Bill Mazeroski's Fielding-An Essay on Fielding Statistics
By Clifford Blau
Several weeks ago, Mike Emeigh posted an article to rec.sport.baseball
and to his Web site (http://www.geocities.com/Colosseum/Stadium/8957/index.htm).
In it, he analyzed the value of Bill Mazeroski's fielding. He
calculated
the value of the extra putouts, assists, and double plays made by Maz.
He found that Mazeroski saved the Pirates an average of 57 runs
compared
to the other National League second basemen during the years 1957-68.
(Mr.
Emeigh subsequently expanded the study by two years and produced a
figure
of 51 runs per annum. His most recent revision resulted in a finding of
25 runs per 162 games. I believe Mr. Emeigh has dealt as well as
possible
with the question given the available data. After my original article,
I have summarized some comments and data provided by other SABR members
on this question.)
I realized that 57 runs is a lot of runs, and that if it were true
that
Mazeroski's fielding were that valuable, it would be possible to
corroborate
the fact in other ways. The typical difference between an average
defensive
team and the best is about 100 runs, so if one player were accounting
for
over half that much with his fielding, it should stand out. I will
state
here that I found no certain answers. While I think it unlikely that
anyone's
fielding could be so valuable, I will here present what evidence I
have.
I will make use of comments made during a discussion of Mr. Emeigh's
article
on the SABR e-mailing list as well as other research. Any other facts
that
anyone can contribute to this study are welcome.
The Big Picture
Mr. Emeigh, in making a case for Mazeroski's Hall of Fame candidacy,
stated
that Maz' value was equal to that of an average fielding second baseman
who created 57 runs above average each year. I decided to test this by
looking at the performance of teams who had such players. I looked for
players who created 57 runs above average (for the league, not
position.)
There were very few players who did that for as many as a dozen years.
I identified 14 such players. Leaving Ruth and Gehrig out, since they
were
teammates for much of their careers, I looked at a dozen great hitters.
They collectively created an average of 55 runs above average over
their
best 12 consecutive seasons. Their teams had a collective winning
average
of .565. 57 runs over a 162 game season should produce approximately 6
extra wins, or a winning average of .537 for an otherwise average team.
The Pirates during the years 1957-1968, when Mazeroski was their
regular,
full-time second baseman, had an overall winning average of .517. More
to the point, the teams with the great hitters scored a park-adjusted
50
runs above average, very close to what was produced by their leading
hitters.
The Pirates during 1957-68 allowed just 21 runs below the league
average,
park adjusted. On this basis, it seems unlikely that Mazeroski was
saving
57 runs a year. It should be noted that Rogers Hornsby's teams during
his
prime had only a .499 winning average, and that Stan Musial's Cardinals
scored only 25 runs above average while he was creating 60. Thus, it is
not outside the realm of what we may construe as normal variation that
the other Pirates may have been far enough below average defensively
over
those dozen years to neutralize some of the effect of Mazeroski's
fielding.
Further, if Mazeroski's fielding were so valuable, then pitchers
being
traded to the Pirates would tend to have a lower ERA than they'd had
the
previous year. Likewise, pitchers traded by the Pirates would see
theirs
rise. In fact, a group of 7 starting pitchers traded to the Bucs during
those years saw their adjusted ERAs rise by .07, while the one starter
traded away by them improved his adjusted ERA by .17. On the other
hand,
in 1969, when Maz missed most of the season due to injury, the Pirates
adjusted ERA went from 107 to 97.
A second claim that Mr. Emeigh made is that Mazeroski took away an
average
of 46 hits per season compared to his peers. If this were so, the
Pirates,
all other factors being equal, would have had an opponents' batting
average
.008 below average. Their actual batting average allowed was .004 above
average. Part of this .012 difference could be due to Forbes Field.
It was a slight hitters' park, although it was very tough to hit home
runs
in. To what extent it increased batting average, I do not know, but
obviously
it did. (My subsequent research covering the years 1957-1962 shows that
opposing players' batting averages were .002 higher at Forbes than in
Pirates'
road games.)
Methodological Flaw
Mr. Emeigh gave Mazeroski credit for all plays he made above average,
without
consideration of factors other than Maz' skill that may have influenced
the statistics. He converted the extra outs to runs using Pete Palmer's
Linear Weights formula. To see the error of this, take an extreme
example.
Imagine a pitcher throws a perfect game, consisting of 27 routine
ground
outs to second. If the league average number of chances per game for a
second baseman were 5, the second baseman would get credit for taking
22
hits away from the opposition and making 22 additional outs, yielding a
runs saved value of 16. The pitcher, assuming a league ERA of 4.00,
would
get credit for 4 pitching runs. The other 8 fielders would presumably
be
charged with allowing 16 runs above average, although they had no
chance
to save or allow runs. The point is that while opportunities to make
plays
will not be as absurdly distributed as in this example, they do not
even
out. All chances above average do not represent plays that an average
player
would not have made in the same circumstance. Many factors affect how
many
chances a second baseman has besides his skill, and we cannot make
proper
allowance for all of them.
