MEASURING AND PREDICTING UW BADGERSS PERFORMANCE BY QUARTERBACK

MEASURING AND PREDICTING UW BADGERS’S PERFORMANCE BY QUARTERBACK AND RUNNING BACK STATS By: Tyler Chu ECE 539 Fall 2013

Reasons to Predict • Millions of Badgers Fans who want to know how their team is going to do • Immense amounts of money go into the NCAA football programs

Main Problem & Goal • Problem: • Most predictions available have a human bias in it which stems from personal opinions that could result in errors with the predictions. • Goal: • Eliminate the human error by having a Multi-layer Perceptron to perform the prediction

Why MLP • Teams can win in a variety of ways • No linear mapping exists to determine the outcome • No one piece of the data always correlates to a win or loss as there are many ways in which a team can win or lose.

Why MLP • MLPs • Multi-Layer Perpceptrons are capable of predicting outcomes of non-linear data. • Multi-Layer Perceptrons reduce the problem to a Neural Network prediction problem and remove the human personal bias of a teams performance from the prediction.

Data Collection • Data was to be available the web’s many different sport statistic sites. • A large data set was required to represent the large number of ways to win • Used Sports References’s website • Used Excel’s web query feature to acquire tabular data

Data Collection • Many feature vectors were collected • • • Passing Completions, Attempts Yards per attempt Touchdowns Interceptions Passer Ratings Rushing equivalents for RB’s

Preliminary Results • Data was formatted in Matlab and then fed into a modified MLP Matlab program provided from the class website. • Multiple tests run using the same variables for alpha and momentum set to default values of 0. 1 and 0. 8 respectively • Average of initial results on the data with one hidden layer and neuron was a 73. 6842 classification rate

Initial Test

Secondary Test

Results • Additional hidden layers and neurons eventually converged to a 95% classification rate • Decided to predict future seasons based upon if the current quarterback and running back stay – generally large difference if they do not

Results • Use a linear formula between each consecutive season • Found that UW would improve to a 9 win season if Stave and Ball both stayed • Currently at 9 wins with one game to go

References • Newman, M. E. J. , and Park, Juyong; A network-based ranking system for US college Football. Department of Physics and Center for the Study of Complex Systems, University of Michigan, Ann Arbor, MI 48109. ar. Xiv: physics/0505169 v 4 31 Oct 2013 • ESPN, ESPN College Football. 8 Dec. 2013 http: //espn. go. com/college-football/team/_/id/275/ • Sports References. SR College Football. 8 Dec. 2013 http: //www. statfox. com/nfllogs. htm