About three weeks ago I wrote the following:
My longtime friend Craig Wright produces a subscription service called “A Page From Baseball’s Past”, which is just excellent; I learn a lot by reading it . ..I think you can reach them at support@baseballspast.com. I hope. Anyway, in a recent article Craig was writing about Jeremiah Denny, 19th century third baseman, and he departed into the issue of Denny’s defense, thus bringing up a long-unresolved research issue that the time has probably come to move against. I give little or no weight to putouts by a third baseman, in part, because bad teams have more putouts by third basemen than good teams. It seems to me that if it is characteristic of a bad team, then it can hardly be a characteristic of a good player, unless there is something unusual going on that we don’t quite understand.
But our ability to study that issue now is quite a bit better than it was ten years ago, when I was writing the Win Shares book, so I thought I would take a better look at it. I am really asking here two questions:
1) Is it actually true that bad teams have more putouts at third base than good teams?, and
2) Assuming that that is true, is it equally true over time? Or is it something that might be true in our time, but not in the 1890s, for example. Or vice versa.
I studied the issue in this way. First, I divided all teams in baseball history into seven eras—
1 1876-1899 (Cap Anson era)
2 1900-1919 (Ty Cobb era)
3 1920-1939 (Babe Ruth era)
4 1940-1959 (Stan Musial era)
5 1960-1979 (Pete Rose era)
6 1980-1999 (Rickey Henderson era)
7 2000-2007 (Willie Bloomquist era)
In each era, then, I divided the teams into four levels by their winning percentage:
a) Winning percentage of .560 or better
b) Winning percentage .500 to .559
c) Winning percentage .440 to .499
d) Winning percentage less than .440
I then looked at the average fielding records for third basemen on each group of teams.
To make this short and sweet, it is clearly true in every era of baseball history that putouts by third basemen are inversely related to the quality of the team. However, this effect may be less in Jeremiah Denny’s era than in any other.
Starting with the Ty Cobb era, this is the average defensive performance of the third basemen, by quality of team, with the best teams on the top line, the weakest teams on the bottom line:
|
|
G
|
PO
|
A
|
Err
|
DP
|
FPct
|
108
|
|
156
|
183
|
310
|
36
|
19
|
.932
|
66
|
|
154
|
187
|
300
|
38
|
20
|
.928
|
61
|
|
156
|
191
|
301
|
37
|
20
|
.930
|
93
|
|
154
|
194
|
307
|
43
|
21
|
.920
|
Third basemen on the best teams in this era recorded an average of 183 putouts; on the worst teams, 194 putouts. The number at left is the number of teams in the group.
In the Babe Ruth era putouts and errors by third basemen went down, double plays and fielding percentage went up:
|
|
G
|
PO
|
A
|
Err
|
DP
|
FPct
|
91
|
|
161
|
163
|
296
|
24
|
25
|
.951
|
84
|
|
161
|
170
|
310
|
26
|
26
|
.948
|
57
|
|
162
|
170
|
305
|
27
|
27
|
.946
|
88
|
|
162
|
173
|
309
|
28
|
28
|
.945
|
However, we have essentially the same ratio of putouts on good teams to bad teams as we had before. In the Stan Musial era assists and double plays by third basemen increased, while fielding percentages stabilized:
|
|
G
|
PO
|
A
|
Err
|
DP
|
FPct
|
91
|
|
170
|
166
|
320
|
24
|
31
|
.954
|
75
|
|
170
|
167
|
314
|
26
|
29
|
.950
|
69
|
|
168
|
169
|
327
|
26
|
31
|
.950
|
85
|
|
166
|
175
|
319
|
27
|
30
|
.948
|
The same ratio as always between putouts on bad teams and putouts on good teams. . .6, 7% more putouts by third basemen on bad teams. It’s a pretty significant difference. The fact that there is that much of a difference in putouts for teams of different quality suggests, to me, that there are large number of plays included there which have a minimal skill component. In the Pete Rose era there was a substantial shift in the ratio of assists to putouts among third basemen:
|
|
G
|
PO
|
A
|
Err
|
DP
|
FPct
|
98
|
|
184
|
140
|
341
|
24
|
30
|
.953
|
141
|
|
181
|
141
|
335
|
24
|
30
|
.951
|
109
|
|
179
|
144
|
341
|
24
|
30
|
.952
|
96
|
|
182
|
149
|
338
|
27
|
31
|
.947
|
The schedule got 8 games longer in 1961/62, which helps explain the increase in assists, but what explains the decrease in putouts?
