MLB Regression via Game Score

Regression measured thru Game Score

What is regression and Game Score and why should it matter in handicapping? Let’s get some definitions we can work with. First is for regression. I am using the terminology of “regression to the mean”. This is copied directly from a Google search. “In statistics, regression toward (or to) the mean is the phenomenon that if a variable is extreme on its first measurement, it will tend to be closer to the average on its second measurement—and if it is extreme on its second measurement, it will tend to have been closer to the average on its first.” Think in terms of outliers. Going in the direction more extreme than another, the outlier is not the mean (average bar). Humans are not machines and have progression and regression toward the mean. If we could gauge those “outliers”, then we could bet on those opportunities they present. Here is where Game Score comes in.

What is Game Score? The next few lines are copied directly from Fangraphs. “Game Score was originally created by Bill James to measure the quality of individual starts.” “Calculating Game Score (original or v2) is extremely simple, requiring some basic addition, subtraction, and multiplication. The original Game Score is calculated like this:

Game Score = 50 + Outs + 2*(IP Completed After the 4th) + Strikeouts – Hits – 4*Earned Runs – 2*Unearned Runs – Walks”

Game Score tries to answer the fundamental question, “how good was that start?

Fundamentally, it is important to use Game Score because the Win stat for a pitcher does not indicate the performance of the pitcher nor if the performance was above or below mean average.  Ok. So now we know what Game Score is, how to calculate it and why it is important. 

Now how do we use it? I string together the last 7 starts for a pitcher and cut them up into segments. Seven games are roughly 21% of a season and at a glance is a good representation of current performance. Here are the charts I use on my spreadsheet.

The charts on the left are the actual GSc from the past seven games. The chart on the right shows the average GSc for a segment of games. The 1^st box is last 7 games, then the last 3 games, then the last game and finally the year to date. The box GSc vs OPP is the GSc vs the team pitching against. In the example above, Graveman is 58 over his L7 which is significantly better than ytd. His last 3 and his last game also indicate he is above his ytd. The ytd is the mean. This tells me that Graveman is going to regress soon. It may not be today but other factors may lead to believe that. Such as how he fares on the road, if key hitters light him up if he struggles vs the team he is facing and so forth. I assure you though, he will regress soon. The colors in the chart reflect poor performances (green light) or great performances (yellow light). They are also great information. Getting to a mean requires outliers, and those are them. Yellow means the pitcher is like to regress back to his average mean and green means the pitcher is likely to progress to his mean. Over time, you will be able to pin down opportunities you didn’t know were even there. A Sabermetric way to way performance that predicts the future. Pretty cool!