Not Your Typical Pitching Stats

by my calculations

nkSLG would more likely be lower for a “pitch-to-contact” pitcher? The leaderboard has pitchers who throw up the junk and let the batters make weak contact whereas the two trailers are considered pretty good strikeout pitchers.

Are there year-to-year consistencies and regression for nkSLG similar to BABIP?

logan kensing 2.4 (Rauch lands at .95, 8th worst)

I think Livan checks in at about the same, IIRC

I may have been confused too much by this season’s spread. It still seems like the three measures (rv100, FIP, tERA) should be fairly parallel to each other.

See this comment from Harry below.

Every batted ball outcome (BB, HR, K, GIDP, 1B, etc.) has a linear weight assigned to it. We use these for all types of things. For a pitcher, all linear weights combined corresponds to ERC (composite ERA, which is putting a pitcher’s events together in a context-free manner). However, you can also define expected linear weights by count, not just outcome. By definition, the LW of a 0,0 count is 0 (exactly average). At 0,2 the LW is significantly negative, but not quite as negative as a full out. At 3,0 the LW is significantly positive, but not quite as positive as a BB. On a pitch by pitch basis, you assign the change in LW to the pitch. If the ball’s put in play you also assign the change in LW to the pitch based on the count and end result. For example, a single on an 0,2 count costs more to the pitcher than .5 runs, which is the usual amount, because an 0,2 count has a negative LW, say -.2 runs, which isn’t quite the full -.3 runs for a K. That single costs .7 runs to the pitcher.

Of course, if you’re looking at all of a pitcher’s pitches together, you don’t have to go pitch by pitch, you just care about the final result, because of the commutative property. But if you want to judge pitchers by their individual pitches, you add up the individual run value of all curveball or all fastballs or whatever.

I suppose you could do the same thing using run expectancy instead of linear weights, too, where the value of a single or going from one count to another changes based on the base-out state. Or ever with win probability.

For a pitcher, if we’re looking at all of his pitches, the linear weights approach is directly analogous to ERC. And the run expectancy approach is analogous to RA. Well for full innings at least. If a pitcher comes in late or leaves an innings early, run expectancy is different from the way actual runs allowed are credited to pitchers. But over full innings, the sum of RE will always equal actual runs scored and be discrete.