It’s the thin-handled bats, I think. As bat handles got thinner—which has a long history, but really begins with Ernie Banks—as bat handles got thinner there were fewer balls popped up off the fist high enough that the third baseman could catch them. This trend continues into the 21st century; if you hit a player on the first now the bat just shatters in his hand, and you have a weak ground ball. This is the data from the Rickey Henderson era:
|
|
G
|
PO
|
A
|
Err
|
DP
|
FPct
|
104
|
|
184
|
114
|
307
|
21
|
26
|
.952
|
168
|
|
185
|
118
|
314
|
22
|
27
|
.951
|
165
|
|
181
|
115
|
308
|
23
|
28
|
.949
|
101
|
|
181
|
118
|
312
|
24
|
28
|
.947
|
There are still more putouts by third basemen on bad teams than good, but the difference is declining; it’s 3% here, not six or seven. Putouts are down, assists are down a little bit in part due to the 1981 and 1994-95 strikes, and in part to increases in strikeouts. And this is the data from the Willie Bloomquist era:
|
|
G
|
PO
|
A
|
Err
|
DP
|
FPct
|
57
|
|
183
|
117
|
317
|
19
|
29
|
.957
|
69
|
|
181
|
115
|
313
|
20
|
29
|
.955
|
55
|
|
182
|
118
|
317
|
22
|
31
|
.951
|
59
|
|
181
|
119
|
316
|
23
|
30
|
.951
|
The database I am using here still hasn’t been updated for 2008. Anyway, the data is changing a little bit. Fielding percentages by third basemen—stable for 80 years—have apparently started to edge up. The putout difference between good and bad teams appears to be narrowing. Note the de-centralization of teams. From 1900 to 2000, more and more teams were drifting into the “mid-range” categories, with winning percentages of .440 to .560. In the Ty Cobb era there were 201 “good or bad” teams, 127 mid-range teams. By the Rickey Henderson era there were 205 “extreme” teams, but 333 mid-range teams. In the twenty-first century, at least so far, it’s going back the other way.
Now, the data for the Cap Anson/Jeremiah Denny era:
|
|
G
|
PO
|
A
|
Err
|
DP
|
FPct
|
112
|
|
120
|
158
|
246
|
57
|
17
|
.875
|
64
|
|
123
|
167
|
253
|
64
|
17
|
.868
|
50
|
|
121
|
162
|
245
|
59
|
17
|
.873
|
100
|
|
115
|
155
|
237
|
68
|
17
|
.853
|
The data from this era is difficult to interpret because the game was so un-stable. Teams went out of business in mid-season, and it was never the good teams that did that, but the standard deviation of winning percentage changed a great deal between 1876 and 1899, as did the schedule length, normal fielding percentages, and everything else. It’s hard to know what to make of the data.
However, the putouts/game from third basemen on good teams in this era averaged 1.31, while the putouts/game from third basemen on bad teams averaged 1.35. At a minimum, we can conclude that putouts by third basemen have always been inversely related to the quality of the team, for the simple reason that
a) when you have more baserunners on against you, you have more plays at third base, and
b) when you have more baserunners on against you, you lose more games.
