Just for the Halibut, I made up two teams of what appear to be different but equal players. First, I sorted out all players in baseball history who had Win Shares Winning Percentages, for the season, between .590 and .610. ..that is, two teams of .600 players. Good players, not MVPs or anything. Second, I eliminated all players who (a) had less than 400 plate appearances, or (b) played before 1940.
This left me with a field of 690 players. I ranked them 1 through 690 in two areas:
1) The ratio of walks to plate appearances, and
2) The ratio of on base percentage to slugging percentage.
Combining those two, I then had all of the high walk players on one end of the scale, and all of the low-walk players—but players of overall equal quality—on the other end of the scale. Given that list, I then chose All-Star teams—the Joe Cunningham team, of players who walked a lot, and the Joe Carter team, of players who were good hitters but did not walk. Although, as it happened, neither Cunningham nor Carter was on either team, as neither man happened to have a season that landed in the appropriate range.
Anyway, this is the Joe Carter team:
Player
|
G
|
AB
|
R
|
H
|
2B
|
3B
|
HR
|
RBI
|
BB
|
SO
|
SB
|
CS
|
Avg
|
OBA
|
SPct
|
OPS
|
Juan Samuel, 1986, 2b
|
145
|
591
|
90
|
157
|
36
|
12
|
16
|
78
|
26
|
142
|
42
|
14
|
.266
|
.302
|
.448
|
.751
|
Alfonso Soriano, 2007, lf
|
135
|
579
|
97
|
173
|
42
|
5
|
33
|
70
|
31
|
130
|
19
|
6
|
.299
|
.337
|
.560
|
.897
|
Juan Gonzalez, 1997, dh
|
133
|
533
|
87
|
158
|
24
|
3
|
42
|
131
|
33
|
107
|
0
|
0
|
.296
|
.335
|
.589
|
.924
|
Adam Jones, 2013, cf
|
160
|
653
|
100
|
186
|
35
|
1
|
33
|
108
|
25
|
136
|
14
|
3
|
.285
|
.318
|
.493
|
.811
|
Jorge Cantu, 2005, 3b
|
150
|
598
|
73
|
171
|
40
|
1
|
28
|
117
|
19
|
83
|
1
|
0
|
.286
|
.311
|
.497
|
.808
|
Alexei Ramirez, 2008, ss
|
136
|
480
|
65
|
139
|
22
|
2
|
21
|
77
|
18
|
61
|
13
|
9
|
.290
|
.317
|
.475
|
.792
|
Lee May, 1973, 1b
|
148
|
545
|
65
|
147
|
24
|
3
|
28
|
105
|
34
|
122
|
1
|
1
|
.270
|
.310
|
.479
|
.789
|
Cory Snyder, 1986, rf
|
103
|
416
|
58
|
113
|
21
|
1
|
24
|
69
|
16
|
123
|
2
|
3
|
.272
|
.299
|
.500
|
.799
|
Salvador Perez, 2017, c
|
129
|
471
|
57
|
126
|
24
|
1
|
27
|
80
|
17
|
95
|
1
|
0
|
.268
|
.297
|
.495
|
.792
|
And this is the Joe Cunningham team:
Player
|
G
|
AB
|
R
|
H
|
2B
|
3B
|
HR
|
RBI
|
BB
|
SO
|
SB
|
CS
|
Avg
|
OBA
|
SPct
|
OPS
|
Brett Butler, 1988, cf
|
157
|
568
|
109
|
163
|
27
|
9
|
6
|
43
|
97
|
64
|
43
|
20
|
.287
|
.393
|
.398
|
.791
|
Pee Wee Reese, 1950, ss
|
141
|
531
|
97
|
138
|
21
|
5
|
11
|
52
|
91
|
62
|
17
|
0
|
.260
|
.369
|
.380
|
.750
|
Don Buford, 1970, lf
|
144
|
504
|
99
|
137
|
15
|
2
|
17
|
66
|
109
|
55
|
16
|
8
|
.272
|
.406
|
.411
|
.816
|
Larry Walker, 2003, rf
|
143
|
454
|
86
|
129
|
25
|
7
|
16
|
79
|
98
|
87
|
7
|
4
|
.284
|
.422
|
.476
|
.898
|
Mickey Tettleton, 1994, c
|
107
|
339
|
57
|
84
|
18
|
2
|
17
|
51
|
97
|
98
|
0
|
1
|
.248
|
.419
|
.463
|
.882
|
Ferris Fain, 1947, 1b
|
136
|
461
|
70
|
134
|
28
|
6
|
7
|
71
|
95
|
34
|
4
|
5
|
.291
|
.414
|
.423
|
.837
|
Mike Hargrove, 1980, dh
|
160
|
589
|
86
|
179
|
22
|
2
|
11
|
85
|
111
|
36
|
4
|
2
|
.304
|
.415
|
.404
|
.819
|
Eddie Yost, 1952, 3b
|
157
|
587
|
92
|
137
|
32
|
3
|
12
|
49
|
129
|
73
|
4
|
3
|
.233
|
.378
|
.359
|
.738
|
Willie Randolph, 1985, 2b
|
143
|
497
|
75
|
137
|
21
|
2
|
5
|
40
|
85
|
39
|
16
|
9
|
.276
|
.382
|
.356
|
.738
|
