Save-Equivalent Innings
November 29, 2004
How important are those innings pitched by closers? Take a look.
Tangotiger invented a statistic called Leveraged Index (or Crucial Innings), which basically rates the relative importance of a pitcher’s innings. Tango developed his Leveraged Index by assessing the potential game impact of each relief pitcher’s appearance using a method called Win Expectancy, which requires play-by-play data.
Bill James attempted to estimate the same thing in Win Shares, but historical play-by-play data doesn’t exist. So Bill created an estimate by weighting the number of save opportunities and holds that each pitcher had. He called them Save-Equivalent Innings.
I’ve actually changed Bill’s formula a bit, because I think the original formula overstated the impact of saves and holds in today’s save-happy environment. The specific formula I use is 1.5 times Saves plus .5 times Holds. This creates Save-Equivalent Innings. I then go ahead and create my own Index via this formula:
(IP + Save-Equivalent Innings)/IP
The formula maxes out at 1.9, which is also consistent with James’ approach and Tango’s general findings. Here are the links to the sortable Save-Equivalent Index listings:
American League
National League
Studes,
Interesting list. I’ll have to compare it to mine at some point.
***
I see no reason to max anything out. This is one of those Bill James fudge factors that has no good basis. While Bruce Sutter’s career LI is 1.9, he had plenty of seasons where his LI was over 2. I think the highest LI I found for any reliever was 2.3.
Note also that relievers of today have an easier chance to get higher LIs, simply because they pitch in an environment where scoring is higher. It’s tougher to hold onto a 2-run lead in 1994 than in 1968, and so, there’s a bigger potential swing.
I think Percival from 99-02 was at 2.17 IIRC, using the 5 RPG environment, and was at 1.90 using the 4.3 (and wrong) RPG environment.
If Bill James insists on an artifical fudge level because his equation is a rough approximation, then he should set this fudge level based on the run environment.
Posted by tangotiger on 11/30 at 08:55 AM
Page 1 of 1 pages
