keyboard_arrow_uptop

Starting today, we will be periodically running some of the best content from the new, super-charged Baseball Prospectus archives. Those new to BP may be reading this content for the first time. Long-time readers can rekindle old debates. We begin today with Keith Woolner‘s look at the conversion of balls in play into outs, from 2002. To do your own mining, go to BP’s Search function. To request a specific article from the archives, e-mail jkeri@baseballprospectus.com.

This week’s question comes from Robert Shore, who asks:


Like many people, I was mightily impressed by Voros McCracken’s work, which
strongly suggested that pitchers have essentially no effect on the conversion of
balls in play to outs. It occurred to me to wonder about the converse question.
Are some batters better than others in converting balls in play to base hits?


This is a question that has come up several times in the past year, since
McCracken published his results,
and in fact was addressed a bit in a
BP Mailbag article.

To restate, McCracken’s work shows that, once you remove strikeouts, walks, and
home runs (events that the pitcher is solely responsible for), and look only at
the results of plate appearances where the defense participates (a ball in
play), the range of differences between pitchers is small enough to be
negligible in most cases. The difference between the likelihood of a hit off of
Pedro Martinez and or Dave Mlicki given that the batter doesn’t strike out,
walk, or hit a home run, is essentially the same. There’s some
evidence that
pitchers do have minor differentiation in ability to prevent hits on balls in
play
, but you have to look over several years worth of data to see the trend.

Does the same result hold for batters? Are they equally likely to get a hit once
you remove the “Three True Outcomes”?

To answer the question, I turned again to the
free Lahman database available
from www.baseball1.com. I looked for batters
who had 300 or more at bats in
consecutive seasons, to see if their rate of hits on balls in play (ball-in-play
hit rate, or what I’ll label BIPr) was consistent from year to year. Note that
this will exclude hitters who weren’t more or less regulars in two straight
seasons, whether due to injury, lack of opportunity or ability, or retirement.
In particular, we’d expect that it leaves us with the better hitters in the
league–and BIPrs that are higher than the league average–if an ability exists.
I further broke down different eras in baseball history, to see how that ability
may have changed over time. Data is through 2000.

In the post-1893 era (the introduction of the 60’6″ pitching distance), the
year-to-year correlation of BIPr for regular batters was +0.501. Recalling that
0 indicates no relationship, and +1.00 indicates a perfect linear relationship
between the two sets of values, we can see that the ability for batters to
create hits on balls in play is much more persistent from year to year than the
near-zero correlation in a pitcher’s BIPr.

I split up the eras of baseball into (a) pre-1893, (b) 1893-1920, (c) 1921-2000,
(d) 1921-1950, and (e) 1951-2000. Note that the (d) and (e) are a further
breakdown of (c), as I wanted to isolate the lively ball, pre-WWII game from the
1950-present. I also looked at the 1960’s expanded strike zone years, and the
subsequent post-expansion period.


Years     Corr Median
1951-2000 0.45  0.285
1921-2000 0.46  0.288
1921-1950 0.44  0.297
1893-1920 0.61  0.285
pre-1893  0.48  0.282

    Years Corr Median
1969-2000 0.46  0.287
1963-1968 0.41  0.276


There are several interesting things to note here: first, a batter’s ability to
get hits on balls in play has been a consistent part of the professional game
since its inception, with a remarkable cross-era consistency in year-to-year
correlation. The one outlier is the 1893-1920 game, where the batter exerted
greater influence on the outcome, resulting in a correlation of +0.61. I thought
perhaps there was one stretch of time that was skewing the results, but
splitting out the 1893-1920 era in the NL-only high -offense years 1893-1900
from the two-league “dead ball” years 1901-1920 yielded almost identical
correlations. It seems that batters demonstrated more year to year consistency
in getting a hit on a ball in play in those years.

Another fascinating observation is that the median BIPr is remarkably consistent
across eras. The median 1950-2000 BIPr is identical to the 19th century BIPr
before the introduction of the modern pitching mound. Again, there are
interesting exceptions, namely the 1921-1950 era, which had an unusually high
.297 median BIPr, and the 1963-68 era, the low point at .276. These, perhaps
unsurprisingly, correspond to a high-offense era of the ’20s and ’30s, and the
expanded strike zone and record-low offense of the ’60s. Indeed, the 1921-1930
era had a .309 BIPr, the 1930’s had a .299 BIPr, and the 1940’s fell to a more
typical .282 BIPr. The regular batters of the ’20s and ’30s not only started the
love affair with the home run, but they also tore into opposing defenses at
rates unseen in any other time.

These results may be easier to visualize through a chart. The following two
charts show the year-to-year relationship in BIPr for regular batters. The first
is for 1921-99, the latter for 1893-1920. Two things to look for are that the
scatter plots are not symmetrical, but are somewhat flattened, like an oval. If
they were round, it would indicate that a value in year 1 had no predictive
value for the value in year 2, i.e., there’s no ability being shown, which is
very close to what we see for pitchers’ BIPr. The degree to which the point form
a straight line (along the center of the oval), indicates the strength of the
relationship, or, how much the value in one year helps predict the value in the
other. The red dots form a “flatter” shape than the blue, reflecting the higher
correlation mentioned above for the 1893-1920 era.