The Joe Carter team, you have to construct the lineup by finding the two players who could possibly lead off, and then everybody else is a middle-of-the-order type hitter. The Joe Cunningham team, you have to identify the few players who would naturally fit in the middle of the order, and then everybody else could hit leadoff.
In the aggregate, both lineups have similar playing time—1,239 games, 4,866 at bats for the Joe Carter team, 1,288 games, 4,530 at bats for the Joe Cunningham team. The Joe Carter team has huge advantages in home runs (252-102) and RBI (835-536), and also an advantage in batting average (.282-.273). They have a good advantage in doubles (268-209). The Joe Cunninghams have the edge in runs scored (771-692), triples (38-29), stolen bases (111-93), and also have a huge advantage in walks (912-219) and on base percentage (.399-.315). Their OPS is about the same (.819-.802, Carters), although we probably all know that walks are under-valued in OPS. Each set of nine players is credited with 176 Win Shares.
No particular conclusion here; I just wanted to do that. Maybe this would help to explain it. . ."it" being the general problem we were talking about earlier. We have "knowns", "unknowns" and "game condition variables".
The question posed by 337 was, Could you use a simulation to run a team with 16 position players/9 pitchers against a team with 12 position players and 13 pitchers, and thus decide which is the more effective strategy? Well, no, because that introduces unknowns. When you combine unknowns with game condition variables, you just produce more unknowns.
The problem of the high-walk team against the low-walk team combines KNOWN elements with game condition variables. In a simulation, I know for certain how that would work out. The simulated game conditions WOULD alter the output. They would alter it in favor of the players who walk. I have run those kind of simulations: I know that is what happens. You put a bunch of guys with high on base percentages together, the pitchers run out of bases on which to put them. In a simulation, the Joe Cunningham team (above) would easily defeat the Joe Carter team, even though the teams are theoretically equal. The teams are equal on their expected impact on a normal team context. If you alter the context so that the walks can build up higher and higher, that neutralizes the weakness of the Cunningham types—their relatively weak ability to drive in runs—and maximizes their strength, which is their ability to create scoring opportunities. There are only so many scoring opportunities that you can create before runs start scoring.
Marisfan suggests that there is a DIFFERENT game condition variable at work there somewhere. . . .I don’t know exactly. He suggests that there could be a "real life" variable which is not replicated in the simulation—like the advantage of pitchers pitching in short windows. Hitters of that type lose effectiveness when they are stacked together. It doesn’t really make any sense, and I don’t believe it, but there could be something there that we just don’t understand yet. You can’t ASSUME that there is some unknown variable that would be at work; you’d have to prove it. You’d have to articulate exactly HOW this real life variable occurred, and why it occurred, so that we could guess WHEN it would occur, so that we could search for evidence of it in real life. But. . .can’t rule out the possibility that there is something there.