formula, but how do you assign a number to President Barack Obama’s diplomatic prowess or the reaction that Israel would have to a certain action in regard to Iran?
Sadly, I don’t think computers will ever accurately predict the future using math. They might be able to make the calculations for us, but we would probably fail to get the proper variables into them.
That doesn’t mean it hurts to try. Looking at my random generator, I now ask it who will win the Super Bowl. And the answer is…the New York Giants. You heard it here first. The computer is always right, except when it’s not.
John Breeden II is a freelance technology writer for GCN.