3-6-3
In 2010 a man named Chris Chestnut posted an article on Retrosheet reporting on his research about 3-6-3 double plays—specifically, about how many 3-6-3 double plays are started by each first baseman. For some reason I just discovered this article about three days ago, and I was fascinated by the research. I tweeted out a comment from the research (Keith Hernandez started more 3-6-3 double plays in his career than any other first baseman), and also contacted Mr. Chestnut to ask if he had updated the research—not only because ten years have passed, but also because, in those ten years, Retrosheet has pushed their research backward in time, so that it would now include. . .well, George Sisler. I did that in the wrong order; I should have contacted Mr. Chestnut BEFORE I tweeted out about Hernandez, because, as it turns out (a) Mr. Chestnut’s old e-mail address, published with his article, still works, (b) he HAS, in fact, updated the research, (c) he was generous enough to share his new research with me, and (d) Hernandez no longer is #1 on the list, although he is still near the top. I should have talked to him BEFORE I tweeted that out; sorry. Apparently prompted by my enquiry, Mr. Chestnut has published his updated research, also on Retrosheet.
I have been interested in "this subject" for many years, and I have written about it several times. "This subject" is first basemen’s meaningful defensive plays. The original defensive statistics of first basemen are very badly designed, and thus fail to distinguish between the types of plays that a first baseman makes. This has always bothered me because there is something in there which is very measurable, not in any sense intangible or even elusive, but a very instructive something that could routinely be counted, but which we just don’t BOTHER to keep track of. Mr. Chestnut bothered. Can I nominate him for an Award here? This is good research.
Mr. Chestnut counted three things for each first baseman. These are his definitions:
3-6-3 Double Play – the first baseman fields a ground ball and throws to the shortstop to retire the runner on first. The shortstop returns the throw to the first baseman to retire the batter/runner.
Opportunity – all ground balls (including bunts, excluding base hits) fielded by the first baseman with a runner on first (additional runners may be on second and/or third) and less than two outs.
Attempt – all opportunities where the first play by the first baseman is a throw to the shortstop.
I would say that I essentially agree with all of those definitions. If I had been doing this study, I think those are the definitions that I would have used. The only problem I see is that it is unclear how an event enters the chart if a first baseman fields the ball and throws it to second base either (a) too late, putting an extra runner on base, or (b) resulting in an E-3, or (c) resulting in a 3-6-1 Double Play. I would guess that the first two of those categories are not accessible from the available records. There probably is just no way of knowing.
I have some disagreements with Mr. Chestnut about his preparation of the charts of data resulting from this research. He figures and reports successful double plays as a percentage of attempts, rather than as a percentage of opportunities. This seems obviously wrong to me. What he seems to be implicitly assuming is that if the first baseman does not attempt to make the play at second, it is generally because there is no play there to be made. We should accept the first baseman’s judgment on that play, and evaluate the success of the first baseman based only on when he does make the play to second. I think it would be pretty easy to demonstrate that he picked the wrong side of that argument. I’ll get back to that issue.
Also, Mr. Chestnut fills up his charts with completely useless rankings, making it much more difficult to find the useful information hiding among the trees. But these are trivial objections. What matters is that he did the work, and he shared his data. I can extract his data and put it in my preferred form easily enough.
OK, first conclusion; these are the top 10 first basemen in the number of career 3-6-3 double plays started (and, I suppose, ended).
Last
|
First
|
T
|
DP
|
Pujols
|
Albert
|
R
|
81
|
McQuinn
|
George
|
L
|
73
|
Hernandez
|
Keith
|
L
|
72
|
Vernon
|
Mickey
|
L
|
72
|
Mattingly
|
Don
|
L
|
71
|
Kuhel
|
Joe
|
L
|
68
|
Hrbek
|
Kent
|
R
|
66
|
Grimm
|
Charlie
|
L
|
64
|
Murray
|
Eddie
|
R
|
64
|
Scott
|
George C.
|
R
|
60
|
Those are all good first basemen, but it is surprising to see Pujols—a right-handed first baseman who won two Gold Gloves—finishing well ahead of Keith Hernandez, a left-handed first baseman who won ten Gold Gloves.
