Changes in home run rates during the Retrosheet yearsby Tom M. Tango
February 15, 2008
I make no bones about it. Retrosheet is a monumental achievement, and equally effective in diminishing worker productivity in corporate America.
The Colorado Rockies did two things in 1993: They were part of the expansion of the league, and they brought with them a park that decidedly favored hitters. The expectation is that home runs would explode. From 1988 through to 1992 (pre-Rockies, pre-expansion), 593,975 times a batter had a plate appearance in which he put bat to ball (i.e., all PA, excluding strikeouts, walks, hit batters, interference and sacrifice bunts), and 16,001 went for home runs. That's 2.7 percent of all contact PA going for homers.
From 1993 through to 1997, we had 584,918 contact PA (a virtual match to the previous time period), with 21,019 home runs, for a rate of 3.6 percent. That's 33 percent more home runs compared to 1988-1992. Furthermore, the year-by-year rates in each group were fairly stable. It was a sudden jump from one time period to another, occurring right at 1993, perhaps 1994. It was sudden, it was quick, it was dramatic. And by definition, it was a one-time shift.
What can cause changes in output? There are five groups of events:
- Player-controlled events (drugs, conditioning)
- Umpire-controlled events (strike zone)
- Team-controlled events (parks, ballplayers)
- League-controlled events (expansion, configuration of ball, bat)
- Nature-controlled events (climate, wind, weather)
Hitters, pitchers, and parks
Here are the HR rates from 1982 through to 1998:
1982 2.7% 1983 2.7% 1984 2.7% 1985 3.0% 1986 3.2% 1987 3.7% 1988 2.7% 1989 2.6% 1990 2.8% 1991 2.9% 1992 2.5% 1993 3.1% 1994 3.7% 1994 3.7% 1995 3.6% 1996 3.9% 1997 3.7% 1998 3.7%
You will note that 1987 stands out, both in terms of the huge jump from 1986, and an even more dramatic fall to 1988. The difference between 1987 and the 1993-1998 period is that what happened in 1987 was a sudden jump, which reverted back, while 1993 was a sudden jump and sustained itself from 1994 onward.
So, when trying to explain 1987, we can discard the first group of events (player controlled). It's implausible that hitters, as a group, would be able to do anything in the offseason of 1986 to give them a decided advantage over pitchers. Anything they would have done to their minds, bodies and souls would have resulted in a more gradual shift over a period of years. We can also discard the third group, since again, even if teams had decided that the Vince Colemans and Willie Wilsons and Gary Pettises of the world were not for them, this wouldn't happen suddenly in one offseason.
That leaves us with the likely culprits of: mother nature, juicing the ball, and the umpires calling a different strike zone. Each of these possibilities satisfies the conditions that something could happen overnight, and it could be undone a year later.
When it comes to 1993, these three remain as plausible scenarios (with mother nature likely being a less reasonable answer; comparisons to minor leagues at the time would be instructive). However, now we also have to consider the parks and expansion. Can we figure out how much effect those two had? Remember my ode to Retrosheet?
What if we reconstruct our league totals, but we throw out all the expansion hitters and expansion pitchers? While we're at it, let's throw out all the expansion parks. What we'll be left with in 1993 is the same group of players from 1992, and the same parks. But, that's still not enough, since not each hitter will face the same pitcher in both years, or for the same number of PA, or even in the same parks.
Let's go back to 1987 for a second. In that season, Tim Raines hit two home runs against Doug Drabek at Olympic Stadium on six contact PA. In 1988, he hit zero home runs on two contact PA. For Raines, 1987 was a good year, but it was a bad one for Drabek, in terms of this tiny sample. Pro-rating Raines' 1987 down to two PA, he had 0.67 HR in 1987 and 0 in 1988, on two PA. What we did here was look for the exact same hitter and the exact same pitcher in the exact same park with the exact same number of (pro-rated) PA, over back-to-back seasons. We are in effect controlling for the players and the parks, leaving only one parameter (the year) as variable.
If we repeat this step for all hitters, pitchers, and parks in 1987/1988, we end up with 29,212 matching PA, which is roughly 25 percent of the entire population of plate appearances in those seasons. While the other 90,000 or so discarded PA are some combination of non-matching hitters/pitchers/parks between 1987 and 1988, these 29,212 PA are an exact match of hitters/pitchers/parks in those two years. The result here will tell us exactly the effect between those two seasons.
In 1987, the matching combinations had 4.1 percent of their contact PA as home runs, while it was 2.9 percent in 1988. If we look at the non-matching data for those two seasons, the home run rate was 3.6 percent and 2.6 percent, respectively. We can see, therefore, that the two groups mirrored themselves: The matching combinations dropped their homer rate by 30 percent, while the non-matching dropped it by 28 percent. We really didn't expect a different answer. It's not like we had a new Coors field to contend with, or a bunch of new hitters and pitchers to throw things off.
