March 12, 2013

Predicting Runs

The series using the Lineup Analysis Tool (LAT) to look at lineups is coming up. With five years of data under our belts, here is how the predictions worked out:

Run Prediction Graph

The 2010 Mariners, predicted to score 4.92 runs per game, actually scored 3.17 per game. They are such an outlier that I’m tempted to take them out of the regression equation. That would likely make the slope of the line steeper. You can see that due to the outlier, teams at either extreme are not evenly distributed around the mean.

For 2012, the correlation between prediction and actual runs scored was 61%. The regression formula I’ll use based on the five years of data will be y = .7161x + 1.018.

You can watch follow the data in this spreadsheet.

Leave a Reply

Your email address will not be published. Required fields are marked *