Pujols not only beat Hernandez in 3-6-3 double plays started, but did so in fewer innings, and in fewer opportunities. Let’s add opportunities and innings to the chart:
Last
|
First
|
T
|
Innings
|
Oppor
|
DP
|
Pujols
|
Albert
|
R
|
16395
|
544
|
81
|
McQuinn
|
George
|
L
|
13231
|
426
|
73
|
Hernandez
|
Keith
|
L
|
17279
|
688
|
72
|
Vernon
|
Mickey
|
L
|
19595
|
658
|
72
|
Mattingly
|
Don
|
L
|
14132
|
459
|
71
|
Kuhel
|
Joe
|
L
|
18203
|
505
|
68
|
Hrbek
|
Kent
|
R
|
13660
|
386
|
66
|
Grimm
|
Charlie
|
L
|
18771
|
621
|
64
|
Murray
|
Eddie
|
R
|
21151
|
642
|
64
|
Scott
|
George C.
|
R
|
15190
|
570
|
60
|
Hernandez had about 5% more innings than Pujols, and 26% more DP opportunities, but Pujols started more double plays.
It has been so long now since Pujols has been what he was that it is hard to remember. He really was a magnificent player. What I remember about him is how alert he was. Anything that happened on the field, he reacted to it instantly. He had a good arm, a third baseman’s arm, although he wasn’t really quick enough to play third. But it is very surprising that he is at the top of this list; not surprising that he is above average, but surprising that he is number 1.
Here’s one way we know that "opportunities" are the correct basis for comparison, not "attempts". If you compare left-handed first basemen to right-handed first basemen based on 3-6-3 double plays PER ATTEMPT, left-handed first basemen are only 1.3% better than right-handed first basemen. Left-handed first basemen get the double play on 20.85% of attempts; right-handed first basemen, on 20.58%.
But if you compare them PER OPPORTUNITY, the difference is not 1%; it is 15%! Left-handed first basemen get the double play when a ball is hit to them in a potential double play 7.9% of the time; right-handers, only 6.9%. Right-handed first basemen don’t make the play not because they try and fail, but because they don’t try as often.
If "attempts" were the right basis to be evaluating individual first basemen, how would you explain that? It would seem to me that you can’t. These are the first basemen who turned the most double plays, per opportunity (minimum: 100 chances):
Last
|
First
|
T
|
Oppor
|
DP
|
Pct
|
Goodman
|
Billy
|
R
|
125
|
23
|
.184
|
Stevens
|
Lee
|
L
|
237
|
42
|
.177
|
McQuinn
|
George
|
L
|
426
|
73
|
.171
|
Hrbek
|
Kent
|
R
|
386
|
66
|
.171
|
Tabler
|
Pat
|
R
|
129
|
21
|
.163
|
Goldschmidt
|
Paul
|
R
|
339
|
55
|
.162
|
Overbay
|
Lyle
|
L
|
367
|
58
|
.158
|
Mattingly
|
Don
|
L
|
459
|
71
|
.155
|
Pujols
|
Albert
|
R
|
544
|
81
|
.149
|
Morneau
|
Justin
|
R
|
302
|
44
|
.146
|
Billy Goodman was an all-over-the-infield-and-occasionally-the-outfield player. The norm for 3-6-3 double plays per opportunity is .073, with a standard deviation of .0345. Goodman is 3.45 standard deviations above the norm. Only one player is even two standard deviations below the norm:
Last
|
First
|
T
|
Oppor
|
DP
|
Pct
|
Thomas
|
Lee
|
L
|
120
|
0
|
.000
|
Hatteberg
|
Scott
|
R
|
153
|
1
|
.007
|
Jordan
|
Ricky
|
R
|
139
|
1
|
.007
|
Merkle
|
Fred
|
R
|
123
|
1
|
.008
|
Dunn
|
Adam
|
R
|
110
|
1
|
.009
|
Bell
|
Josh E.