The Coors expansion
Ah, but what about 1992/1993? In this case, we had 22,622 matching PA, with 629 homers in 1992 and 744 in 1993. Remember, we are looking only at the same hitters against the same pitchers in the same parks with the same number of PA here. And we see a jump of 18 percent in number of home runs hit. The rest of the league had a jump of 24 percent. So, we can see two things here: 1. There was a huge jump in homers, without considering parks or expansion, at 18 percent 2. The rest of the league jumped even more, at 24 percent, with the extra jump possibly because of parks and/or expansion.
Continuing with 1993/1994, there was a 20 percent jump in homers for the matching combination, with only a 15 percent jump for the non-matching combination. If we combine the two groups of years, 1992/1993 and 1993/1994, we get a 42 percent jump with the matching combination of players, and a 43 percent jump with the non-matching combination of players.
It would seem, therefore, that the league-wide jump in home runs between 1992 and 1994 (43 percent) affected the matching combination and the non-matching to a nearly identical extent. That is, while the matching combination only considered the same players in the same parks in back-to-back years, and the non-matching considered everything else (new players, new parks, etc), the overall effect was the same.
The conclusion would then follow that expansion and parks had very little to do with the jump in homers between 1992 and 1994.
The real culprit
I don't know. But, juicing the ball seems to be a possibility. Bob DuPuy of MLB, using his best legal training, says this:
“We are satisfied that the ball comports with all major league specifications,” DuPuy said. “Beginning in 2000, we have had annual independent testing done by UMass at Lowell, baseball research center, under the direction of Dr. James Sherwood, and those tests have showed full compliance with standards.”
You know he's trying to tell you something without telling you anything by using technical words: specifications, independent, compliance, standards and doctor. Talk about a lawyer pitching his perfect game: five key words in only two sentences! Very impressive. And, he tells you nothing at all, which must have gotten him the lawyer-of-the-day at his local club. Now, consider MLB's own ball tester, whom Mr. DuPuy was kind enough to include in his statement, Dr. James Sherwood:
That’s part of what frustrates Sherwood… “Their testing window is this big,” he says, his hands a foot apart. “I don’t know why it was ever set that wide.” A ball testing at the high end could travel as much as 50 feet farther than one falling on the low end, he says. That’s the difference between a lot of home runs and a whole lot of home runs. Sherwood would love to bring the testing procedures into the modern era. Upstairs, his computerized machines can control a baseball bat with the precision of Barry Bonds. He has air cannons that can fire a ball at 180 mph. But the league doesn’t like change. Sherwood estimates the MLB hasn’t altered ball design since Babe Ruth played. Sherwood says there’s some evidence that firing a baseball at 58 mph may not be fast enough to accurately determine its liveliness. “Has there actually been data on that?” Drane asks. “Yeah,” Sherwood says, “We’re just going to explore looking at the higher speeds and present that to the league. Maybe they’ll change their minds.”
Mr. DuPuy must have known about Dr. Sherwood's thoughts, no? DuPuy is the contact person for MLB, and Sherwood is the contact person at the testing facility. They must have contacted each other at some point, no? Does Mr. DuPuy's statement represent Dr. Sherwood's position? In a pure legalese sense, I suppose it does. But, by the reasonable man test? Not at all.
In 2006 there were 1,454 homers in the “Just Enough” category, which means clearing the fence by approximately 10 feet or less. They are spread fairly smoothly from 0-10 feet of clearance.
Greg allows us to estimate that at 8.7 feet of clearance, 1,265 homers would have cleared the fence. That is, if you can change the composition of the ball to reduce the length of the long flyball by 8.7 feet, you'd end up with 1,265 fewer home runs. In 2006, we had 5,386 home runs on 135,626 PA, for a rate of 4.0 percent. Knocking out 1,265 would bring the homer rate down to 3.0 percent. The average home run rate from 1982-1992 was 2.9 percent. So, all we have to do is reduce the length of homers by around 8.7 feet, and we'll get home run output rates from the 1980s. Is this possible? Quoting the good doctor once more:
The Major League balls are manufactured in Costa Rica and have a compressed cork sphere per the specifications. The Minor League balls are manufactured in China and have a cork center as specified in “1996 Minor League Baseball Proposal”. This cork center is the likely source for the decrease in performance, which results in a comparable Minor League ball hit of 391.8 ft under the same conditions as the Major League balls. Small samples of 1998 MLB baseballs were also tested. The 1998 MLB baseball had a comparable batted-ball distance of 400.5 ft.
And 400.5 feet minus 391.8 feet equals 8.7 feet.
Tom has co-authored a baseball strategies book called The Book - Playing The Percentages In Baseball, which is available at Inside The Book. Readers are encouraged to post comments on the author's blog, or to send him an email.