Analysis on the predictive capability of statistical metrics

Matthew S.

Order the writing of a tailor-made Management Case study

Free quote online

Case study Format .doc

Analysis on the predictive capability of statistical metrics

Download

Read an extract

Themes

Statistical metrics, overall performance

Reader
Abstract
Contents
Extract

Abstract

As the largest professional football game of the year approaches, Super Bowl 49, sports fans are inundated with team and player statistics, records, and superstitions – all in an attempt to predict how the ‘big game' will turn out. While attempting to ultimately make an inference as to who is thought to win, the focus of this paper is to verify the accuracy of the famous Bear Bryant quote – Does a strong offense make it more likely to win a Super Bowl? Statistically speaking, is a strong defense more important?

Selecting factors in which to analyze was not as straightforward as it might seem. There is a massive amount of data available on individual players as well as teams. Every ‘snap' is quantified, every run tabulated, there is a plethora of information from which to choose from. Ultimately the decision was made to analyze two important statistics: the number of points scored by a team during regular season play (a gauge of a team's offensive strength), as well as the number of points scored against a team during the regular season (a gauge of a team's defensive strength). Of all factors to choose from, it was felt that the points scored data most accurately represented the overall performance of team metrics, an ‘all encompassing' statistic (for instance the use of quarterback passer rating focuses too heavily on an individual player).

Get this table of contents for free after login.

Extract

[...] Data was gathered from the entire current 48-game population of past Super Bowls for comparison's sake. In this case, though an entire population will be analyzed, the population will grow each year, allowing us to use this current population as a sample for future projections of the new population. Initially, we seek to determine if the above statement is accurate for the NFL, and then follow through with a fun prediction for this year's game Data Data collection was relatively straightforward, with all the necessary information for every Super Bowl ever played available on the NFL's website (www.nfl.com). [...]

[...] Analysis To begin, we defined population 1 as the better defense (lower points against) and population 2 as the better offense (higher points for). Proportionally, we determined the following parameters: This data can be visualized in the Chart 1 and 2 for population 1 and respectively. Chart Proportion of teams with the better defense that won the Super Bowl. Chart Proportion of teams with the better offense that won the Super Bowl. Note that unlike a proportion of winners and losers of these games (which would be 50/50), the proportion of winners with the best offenses and best defenses adds up to over half of the games won. [...]

[...] Probability table using data from table Statistical Inferences Examining our first population, teams with the better defense, gives the following margin of error at 95% confidence: Therefore, our confidence interval for a team with a better defense is Likewise, for teams with a better offense, we find the following margin of error at 95% confidence: Therefore, our 95% confidence interval states: Finally, getting to the heart of our original question, we first determine whether the proportion of teams with the better defense winning the game is statistically different than the proportion of teams with the better offense. We define the following: Hypothesis Test Can we say with a 0.05 level of significance that there is a difference in predicting the winner of the super bowl using points for and points against? a. : - Points scored on a team during the regular season is not a valid predictor of who will win the super bowl. [...]

[...] This could be either due to the lack of a comparison between the other ranks (ie a second ranked defense and a fifth ranked offense may be better than a top ranked defense and a 19th ranked offense), due to another statistic having importance, or due to the presence of an intangible on the field. One obvious, but possible predictor is if a team has both the better offense and defense. In these cases, the overall proportion of winning teams was around suggesting a higher statistical relevance. [...]

[...] “Offense sells tickets, Defense wins championships” -Bear Bryant Analysis on the predictive capability of statistical metrics 1. Introduction “Offense sells tickets, defense wins championships” -Bear Bryant As the largest professional football game of the year approaches, Super Bowl 49, sports fans are inundated with team and player statistics, records, and superstitions all in an attempt to predict how the game' will turn out. While attempting to ultimately make an inference as to who is thought to win, the focus of this paper is to verify the accuracy of the famous Bear Bryant quote Does a strong offense make it more likely to win a Super Bowl? [...]

doc