OK, I’ll get back to the Greatest Pitcher’s Duels tomorrow or the next day, but I wanted to take off today to study a couple of other things that I can study with this wonderful data base that Isaac created for me, which has all the game lines for pitchers in the Retrosheet files (Thank you, Retrosheet.)

            It is universally known and often observed that modern pitchers do not stay in the game as long as they once did.   It was once common for pitchers to pitch complete games—and quite common for pitchers to stay in the game, if necessary, beyond the 9^th inning.  This is no longer common.

            However, there is another side to this issue which is, as far as I know, totally unexamined and undocumented.    It has long seemed to me that, as long outings have disappeared, so have short outings.   I spend a lot of time going through Retrosheet box scores, looking at games from decades past.   With what seems to me stunning regularity, you find these old games in which the starting pitcher gives up two, three runs in the second inning—and comes out of the game.

            The way they thought about it was different from the way we think about it.  The way they thought was, “if this pitcher doesn’t have his good stuff today, we’ve got to get him out and get somebody into the game who does.”   This would happen not only to second-line pitchers, but even to good pitchers—more often to second-line pitchers, of course, but not exclusively.   It would happen more often in a big game.   If it was a big game, in the way that 1950s, 1960s managers thought about it, then we really need to get the pitcher out of there quickly if he doesn’t have his good stuff.

            This was part and parcel of the irregular usage patterns of baseball in years past—pitchers switching between the bullpen and the starting staff, coming into the game off-rotation, starting on long rest or short rest.   Another part of it—I have thought, but never documented and never seen documented—was that pitchers also got jerked out of the game very early in some games.

            OK, is that true?

            It’s much more true than I would have guessed.   Since the 1950s, the number of “unexplained early departures” has decreased in every decade—and has decreased, start to finish, by almost 80%.

            I defined an “unexplained early departure” as “any game in which the starting pitcher is taken out of the game before completing four innings and before allowing five runs, but not including games in which the starting pitcher is taken out before allowing any runs.”   Earned runs or un-earned; if the pitcher has given up five runs, then we can understand his being taken out of the game.  We didn’t include games in which the pitcher gave up no runs, because those must be injury games or special situations of some kind, rather than the events we are looking for.

            In the 1950s we have data for 18,362 games (9,181 games, but both sides.)    In 2,072 of those games, the starting pitcher came out of the game having allowed less than five runs and having thrown less than four innings.   That’s 11.3%.    In the 1960s, this went down to 9.5%:

            By the 1980s, it had dropped to 5.7%:

            And by the last decade, to 2.4%:

            Not to hype the discovery, but. . .it seems like a pretty important thing to know.   People talk a lot about the pitch limits, about pitchers not staying in to finish the game the way they did years ago, etc.   It seems to me that the real situation is a little different than the way people imagine it to be.    It is not merely that pitchers come out earlier, but also that they stay in the game now, in some situations where, in years past, they would come out.   Not saying that the two sides of the sword are the same or that they balance, but. . .it doesn’t just cut one way.

            I looked at the data in a second way, to make certain that I wasn’t accidentally creating the pattern by the way that I defined an early exit.   I made up an “exit score”, which is this:

            One point for each out recorded by the pitcher, plus

            Six points for each run allowed, plus

            the Square of innings pitched. 

            So if a pitcher pitches 7 innings and allows 3 runs, that would be an exit score of 88—21 for the 21 outs recorded, plus 18 for the 3 runs allowed, plus 49 because 49 is the square of 7.   If it was 7 1/3, I’d add 53.78, as 7 1/3 squared is 53.78.   That’s about where most pitchers come out of the game—around 85, 90.   Sometimes, of course, a little earlier, sometimes later, but somewhere in that range.

            When a pitcher comes out of the game with an exit score less than 30, that’s a very early exit.    In this study, I counted all games in which the pitcher left with an exit score of less than 30.   This shows the same pattern as before.   The percentage of very early exits has gone down in every decade within the range of the study, and, start to finish, almost 80% of the very early exits have disappeared:

While I was doing that, I decided that I might as well look as well at the number of “Hard Work Exits”, defining a “Hard Work Exit” as any game with an exit score greater than 125.   This shows what you would expect it to show:

            The number of long, arm-destroying outings has also declined in every decade except the 1970s, and the overall decline is about 97%.

            I also did one more test.   For each decade, I figured the average of the 100 MOST abusive outings by a starting pitcher.  This is not exactly standard, since the number of outings has increased; if we were really pursuing a conclusion here, rather than illustrating what is already known, we would have to use the average of the top 1% or something.   But I just used the average of the top 100, the hundred “hardest” outings for pitchers.  This is the data:

            But we knew that anyway.   Let me try again to explain the significance of what we did not know.   People think that modern managers are only concerned, and are over-concerned, with not over-working their starting pitchers.   That’s not exactly true.   Rather, in the modern world we believe in regular workloads.   Whereas managers in the fifties and sixties believed that some days pitchers just didn’t “have it”, and jerked the pitcher out of the game quickly when they thought he didn’t have it, we believe now that you get maximum production and minimal risk of injury with a  regular, predictable workload.   And, really, this is progress, because the whole thing about the pitcher just “not having it today” is mostly just nonsense.   If you give up a couple of runs in the second inning, that doesn’t mean you don’t have it today; it just means you gave up a couple of runs in the second inning.   It is a mistake to let that kind of thing pull you off course from what you are trying to accomplish.

            OK, my other research issue here is, by contrast, just idle curiosity.   For many years I have wondered what was the most common pitcher’s box score line.    You know those lines that pitchers have in the box score, like this:

            For years I have wondered. . .how many different lines are there?  Does each “line” get used every year?   Do pitchers repeat lines?    Since I have this data, I can now examine those kind of questions. . .not that it matters a whit to anybody, but I was just curious.

            These are the most common box score pitcher’s lines, by decade, within the data that we have:

            There are, it turns out, a very, very large number of pitcher’s box score lines that occur in any season, so that if you look at one season by itself, you will find that only a fairly small percentage of the lines are duplicated at all.   Even the lines that occur most often—those above—account for less that 1/1,000^th of all game lines, so that the “most common” lines are not typical lines. 

            Taking the six decades as a whole, these are the 22 most common box score lines, and the number of times that each has occurred, and the won-lost record when it has occurred:

            The most common box score line that usually results in a loss is  6   7  3  3  2  3.  That’s occurred 62 times in our data, with a won-lost record of 18-24.

            In the 1950s, with 18,362 game lines in our data, there were 13,581 different pitcher’s lines that occurred.   40% of game lines were lines occurred more than once; 60% occurred only once.

            As you have more games, of course, you have more lines that are “hit” more than once, so the percentage of “unique lines” has gone down every decade since.   Here, let me chart the data:

The percentage of multiple-use pitcher’s lines has gone up, I think, not only because there are more games, thus more opportunities to step on a line that has already been used, but also because of the phenomenon noted in the first half of this article, which is the disappearance on both ends of the “extreme games”.   Pitchers don’t pitch 11 innings in a game anymore—and they don’t come out after two.   This tends to concentrate more games in the middle.   But still, even in modern baseball lines, 37% of the box scores lines for pitchers are lines that occur only once during the decade.

            In the years 1952-2009 inclusive (as much as we have in the data), there are 81,562 distinct pitcher’s box score lines which occur.   Of these, 50,626 occur only once in the data, and those account for 22.8% of all games.

            If any of this is on Jeopardy! tomorrow, I’ll be expecting a cut.