|
R
|
106
|
1
|
.009
|
Horton
|
Tony
|
R
|
147
|
2
|
.014
|
Oliver
|
Al
|
L
|
212
|
3
|
.014
|
Perry
|
Gerald
|
R
|
203
|
3
|
.015
|
Cepeda
|
Orlando
|
R
|
522
|
8
|
.015
|
Except for Cepeda, those are all guys who didn’t have a lot of opportunities; no doubt Thomas would have gotten a few if he had had more chances. Thomas was a guy from the Yankee farm system, in the minors from 1954 to 1960. He drove in 122 runs in the minors in 1959, didn’t make the majors in 1960, drove in 112 in 117 games in the minors in 1960. He was a notorious hothead, expansion finally got him to the majors in 1961, but probably two years late. He had a couple of good years with the bat. I got to know him a little bit later; he worked for the Red Sox when I was first there. Scott Hatteberg is second from the bottom; his late-in-life transition to first base was documented in Moneyball.
Expected double plays per opportunity are .073. By comparing expected double plays to actual, we can measure the distance between the player and the average, over the course of his career:
Last
|
First
|
T
|
Oppor
|
DP
|
Margin
|
McQuinn
|
George
|
L
|
426
|
73
|
42
|
Pujols
|
Albert
|
R
|
544
|
81
|
41
|
Hrbek
|
Kent
|
R
|
386
|
66
|
38
|
Mattingly
|
Don
|
L
|
459
|
71
|
37
|
Overbay
|
Lyle
|
L
|
367
|
58
|
31
|
Kuhel
|
Joe
|
L
|
505
|
68
|
31
|
Goldschmidt
|
Paul
|
R
|
339
|
55
|
30
|
Judge
|
Joe
|
L
|
442
|
58
|
26
|
Blue
|
Lu
|
L
|
412
|
55
|
25
|
Stevens
|
Lee
|
L
|
237
|
42
|
25
|
George McQuinn was 43 double plays better than an average first baseman.
George McQuinn was Keith Hernandez before Keith Hernandez was. He was a left-handed hitting, left-handed throwing first baseman, outstanding on defense, and with batting stats also very similar to Hernandez’. He played his best years in St. Louis, for the Browns, in spite of which he made the All Star team seven times. He made the All Star team five times with the Browns, which I would guess might be a record, I don’t know, and then made it twice more with the Yankees toward the end of his career. St. Louis, New York. Same as Hernandez.
He hit like Hernandez, too, getting 195 hits each year his first two real seasons in the majors, scored 100 runs each year, had a good strikeout to walk ratio, hit 12 to 20 homers every year and hit a lot of doubles and triples. He was as close to Keith Hernandez as you can get.
At the bottom of the list is Orlando Cepeda:
Last
|
First
|
T
|
Oppor
|
DP
|
Margin
|
Cepeda
|
Orlando
|
R
|
522
|
8
|
-30
|
Garvey
|
Steve
|
R
|
561
|
13
|
-28
|
Grace
|
Mark
|
L
|
712
|
31
|
-21
|
McCovey
|
Willie
|
L
|
550
|
21
|
-19
|
Fournier
|
Jack
|
R
|
321
|
5
|
-18
|
McGriff
|
Fred
|
L
|
579
|
24
|
-18
|
Helton
|
Todd
|
L
|
789
|
40
|
-18
|
Powell
|
Boog
|
R
|
399
|
15
|
-14
|
Trosky
|
Hal
|
R
|
344
|
11
|
-14
|
Hodges
|
Gil
|
R
|
639
|
33
|
-14
|
Mostly those guys were not good first basemen and nobody would argue about it, but there are some surprises on the list. Mark Grace I would have expected to be on "leaders" list, not the list of players on the bottom, and Gil Hodges was a great defensive first baseman, at least by reputation. Todd Helton was a good first basemen.
All of this reminds us that there are other variables in here that we’re not tracking. We’re counting 3-6-3 Double Plays, but ignoring (I think) 3-6-1 Double Plays. Some teams are really insistent on the pitcher covering first on a potential double play; other teams don’t care as much. Probably there are many more 3-6-3 double plays than 3-6-1, but until we study it, we don’t know. Maybe 3-6-1 DPs are 10% of 3-6-3 DPs, maybe they’re 40%. We don’t know. If they’re 40%, then on some teams they might be 80%. There’s a variable there that we’re not tracking.
