Applying Sabermetrics to Hockey III - FIP to ERA is relCF% to relGF%

Part One - Bill James's Pythagorean expectation
Part Two - Pitching Runs Applied to Goalies
It's fun seeing how my plotting skills have improved over time... anyway...

Stats for this blog were provided by

Will try to keep the baseball teaching short in this one and get to the meat... 

FIP is a statistic that was introduced as a way to evaluate pitchers with a slight difference to ERA. FIP, fielding independent pitching, serves as a way of evaluating pitchers on the things that the pitchers can control and does not allow the pitcher to play victim to the defense behind him. The formula: 

FIP = ((13*HR)+(3*(BB+HBP))-(2*K))/IP + constant

The constant exists to bring FIP closer to an ERA scale, and is dependent on around the league factors. 

Click here if you want to read more about FIP

What's interesting about FIP is that analysts began to notice that it was a better predictor for a pitcher's future ERA than ERA was, and thus, the importance of FIP was born.

So naturally, I had the thought. Since shot attempts for are a predictor for goals for, is it possible that an individual player's relative shot attempt % is a better predictor for their future relative goals for % than historical relative goals for %?

The answer is yes.

To test this out, I used three different data sets with splits for all positions, only forwards, and only defensemen.

  1. First half of 2015-2016 versus second half of 2015-2016
    1. Minimum 21 games played in both halves
  2. 2014-2015 versus 2015-2016
    1. Minimum 41 games played in both seasons
  3. 2012-2014 versus 2014-2016
    1. Minimum 82 games played in both two season spans

In all three data sets, relCF% performed better as a predictor than did relGF% This was across all positions, subset of forwards, and subset of defensemen. 

Charts to follow, but here is the summary as well showing all Adjusted R^2 for each data set:

And now, enjoy this inundation of charts [in same order as chart above. GF% first, then CF%]