Before I could get that published, however, and accept the Nobel and Pulitzer Prizes that would surely follow from its publication, Craig wrote about third base putouts again. His Diamond Appraised for July 10, 2009, wrote about Third Base Putouts for six solid pages in a vain effort to steal my Nobel Prize for Third Base Putout Research. Craig was responding directly to, and disputing, my conclusion that third base putouts appear to be of no value in assessing the defensive quality of a third baseman. In the course of this he makes several thousand arguments about third base putouts, a few of which I wanted to respond to. . . .
“In his book Win Shares he explained his belief is based largely on a study he did of three groups of team data at third base from the period 1965 to 1990.”
That’s not exactly true. The “Gold Glove study”, which found that teams with Gold Glove third basemen have no more putouts than teams that have no real idea who their third baseman is, was the final nail in the coffin for third base putouts, but it was not the basis of my conclusion. There were actually a series of studies that tried to find value in third base putout data, and failed at every corner.
“At this level it is possible to speak of someone’s `belief’ without denying the evidence being considered in that interpretation. It is exactly Bill’s `belief’ of what this evidence means that I hesitate over and do not find convincing. I’m not convinced that an inability of putouts to separate a generic group of Gold Glovers for mother third basemen is the same as saying that putouts cannot indicate defensive value that should be assigned to a fielder.”
First of all, Craig is focused on that one study, which really was NOT the basis of the conclusion that third base putouts are useless. The bigger fact is that it doesn’t correlate with winning—in fact, it correlates with losing at a fairly significant level. All I am really asking about the stat is “Is this characteristic of good defensive players?” If something is characteristic of losing teams that suggests that it is not characteristic of good players. Losing teams tend to have few homers, few walks, and many errors. We thus conclude that players who don’t hit homers, don’t walk and make errors tend, other things being equal, to be bad players or, as we call them in Kansas City, Neifi Perez/Yuniesky Betancourt type players.
Sabermetrics, like any science, can be seen as a search for a compelling logic. The question is, is this a compelling logic, or a chosen belief? I’ll leave that to your judgment, but it seems pretty compelling to me. If bad teams have more putouts at third base than good teams—which they do—what other conclusion could one draw from that?
At one point, in attempting to explain why I am wrong, Craig comes perilously close to explaining why I am right:
“Can we logically anticipate reasons why that potential distinguishing characteristic does not actually come through when comparing the better third basemen to the rest? Sure we can. The more mundane reason is that we may be missing an adjustment or a series of adjustments that would bring out the distinguishing element in individual putout statistics at third base. But what I really suspect is more of an issue here is that the individual strategies and abilities that might lead to a third baseman garnering more putouts do not necessarily coincide that well with the strategies and abilities that make for a good overall defensive performance at third base.”
How is that different, exactly, from saying that putouts by third basemen don’t indicate defensive excellence at third base? It seems to me that it amounts to the same thing.
OK, the issue here, I believe, is this. A third baseman who plays IN may have more putouts, because he is closer to home plate and in better position to grab those short blips off the bat, but fewer assists, because some ground balls will get by him before he can react. A third baseman who plays deeper may have fewer putouts but more assists—and more net value. BUT the fact that the better third baseman has fewer putouts does not mean that the putouts don’t ALSO count. It’s like strikeouts and home runs. Good hitters in modern baseball, tend to strike out MORE than weak hitters. The reason is that strikeouts are fellow travelers of home runs. The guys who strike out 200 times a year, like Adam Dunn and Ryan Howard, are not bad hitters; they’re good hitters—but that doesn’t mean that the strikeouts are not harmful. They ARE harmful; they just don’t outweigh the home runs.
Regarding Carney Lansford—who, without any exaggeration whatsoever, would dive for a ball hit two feet to his left and four feet in the air—Craig says that “Lansford edged Bill Madlock for the lowest career rate of assists per inning at third base. But Lansford ranked in the top half in putouts per inning. I’m sure that is partly a result of Lansford playing nearly two-thirds of his innings with the Oakland A’s, who have a massive foul area around their home infield, but with a reasonable adjustment, we might still find that Lansford deserves to have an edge in the small part of his defensive equation.