Significant Factors
The Pirates' pitching staff during the years 1957-68 struck out .45
batters
below average per 9 innings. This obviously meant more chances for the
infield and outfield. Indeed, the Bucs' shortstops over that period
averaged
.4 chances per game more than their peers, their third basemen had an
average
number of chances, and their outfielders collectively had .3 extra
chances
per game. It is not clear from this distribution whether the Pirates'
pitchers
tended to allow an above average number of ground balls. Under the
assumption
that ground ball pitchers allow fewer home runs than average, I
determined
that on a park-adjusted basis, the Pirates allowed .6% more homers than
their peers. Thus, while it appears that some of Mazeroski's extra
chances
were due to there being more batted balls than normal against his team,
we cannot say to what extent he benefited from that.
Stats Inc. tracks the location of all batted balls. Using their
statistics
for 1989, 1990, and 1992, I found that the number of balls hit into the
second baseman's zone varied by as much as 204 between 2 teams in the
same
league in a season. The standard deviation for a season was 48, and
over
the three years, one team had 224 balls below average hit in the second
base zone. Defensive average, which uses different, larger zones, shows
larger differences. Thus, Maz' additional 90 chances per season could
be
entirely due to extra balls being hit near him, rather than to his
skill.
Minor or Unknown Factors
Charles Pavitt concluded in his study of the effect of pitchers on
fielding
statistics that teams with an above average number of innings thrown by
right handers will have an above average number of chances by their
second
basemen. Since the Pirates averaged only 1% more innings by righties
than
the rest of the league, this is not a significant factor in evaluating
Maz' fielding statistics. Likewise, teams that allow a lot of
baserunners
will make more double plays on average. The Pirates during the years
1957-68
had an opponents' on base average .001 above the league average, while
surrendering .135 home runs fewer per 9 innings, and their games were
.01
innings longer than the mean. This boils down to about 20 extra
baserunners
per year, which could account for about 1.5 of the extra double plays
in
which Mazeroski participated. Michael Mavrogiannis and John Henry
Rickert,
using a different methodology, reached about the same conclusion. Thus,
the number of baserunners is not a significant factor in the number of
double plays made by Maz. However, this does not mean that the
remaining
29.5 DPs were due to his skill, since we do not know how many double
play
opportunities he had.
Paul Schwarzenbart, in the initial issue of The Baseball Analyst,
showed
that ballparks have as much, if not more, influence on fielding
statistics
as they do on hitting and pitching numbers. Since we do not know what
the
fielding park factors for Forbes Field were, we cannot tell whether
this
helped or hurt Maz' stats.
Summary
Mr. Emeigh assumed that Mazeroski was solely responsible for all extra
chances and double plays credited to him, and that an average second
baseman
would not have made those plays. Actually, many factors influence
fielding
statistics, and one should not assume that a player's defensive value
is
accurately portrayed by them. Other facts, such as the Pirates record
and
number of runs allowed during Mazeroski's career suggest, though do not
prove, that his defensive value was much less than the 57 runs
calculated
by Mr. Emeigh.
Updated Commentary
Several people contributed ideas and data to the discussion of this
issue on the SABR mailing list, in particular John Rickert, Tom Ruane,
and Rich Hansen, as well as Mike Emeigh. I will herewith attempt to
summarize
the key points.
The ratio of a team's infielders' assists to its outfielders'
putouts
acts as a good surrogate for the groundball to fly ball ratio of a
teams'
pitching staff. While both of these ratios are affected by the quality
of the fielders as well as the pitchers' groundball/flyball tendencies,
they do help to estimate the extra chances Maz may have gotten as a
result
of the type of pitchers used by the Pirates. Also, as noted above, the
Pirates had more balls in play against them than the average team and
had
used right handed pitchers slightly more than the average staff. These
three factors were estimated by Mr. Emeigh to increase Mazeroski's
chances
by 7%.
Other Pirates, filling in for Mazeroski while he was injured, made
nearly
as many double plays per game as he did (.804 versus .815), indicating
that a large portion of the extra double plays may have been due to
circumstance
rather than skill.
A large portion of secondbasemen's putouts, about 55-60% on average,
come on balls which they do not field, mainly forceouts. A smaller
portion
of their assists, about 15%, similarly come on balls others fielded.
Mr.
Emeigh calculated that, due to the groundball tendencies of the
Pirates'
pitchers, all, and then some, of Mazeroski's additional putouts were a
result of simply catching throws and were not reflective of fielding
value.
Tom Ruane found, using a small sample, that about 35% of a second
baseman's
chances are on balls they don't field.
Rich Hansen, using Retrosheet files for the 1967 and 1982 American
League
seasons, published data on double plays. He found for 1967 that actual
double play opportunities, meaning groundballs with a runner on first
and
fewer than two out (accounting for about 70% of double plays,) varied
among
secondbasemen by as much as .54 per nine innings. In 1982, team double
play opportunities for secondbasemen ranged from 190 to 132.
Conclusion
The Mazeroski case presents an interesting example of how difficult
interpreting
fielding statistics can be. Separating fielding from pitching is
perhaps
the most difficult task in baseball analysis. While Mr. Emeigh dealt as
well as possible with the available information, he was forced to use
average
where actual information was unavailable. As Messrs. Ruane and Hansen
found,
actually opportunities can vary widely in a season, or even over
several
seasons. One should be hesitant to base any decision solely on even
well
adjusted fielding statistics. It is best to keep this in mind when
considering
hitting and pitching statistics, too.
Any questions or comments, please e-mail me at CliffordBlau@yahoo.com
Return to homepage