Predicting Win Probability - Ravens 2014

Leif Ulstrup
August 12, 2014

Premise

The probabilty of the Baltimore Ravens (under quarterback Joe Flacco) winning a game in 2014 can be predicted using 2008-2013 Win/Loss data to create a generalized linear model relating wins to Ravens points scored (predictor).

Joe Flacco began as Ravens quarterback at the start of the 2008 Season; same quarterback for 96 games
Strong offense performance reduces the dependence on defense and should be a good predictor of win probability
It is fun for fans to assess win probability in the middle of a game

A Shiny web application was created to enable Ravens fans to predict win probability by inputing the current score.

Predicting Ravens Win Prob vs Ravens Score

Built a generalized linear model (GLM) using all Ravens regular season game wins/losses and score data (n=96) from 2008 until 2013. Data source: NFL.com and wikipedia.com.

plot of chunk unnamed-chunk-1

	Estimate	Std. Error	z value	Pr(>\|z\|)
(Intercept)	-3.26	0.83	-3.91	0.00
RavensScores$Ravens	0.18	0.04	4.55	0.00

Link to Shiny Application

A Shiny application was created to host an interactive version of the model on the cloud based site - shinyapps.io

Click Here to Link to Shiny App for Interactive Version

The source code for the Shiny application and this presentation along with the Ravens 2008-2013 Win/Loss and Scores data can be found here at this link https://github.com/leifulstrup/developing-data-products-ravens

Future Explorations

Potential future predictive model explorations:

Consideration of Home vs Away game as a factor
Quarter of Play & Time Remaining
Greater Weighting to Recent Games and Season
Inclusion of factors such as Turnover Ratio
Inclusion of Defense and Offense Performance Measures

GO RAVENS!