NFLGamePredictor

This presentation covers the Shiny application implement, proof of concept, adhering to the customer supplied requirements. The prototype will be described followed by a phase two and three proposal.

The Proof of Concept prototype can be found here:

https://donhescht.shinyapps.io/NFLGamePredictor/

The follow team analysis fostered the simple PoC solution.

• Historical Turnovers Forced (TOF) improves a team's points.
• Historical Points For PF (offense) predicts future scoring.
• Historical Points Allowed PA (defense) predicts opponent scoring.

• Created NFL Database using https://github.com/BurntSushi/nflgame
• Formula below uses data from 2009-2014 to predict score difference
• Compare actual game score difference to predicted.

\[ Pts\left(t,o\right)=PFGain\cdot PFAvg\left( Team \right)\; \cdot \; PAGain\cdot \left( \frac{PAAvg\left(O\right)}{PA.NFL.Avg} \right)\; \; +\; \; TOGain\cdot TOF\left( team \right) \]

• Predictor shows promise for predicting winner and spread.
• Phase 2 ($100K)
  • Improve model with machine learning
  • Increase NFL features for learning
  • Add page to show prediction of next game week
• Phase 3 (+$300K)
  • Implement model for Web and Mobile devices
  • Use Boot Strap as framework