Charles Rodenkirch

11/07/2013

ECE 539 Project Proposal

Predicting Individual Overall Placement in Collegiate Waterski Tournaments

ABSTRACT

I am currently president of the UW-Madison Waterski and Wakeboard Team. Collegiate waterski tournaments consist of three events (Jump, Slalom, and Trick) for which each team submits 5 men and 5 women competitors. Each skier is given final overall ranking at each tournament. This is often considered the ultimate classifier of skier skill. However due to the complexities in determining final ranking in tournaments it is very hard for a skier to predict how he will place before the tournament concludes. This is because there is no absolute scale for final ranking. In each event skiers are scored based on a standardized point guide, these scores are then ranked against other competitor’s scores in the same event and then the skier is given a final point value for that event based on how his scores rank against the other competing skiers scores. The point values from each event are then summed to find a final composite score which is used for the overall placement. This means that any skier’s overall placement depends on how well other skiers perform which is often hard to predict, as each competitor only gets one opportunity in each event (jump, trick, and slalom). Skiers often know what scores they hope to receive in each event due to their performance in the week beforehand; however, any attempts to estimate what their resulting final ranking would be in the tournament based on these scores is nontrivial due to having to predict how the other skiers will perform. I would like to investigate using a multilayer perceptron as a tool to predict final skier ranking using historical tournament results.

PROPOSAL

The multilayer perceptron system would receive the following inputs values:

·  Weather conditions from all previous tournaments are available from weather.org [1]

·  The skier’s school affiliation as certain schools consistently perform better than others due to size and recruitment

·  The skier’s sex to determine whether he is being ranked against men or women

·  The skier’s raw scores in each competition before any ranking is performed. (Trick scores have a point value, jump scores have a distance value, and slalom scores have a buoy count.) These scores will be pulled from USA Waterski’s score archives [2] for testing and training and if the system was used for prediction would be based off the skier’s best scores in the practices leading to the tournament.

·  Number of competitors in the tournament

·  Ski Lake that is hosting the tournament

The outputs from the multilayer perceptron would be the following:

·  Estimated overall placement against skiers of the same sex in the tournament (given as a category: 1st through 5th, 6th through 10th, 10th through 15th, etc.)

TIMELINE

There is roughly 4 weeks remaining before the project must be presented. The first week will be used to compile training and testing data sets. The second and third week will be testing different multilayer perceptron configurations to determine optimal prediction. The final week will be used to write up the information gained from the project.

REFERENCES

  1. http://weather.org/weatherorg_records_and_averages.htm
  2. http://www.usawaterski.org/rankings/view-tournamentsHQ.asp?sl=on&tr=on&ju=on&sTourLevel=Collegiate&sTourRange=5