Do you see my point?”
Well. . .no, actually, I don’t. How does Carney Lansford getting a lot of putouts at third base because he played in the Oakland Coliseum demonstrate that putout data by third basemen has value? I don’t get it.
Anyway, the real reason for responding to Craig’s material. . . .Craig made an effort to sort putouts by third basemen into types of plays.
“The cheapest putout on the infield is taking a good throw on a force play.”
“The next cheapest putout is catching an easy pop fly.”
“The other easy putout is the line drive right at you.”
“But now let’s talk about the putouts that more clearly require (skill). . .(such as) coming up with a bad throw on a force play.”
“Then there are tag plays. ..particularly on long throws from the outfield.”
“We have the short foul pops over by the stands where quick feet (and other assets) come into play.”
“There are groundballs you come up with in a force situation.”
“Finally there are the putouts on the difficult line drives that do require a degree of superior talent.”
This got me to wondering how putouts by third basemen divide among these categories or others. We have a lot of data now, from Retrosheet (All Hail Retrosheet; may Retrosheet be praised). . ..we have a lot of data now, and I thought perhaps we could learn something more about the issue by studying it. I can’t program a computer to do anything more useful than scramble the letters in somebody’s name to try to find anagrams, but my son Isaac is a computer programmer, so I hired Isaac to search ten years of Retrosheet data (1999-2008) and ask a few questions. The first thing we did was to sort putouts by third basemen into four categories:
Type A are pop outs and fly balls (and bunt pop outs, for those of you who are into Retro-code.)
Type B are line drives.
Type C are force outs and plays that could be force outs.
Type D are tag plays and runners doubled off third.
In modern baseball:
56% of third base putouts are Type A.
20% are Type B.
13% are Type C, and
11% are Type D.
In a season, a modern team will have about 117 putouts at third base, of which:
66 are pop outs and fly balls,
23 are line drives,
15 are force plays, and
13 are tag plays and runners doubled off.
The variation from team to team seems essentially proportional to the total. The standard deviation of putouts at third base is 16.6—not a very large number. The standard deviation is:
12.3 for pop outs and flies,
5.7 for line drives,
5.0 for force plays, and
3.9 for tag plays.
Force plays at third have a somewhat higher standard deviation relative to their number than do the other types.
The standard deviation of putouts at third base in road games is 8.5; in home games, it is 10.7. This suggests that there are significant park effects in the category—otherwise, the standard deviation at home should be about the same as the standard deviation on the road.
The most putouts by a third baseman in this era was 152, by Ryan Zimmerman in 2006. Zimmerman—who is certainly a good third baseman, except that he sometimes throws erratically—had 99 Type A putouts, so essentially all of his “above normal” performance was in the catching of pop outs and fly balls. He was +33 in those plays, about +35 overall. He apparently had a park effect working for him at RFK; anyway combining 2006 and 2007 he had 42 more putouts at home than on the road. In 2008, in the new park, he had 14 more on the road than at home.
The largest home/road differential for putouts was by Brandon Inge in 2006—84 putouts at home, 51 on the road. Here is my point again. The Detroit Tigers in 2007, with Brandon Inge playing third base, had 101 putouts at third base, the second-lowest total in the American League. In 2008, with Carlos Guillen and Miguel Cabrera playing third base, they had 149 putouts at third base—easily the highest total in baseball. It’s hard to see how this tracks with quality defensive play at third base.
But the data also does not obviously promote the idea that a large foul territory equals more putouts at third base. Mike Lowell in 2006 had 143 putouts—one of the highest totals over the ten-year period—and these included 79 in Fenway Park, 64 on the road.
Lowell was at that time an excellent third baseman, and one of his assets was that he responded quickly and aggressively to foul popups. Fenway, of course, has limited foul ground, but this doesn’t seem to reduce putouts by their third basemen very much. Billy Mueller, in his three years as the Red Sox third baseman (2003-2005), had 125 putouts in Fenway, 109 on the road. Eric Chavez has a career rate of putouts only 11% higher in Oakland than on the road—not a massive bias.
