Re-Working Hitter Game Scores
You may remember. . . or may not; I don’t talk about it a lot. . . .but you may remember that two or three years ago I finally hit upon a way to figure a "Game Score" for a hitter, The mistake that I had been making, which prevented me from being able to make this work for 30-some years after I developed the pitcher’s Game Score, was trying to center hitter’s Game Scores at 50. Doing that contradicts the essential nature of a hitter’s production, which is not symmetrical by games.
Basically, a Hitter’s Game Score as I developed it at that time, was:
15
+ 10 for each run that he created,
+5 for a run scored,
+5 for an RBI
Conceptually simple, and it works. To make it work, I had to invent a way to estimate Runs Created by a Hitter in a game, but that was worth doing for its own reasons. An average game by a hitter is at 25.0, which is where it has to be, and virtually all games are between zero and a hundred.
However, an issue with this structure is that a hitter starts out at 15 and, in an "average" game, works his way up to 25. Baseball teams historically score 4.50 runs per game, or something very, very close to that. The average hitter in a 9-man offense creates 0.50 runs per game, which also means that the average hitter SCORES 0.50 runs per game, and drives in most of those, so an average hitter in an average game will come in at 15 + 5 + 2.5 + 2.5. More or less.
But this is different than the structure of PITCHER game scores. In a pitcher’s Game Score the pitcher starts out at 50, and goes up or down depending on the good or harm that he does for his team. On average you start out at 50; you end up at 50. For a hitter, you start out at 15 and end up at 25.
Neither approach could exactly be described as "right" or "wrong"; it is just what you are happy with. The "right" answer is that which most accurately answers the question "How good a game was this, for the hitter?" Sometimes I don’t like the pitcher’s system, starting in the middle. It implies that if you don’t do anything, you’re average. A starting pitcher starts the game, walks the first three hitters, gives up a grand slam and is taken out of the game; his Game Score is 29. Doesn’t seem right. He hasn’t done ANYTHING to help his team win; it seems like that should be a zero, or near-zero.
But that’s the exceptional case; mostly, starting in the middle works OK for that purpose. The problem on the other end. . .an average Game Score for a hitter is designed to be 25, but it is actually 24.3, because more than nine players are used in an average game. A defensive replacement who doesn’t have an at bat in the game gets a Game Score of 15. Maybe it would be better if he was at 25, since 25 is the neutral score here?
This bothers me in a case like Jerry Lynch. Jerry Lynch was a very good left-handed hitter in the National League on my childhood, but his defensive limitations limited him to the role of pinch hitter/reserve outfielder. In 1958, when he had his career high in at bats (420), he hit .312 with power. In 1961, when Frank Robinson was the NL MVP and led the league in OPS and OPS+, Lynch batted only 181 times but drove in 50 runs, registering a higher OPS and higher OPS+ than Robinson did. Lynch that year was actually mentioned in MVP voting, with 30% of regular playing time. In 1964, batting just 297 times, he hit 16 homers and drove in 66 runs.
But because he often batted only once a game, Lynch’s average Game Score, for his career, was only 22.1, which is in the Mark Belanger range. That bothers me. So I am thinking about changing the Game Score method to a system that starts out at 25, and ends up at 25 for an average hitter in an average game. Have not decided whether to push the button on this or not, but here is the system that I am thinking of going to. A hitter’s Game Score is
25
  + 8 for each run that he created,
+3 for a run scored,
+3 for an RBI
  +1 for Reaching On An Error
MINUS 2.5 for each out that the hitter is responsible for
The practical effects of the change are as follows:
(1) It is a little bit more complicated than the original formula. That is a negative, but then real life is complicated,
(2) It changes from a system based 50/50 on "real runs" (Runs Scored and RBI) and "expected runs" (Runs Created) to a system that is 57% based on Runs Created and 43% on Runs Scored and RBI. I’m OK with that change, don’t regard it as either a positive or a negative.
(3) It reduces very slightly the number of "out of range" Game Scores, which are Game Scores less than zero or greater than 100, and also creates a slightly more even balance between them. Those things are both positives, but very small positives since less than one game in 2,500 is out of range.
(4) It gives the hitter a little bit of credit for reaching on an error, which I like.
(5) It changes from a system that goes up as at bats go up to a system that in essence filters out replacement level, going up only when the batter’s contribution is on balance positive.
In the original system, outs are accounted for because making outs reduces a batter’s runs created. In this system that is still true, but there is an additional penalty for the out itself, which offsets the fact that the starting point is 10 points higher.
I experimented with/developed the alternative system using a spreadsheet of batter games logs which runs from 1939 to 1993 and includes 324,745 batter games representing the careers of more than 200 hitters. (I have several files like that. I love messing around with batter Game Logs.) In the original system there were 135 "out of range" games, of which 134 were OVER 100 and one was less than zero. In this system there are 126 out of range game scores, of which only 10 are OVER 100, and 116 are less than zero. So. . . a lower total (126 to 135) and better balance (116 – 10 versus 134 – 1.)
In practice, 99% of the differences between the systems are small. In the original system, the ten best games in this data are identified as:
Joe Adcock, 7-31-1954 (4 homers and a double)
Rocky Colavito, 6-10-1959 (4 homers)
Willie Stargell, 5-22-1968 (single, double, three homers and 7 RBI)
Reggie Jackson, 6-14-1969 (10 RBI, the "flat bat" game in Fenway)
Willie Stargell, 8-1-1970 (3 doubles and 2 homers)
Gus Bell, 5-29-1956 (5-for-5, 3 homers and 7 RBI)
Steve Garvey, 8-28-1977 (5-for-5, 3 doubles and 2 homers)
Smoky Burgess, 7-29-1955 (3 homers, 9 RBI)
Rico Carty, 5-31-1970 (4-for-4, 3 homers, 6 RBI and a walk)
Nate Colbert, 8-1-1971 (3 homers, 8 RBI, part of his famous double-header)
In the new system, 9 of the 10 best hitter’s games are the same, with the top 3 in the same order. The Gus Bell game nudges up two spots, with an Ernie Banks game (8-4-1955) pushing Smoky Burgess off the list.
On the other end of the scale there is more change. In the old system, the only game which has a negative Game Score is the game in which Joe Torre went 0-for-4 and grounded into four double plays (7-21-1975). In the proposed new system, that is lifted out of the bottom spot up to third-worst, ahead of a game by Manny Mota (5-24-1973) in which Mota went 0-for-9 with a strikeout and a GIDP, and a game by Luis Aparicio (8-6-1959) in which Aparicio went 0-for-8 with a double play ball AND a caught stealing after he reached on an error.
What does it take to make a negative Game Score, in the alternate method? There are three games in this data set in which a hitter went 0-for-6, struck out 3 times and grounded into a double play. That combination makes a Game Score of negative 0.1.
I haven’t decided whether to make the change or not, and I guess I am asking for your feedback without really expecting to get anything useful. Since none of you make any regular use of the system, you probably don’t understand the ins and outs of it well enough to have an educated opinion, but you might see an advantage or disadvantage somewhere that I didn’t see, or have some reason for a preference one way or another. Thanks for reading, and thanks for your insight.