Before we can go any further with the Runs Saved project, I need to take a break and attend to some other things.  

            We are nearing the end of the Runs Saved Against Zero project, and I was hoping to complete it this week.   However, late in the project, I have acquired an accumulation of small problems, three specifically:

1)      The system assumes that first basemen are initiating as many defensive plays as third basemen.   As was pointed out by Mr. Saeger, this is not accurate, and I need to do something to correct that.   Evaluating first basemen is difficult. 

2)     In the numbers that I just gave you, about second basemen and I think also first basemen, I’m not sure, I had a calculation error, if not two or three calculation errors.   In my explanation for range for infielders, I assigned 10% of the credit for defensive plays made (range) to pitchers, leaving 90% for fielders.  But in doing the actual calculations, I forgot to do this, forgot the remove the 10% that was put on a siderail.   It’s not a big deal; Bill Mazeroski is credited with 19 Runs Saved by Range in both 1960 and 1961, 61 and 59 Runs Saved altogether; if we take away 10% of the Runs Saved by Range he drops to 59 and 57 Runs Saved.   Still, it needs to be fixed. . .and I made the same error in figuring the third basemen and shortstops, which I was going to publish today and tomorrow, so I need to correct those files as well.  

3)     When we get to outfielders, which is work that I have done but not yet explained to you, the system is returning unrealistically low numbers for outfielders.   I’ll need to make some sort of adjustment to transfer more credit for Runs Saved to outfielders. 

None of these three things is actually difficult to correct.   The first base/third base thing is dead simple to adjust for.  The calculation error on infielders is just a matter of re-doing that calculation work, which might take me a day, and I have several ideas about how to adjust the Runs Saved assigned to outfielders consistent with the logic of the system; I just need to spend a couple of days working with the data and sorting it out. 

However, while none of these things is actually difficult, I have been putting aside a lot of other responsibilities to focus on getting this done, and I just can’t do that any longer.  I’ve got some other things that I have to pay attention to.   I’m going to put the Runs Saved project aside for a few days to get some other things done.  

Let me explain about the outfielders, although this is repeating something that I have written earlier but not yet published.   When I got to the outfielders, I figured Willie Mays’ Runs Saved for the 1961 season at 33 runs—and I may have forgotten to take the 10% off for diverted range; I’m not sure, but let’s say 33 runs.   This is far lower than the numbers credited to the best infielders, of course, which is one problem, but here’s a bigger one.  Using modern methods of fielding analysis, Mookie Betts in 2016 came in at +32 runs—32 runs better than an average right fielder.  

It is not technically impossible for both of these estimates to be correct.   Mays in 1961 could be at 33 Runs Saved, but Willie Mays is not the average center fielder.  Maybe the average center fielder is at 18 Runs, the average Right Fielder is at 12, Mookie is at 44; it all works.   It’s technically possible, but it is awkward and improbable.  

We don’t HAVE to agree with modern systems of analysis on every point.   Sometimes we’ll show a player as a good defensive player, they’ll show him as not good; that’s fine.  What we’re measuring is a little bit different, details of the definition, and maybe they’re wrong sometimes; we don’t have to agree. 

We don’t have to agree, but I also don’t want to be in the position of saying that everybody’s work is worthless except mine.   I know that there is great value in the work that has been done by others in this area; if my system has some value, it must also have some degree of agreement, some consistency, with their work.   The Mookie and Mays thing, although I need to know more about it, seems to exceed the boundaries of acceptable off-the-farm-ness. 

There is a fourth and a fifth thing that I MAY do in finishing this system.   The fourth thing is, I may completely revise the way that I incorporate error avoidance into the system.   If I do that, that might take another week, plus of course I will have to revise again all of the "show calculations" that I have done so far.  Also, I may create a "reconciliation system" for those years when the modern data is available; in other words, if my system shows that a player is at +10 runs and the modern data says he is at +20, maybe I’ll revise my estimate to +15.   I don’t know whether I am going to do that or not; I have had that in mind all along, since I created this system in my head, but I don’t know whether I should do it or not, and can’t make a judgment about that until I get to where the data is and I can look at what we have.  

In any case, I’m not two days away from finishing the project anymore, so I need to put it aside and tend to some of the other stuff that I need to do.   Thanks for reading.

Measured Intimidation

       Many years ago, before the national publication of the Abstracts, I was watching a baseball game in which a player hit a home run, and on his next at bat was hit by a pitch.   The announcer said something about that being a common thing that happened, and I had the thought that "I can’t check out THAT, because we don’t have any access to game accounts, but I can check out how often a player hits a home run and is hit by a pitch in the same game."   I did a study of that issue based on box scores in the stacks of Sporting News that I always had setting around, and the study did find that player’s frequency of being hit by a pitch increased significantly when they had a home run in the game.   Measured Intimidation. 

            But that study probably dealt with no more than a thousand games, I am guessing.  I have only very vague memories of it, doubt that I ever published anything about it.  But doing research was so hard, in those days, that all of our research was poor by modern standards.  We didn’t have computers.   We didn’t have access to game accounts.   You really could not do what would now be considered good research.  