Ryan Zimmerman in 2006 had 99 Type-A putouts (fly balls and pop outs), and Vinny Castilla in 2005 also had 99; third and fourth on the list are Mike Lowell in 2002 (in Florida) and Mike Lowell in 2006 (in Boston). In Type-B putouts (Line Drives) the leader was Eric Chavez in 2003 (32), and then a bunch of people with 31—Mike Lowell, Adrian Beltre, Aramis Ramirez twice, A-Rod once and Chad Tracy once.
In Type-C putouts (force plays) the leader was Zimmerman in 2007 (30), then Brandon Inge in 2006, Mike Lowell in 2005.
It appears that the “reverse bias” in third base putouts (the tendency for them to b higher on bad teams) may come heavily not from Type C, which I expected, but from Type D, tag plays. The leaders in Type D putouts in this era were:
Jeff Cirillo, 1999
|
24
|
Mike Lowell, 2002
|
22
|
Adrian Beltre, 2005
|
21
|
Adrian Beltre, 2006
|
20
|
Troy Glaus, 2006
|
20
|
Joe Randa, 2000
|
20
|
And here are the won-lost records of their teams:
Jeff Cirillo, 1999
|
24
|
74-87
|
Mike Lowell, 2002
|
22
|
79-83
|
Adrian Beltre, 2005
|
21
|
69-93
|
Adrian Beltre, 2006
|
20
|
78-84
|
Troy Glaus, 2006
|
20
|
87-75
|
Joe Randa, 2000
|
20
|
77-85
|
A third baseman on a weak team will get more tag plays because, when the team is behind, the other team is more likely to try to go first-to-third and more likely to attempt to steal third. Still. . ..one wouldn’t think that would be a big bias. I don’t know.
The tag play on a first-to-third attempt is a HIGH skill play, a play that often requires tremendous skill. The third baseman must handle a 130-foot throw from the cutoff man or sometimes a 180-foot throw from center field. The throw is likely to be anywhere from 9 feet high to in the dirt to six feet off-line either way, and the runner is running hard, is in your way and is attempting to avoid being tagged out—and the umpire will never give you the call unless you earn it. It’s one of the most difficult fielding plays in baseball, and I would bet that it is NOT made far more often than it IS made. That is, most of the time, even when the throw is there in time, it proves too difficult to get the tag on the runner before he gets in.
Looking at this data. . .well, it kind of looks like the good third basemen in this group do have more putouts than the weaker third baseman—not reliably, but on average. In this data Kevin Kouzmanoff, Chad Tracy, Edwin Encarnacion and Miguel Cabrera had more putouts per inning played than Scott Rolen. Still, the people on the top of the list are like Ryan Zimmerman, Joe Randa, Adrian Beltre and Mike Lowell, whereas the people on the bottom of the list are Tony Batista, Dean Palmer, Chipper Jones and Hank Blalock.
I’m not emotionally invested in the proposition that third base putouts are meaningless and should be disregarded. Craig’s theory, I think, can be stated as “there are many third base putouts which are skill plays, therefore the data should be presumed to be meaningful.” Well, sure. . .I presumed that it was meaningful data until I looked for the meaning and couldn’t find it. It’s a question of signal-to-noise ratio. If you look at the four types of putouts in my research, three of them are problematic.
Pop Outs to third base are subject to foul-territory biases and also to “discretionary” biases, since many of them can be caught by either the third baseman or the shortstop. Also, a very large percentage of them are routine, low-skill plays.
Force Outs at third are inversely related to the quality of the team, and tag plays at third are inversely related to the quality of the team.
That’s a lot of noise in data that has a relatively low standard deviation to begin with. I think there’s too much noise in the data to draw any conclusion from it. If you can figure out a way to suppress some of the noise and make something out of it, good on you.