Making a Sporting Bet: Machine Learning and Sports
Prediction and professional sports has long been a national pastime, maybe even an obsession. Whether it’s Obama filling out his NCAA tournament bracket, or the Nevada books setting the point spread for the Superbowl, we are driven to better understand what will happen on a field, pitch, or court.
We’ve combed the archives of models from the BigML Gallery and highlighted a few interesting sports related models. While these models won’t necessarily help you win bets, they can help you better understand how certain sports statistics have lead to certain outcomes.
NCAA Football Heisman Rank
At BigML, we already have a number of sports related models, many of them available on our public gallery. For instance, here’s a BigML model that predicts the Heisman Trophy rank for a given candidate. This model only works for Heisman Trophy candidates, i.e. the top ten players that were considered for the trophy, for a given year, between the years that the trophy was awarded (1935-2011).
In the absence of any other information about a player, the predicted rank is (unsurprisingly) 5 out of 10. However, one of the most important things you can do to improve your rank as a Heisman candidate is score more than 16 touch downs. Just by accomplishing that, you’ve improved your rank on average to around 4.6.
World Cup Fifa Golden Ball Winners
The Fifa Golden Ball is one of the most prestigious awards a single player can receive. It is given to the most outstanding player at a World Cup Final. In the relevant BigML model, we can see that accumulating 60 points would give you good odds of winning prior to 1993. However, to have good odds of winning in later years you would need over a hundred.
The amount of money that most professional athletes can make is astonishing. Every shot, rebound, pass, or game could be worth tens or hundreds of thousands of dollars. Using some recent active player data from the National Basketball Association and a BigML model, we are able to predict how much an arbitrary player is worth, based on the equivalent stats and salaries of recent players. We can see that making over 12 points a game will increase your average salary from 3 to 8 million dollars. If you can make 6 or more field goals a game, you’re worth closer to 11 million. Of course, there are exceptions to this sort of model: Rookie players and veterans have different caps and limits for their salary. Also, certain player statistics might be negatively affected by midseason injury or trades.
In summary, you don’t have to be a Vegas-style gambler to use machine learning on sports data. Sometimes you just have a burning “what if” question, or you want to see what your favorite player’s chances are of winning an award. BigML can help you no matter what your sports prediction needs are.