            In a game I was watching on Sunday, after a player had hit a home run, the pitcher came in high and tight; didn’t hit the guy, but backed him off the plate.  It suddenly occurred to me that I could do that same study now a thousand times better in a just a couple of hours, by using the file of Game Logs that I have created, which now includes more than 300,000 batter games.

            The games in the file range from 1929 to 2018, but are mostly concentrated in the years 1940 to 2000, with the highest concentration in the 1960s, where the project started.   I have only one player from the 1920s, Earl Averill in 1929, so I didn’t even look at that data.  

            The data from the 1930s is not consistent with the thesis that hitting a home run leads to an increased likelihood of being hit with a pitch, but I only have about 8,000 batter games from the 1930s in the data, so let’s just ignore that.

            From the 1940s I have 19,949 batter game logs—2,086 games in which a player did hit a home run, and 17,863 games in which he did not.   In games in which a player did not hit a home run, there were 279 HBP in 73,558 plate appearances, or one every 264 plate appearances. 

            In games in which a player did hit a home run, the players had a total of 9,449 plate appearances, and hit 2,276 home runs.   That means they had 7,173 plate appearances in which they had hit a home run in the game, but not this time up.  They were hit by pitches 37 times, or one for every 194 plate appearances.  The frequency of HBP increased by 26-27% in games in which the player hit a home run. 

            In the 1950s we have data from 40,710 games—4,418 games in which a player did hit a home run, and 36,292 in which he did not.   In games in which the player did not hit a home run, he was hit by a pitch 732 times in 141,646 plate appearances, or one for every 194 plate appearances.  In games in which a player DID hit a home run, he was hit by a pitch 102 times in 14,646 non-home run plate appearances, or one for every 144.   The frequency of a HBP increased when the player had hit a home run in the game by about the same percentage—about 26-27%. 

            Of course, this approach under-states the underlying effect, since a Home Run cannot "cause" a HBP in a plate appearance which occurs before the home run.  Assuming for the moment that this is a real effect, it seems likely that one-half of the "OPAIG"—the other plate appearances in the game—occurred before the home run, and one-half occurred after the home run, and that there was no increase in the HBP rates before the home run.  That would mean that the actual effect was twice the measured effect, and, if we could focus on the exact one at bat following the home run, probably higher than that in that one plate appearance. 

            It also might not be a real effect; it could be also that, comparing Mickey Mantle and Nellie Fox, that Mantle is simply hit more often by pitches, thus, when you focus on the home run games, you get more Mickey Mantle games than Nellie Fox games, so you get an increased number of HBP.   Mickey Mantle isn’t actually in the study; I’m just using him to illustrate the possible bias. 

            Anyway, moving on to the 1960s.  In the 1960s the players in the "non-Home Run" games had a total of 311,076 plate appearances, and 1,975 hit batsmen.   That’s one hit batsman per 158 plate appearances. 

            In games in which a player DID have a home run, they had 30,128 other plate appearances.  In those PA, they had 217 hit batsmen, or one for every 139 plate appearances.   The HBP frequency increased by a smaller amount—about 12%--but it still increased. 

            In the 1970s, players who did not have a home run in the game had a total of 289,805 plate appearances in the games which are in my file.  In those plate appearances they were hit by pitches 1,707 times, or once for every 170 plate appearances.

            In games in which a player DID hit a home run, the players had 27,033 other plate appearances.   They were hit by pitches 170 times, or once for every 159 plate appearances.   The frequency of HBP increased in the Home Run Games by only about 7%--but it still increased.  

            In the 1980s, players in my data had 148,267 plate appearances in games in which they did not hit a home run.   They were hit by pitches 852 times, or once for every 174 plate appearances.  

            In games in which a player DID hit a home run, he had 16,116 other plate appearances.   He was hit by a pitch 97 times, or once for every 166 plate appearances.   In other words, the rate of HBP occurrence increased by a yet smaller amount—a little less than 5%--but it still increased.  

            In the 1990s the players in my data had 56,697 plate appearances in games in which they did not hit a home run.   They were hit by pitches 353 times in those games, or once for every 161 plate appearances.  

            In games in which the player DID hit a home run, he had 7,235 other plate appearances.   He was hit by a pitch 45 times, or once for every 161 plate appearances.   The rate of HBP within my data—which is limited, just 16,000 games and 64,000 plate appearances, more or less—but the rate did not increase in the 1990s. 

            For the 21^st century, I don’t have enough data to run the numbers.   But the data that I do have narrates a fairly clear and convincing story, subject to future and better research.   In the 1940s and 1950s, when intimidation of the batters was an accepted and common practice of the game, a batter’s likelihood of being hit with a pitch increased by a little more than 50% if he had hit a home run earlier in the game.  

            From about 1960 until about 1990, the use of intimidation as a pitcher’s weapon became gradually less accepted, and the rate of increase in Hit Batsmen following a home run declined to about 25% in the 1960s, to about 15% in the 1970s, and to a little less than 10% in the 1980s.   In the 1990s, the practice of retaliating against a player who had homered earlier in the game probably was or may have been completely eliminated. 

            Thanks for reading.