|
|
|
|
An essay on contemporary sports handicapping: Mathematical Modelling of Sports
Is it actually possible to use mathematics to predict the outcome of sporting events? The answer is yes.
Most of the outcomes of sports can be modeled mathematically. There is much debate over this but the truth is that, with the right data and numbers, you can predict the outcomes of sporting events well enough to beat the spread. The problems is more with getting the right data. There is strong evidence that it is nearly impossible to beat the vig on horse racing. This includes some very sophisticated evidence using all the publically available data. However, other sports are fully beatable. Football and hockey are the most difficult, basketball and baseball are considered to be the easiest to beat.
The tough part about math models of sports is obtaining historical data to justify the model. We are lucky in that many of the necessary statistics are publically available, or are purchaseable.
Variables that are used to predict the outcome of sporting events are both linear and non-linear. For example, a 1 year old horse would not win a race, a 2 year old might, a 3 year old would show significant improvement,.. up to about 8 when the horses performance would begin to drop significantly. Thus, age is considered a non-linear predictor of horse-race performance. A linear predictor in horse-racing may be the average dollars earned per start. You can easily see how the higher this number the better the horse. Thus, this is a linear predictor of racehorse performance.
The optimal tool for the final prediction is to use multiple regression with both linear and non-linear components. Imagine you have pie and each slice of the pie is one of the variables you use to predict the outcome of an event. In the ideal you would complete the pie and know with 100% certainty the outcome of the event. However, this never happens. More realistically you have only a limited % of the correct variables and all you can do is try to improve the % of the 'explained variance' in predicting the outcome. The better the data you have.. the more slices of the pie you can fill up. You will never reach 100%.. but you can reach up to 50% with nearly perfect data. If you can correctly include both linear and non-linear predictors of outcomes in your math formula.. you will beat the spread.
Sounds tough? Well in a way it is. I have received my Ph.D and was lucky to be trained in how to conduct such analysis.
Questions? email at professor@hrmgt.com
| |