A football match is a thrilling encounter -- oh, the well-orchestrated passes, solid defenses and breathtaking offensive. However, the outcome is the most important part of the story. That aside, football prediction has come in handy. Considering chances of making money with slew of betting companies today, correctly predicting an outcome is such a huge task.
Love the NFL? Find out what happens next at ff-winners.com?
NFL just like any other football league is arguably competitive. That means knowing the winner in a game is an uphill task -- but not anymore. Thanks to statistical models, we can use past data, present conditions and future prospects to know what’s likely to happen.
Since knowing football results is perhaps a game of chance, we should up our chances of getting it right. Well, as you have already fathomed, now that gets us deep into probabilistic models. But they are not really difficult. I will spare you the technical stuff -- save a few definitions to make it clearer.
Simply put, logistic regression is an extension of the normal linear regression model. The linear one measures continuous variable (able to take any value in a scale) whilst the logistic extension deals with categorical data (containing categories like Win or Loss).
Take Me to the Pitch Please
Okay, now back to football, we want to know whether the home team will win or lose. Of course, a draw could also be in the math. But we will run away from that for now. However, for those interested in knowing why, a win/loss is the most probable outcome in any match -- that’s if everything doesn’t turn into a game of two equal halves.
Now, many factors affect a football match and ultimately the result. These include weather, motivation, fans morale, current form, style of play, coach capability and so on. But these are the prominent ones we want to use -- style of play and current form.
Combining the two, we can assign an efficiency score. Additionally, result of the previous game (s) is important. So, a win will spur motivation and vice versa. In my illustration, I will use two predictor variables -- efficiency and motivation from past results. Obviously, the dependent variable is Win or Loss, coded as 1 and 0 respectively.
That’s the magical formula. P(Y) is the probability of Win while b0, b1 and b2 are the parameters. X1 and X2 are our predictor variables, efficiency and motivation respectively. Don’t be perplexed by the mysterious e; it’s just an error term. Calculations are pretty easier with statistical software nowadays. For instance, if you want to predict whether or not Baltimore will win against Seattle, use previously recorded efficiency and motivation scores of the team.
After plugging in the values, the logistic regression will give you an outcome ranging from 0 to 1. If the value is closer to 1, the chances of victory are higher. On the other hand, a value closer to 0 is indicative of slim victory chances. Of course, you can make the model complex by adding more predictor variables to increase your chances of arriving at the correct answer. Meanwhile, that’s it for now.