Not in any way suggesting the Chestnut should have counted these as well; it’s work, and it is time-consuming. I appreciate what he did do; I’m not complaining about what he didn’t. But we don’t know. In some eras they start the runner in a double play situation quite a bit; in other eras (like now) they don’t do that. Not all shortstops are created equal; sometimes you don’t get the 3-6-3 DP because the shortstop is slow. I don’t know why Mark Grace and Gil Hodges have low 3-6-3 totals. I’m sure we’ll eventually figure it out.
Now, I may owe an apology to Steve Garvey, because I have written about this subject many times, but here we are again. Not having a good throwing arm is not a moral failing; I don’t have a good throwing arm, either. Garvey very explicitly acknowledged in his autobiography that he had a poor arm, that he did not trust his arm and that he would not make a throw unless he had to. I respect the fact that he took ownership of his weakness.
Garvey is -28 on double plays, which is the second-worst of all time. But the reason I keep coming back to this is: I can’t get people to understand how unusual it was. It’s like the thing I was talking about on twitter yesterday, about Albert Pujols and players who were way above .300 and wound up hitting less than .300. You say that, people start saying oh, this is like Mantle or this is like Cecil Cooper or this is like Jeff Bagwell or this is like Enos Slaughter. No no no no no; you’re missing the point. It isn’t like ANY of those—or anybody else. Pujols was 180 hits above .300. This is more than Mantle, Cecil Cooper, Bagwell and Slaughter combined. It’s a unique thing.
And the same with Garvey’s throwing problem. It’s a unique thing. People will start saying that Steve Balboni couldn’t throw and Frank Thomas couldn’t throw and Cecil Fielder couldn’t throw, etc. No no no no no; you’re missing the point. It’s a unique thing.
And here, because of the work of Chris Chestnut, I finally have the goods to maybe make a few people understand. I’ll give you the data in a minute.
The importance of Chestnut’s research is not that it finishes the subject; it is that it opens it up. It is likely, not certain but likely, that a first baseman who has a high total of 3-6-3 Double Plays also has a high total of 3-6-1 Double Plays. It is likely that he has a higher total of simple 3-6 forceouts, each one of which keeps a runner out of scoring position. It is likely that when he has to make a relay throw home 9-3-2, or when he has to cut it off and throw to second, 9-3-4, it is likely that he is better at it. It is likely that when he has to field a ball with a runner on third and fire it home (3-2), he is good at it. The 3-6-3 double plays are the tip of the iceberg.
And, although Chestnut did not focus on that and did not seem to be aware of it, his research actually shows us another good chunk of the iceberg. Think about it. Chestnut defines a 3-6 forceout OR a 3-6-3 Double play as an "attempt". If you take the "attempts" and you subtract the 3-6-3 Double plays, what do you have left? YOU HAVE THE FORCEOUTS.
It may well be that the most important evidence to come out of Chestnut’s study is not actually what he was focused on, but what he also happened to count, which is the forceouts. According to Chestnut, Steve Garvey had 561 "opportunities" is his career to attempt a double play, but made only 61 "attempts" at starting a double play. That means that, the other 500 times, he just let the runner from first go to second base. He just let the runner from first go to second base! 89% of the time, 500 times in 561 opportunities, he just let him go.
I just invented a new stat there, sort of—"Runners Advanced by Inaction" or "Runners Allowed to Advance" or something. Garvey was at 89.1%.
How unusual is this number? The norm for all first baseman is 64.6%. The average first baseman PREVENTED the advance from first base into scoring position 35% of the time. Garvey did it 11% of the time. This is rather a stunning difference. Garvey is 3.3 standard deviations worse than the norm.
The consequence of this is that Garvey allowed a runner to move uncontested into scoring position 138 more times than an average first baseman. This has, I would think, almost exactly the same effect as 138 stolen bases.
Baseball players are evaluated, over the course of a career, on a scale of thousands of bases. 138 bases is not the difference between a good player and a bad player. 138 bases is not something to sneeze at, either. Each base (first to second) is about .16 runs. The cost of NOT making that throw to nail the runner at second was about 22 runs.
These are the worst first basemen of all time, in terms of just allowing runners to move uncontested from first base to second base:
Last
|
First
|
T
|
Oppor
|
Att
|
Runner Advanced
|
Cost
|
Garvey
|
Steve
|
R
|
561
|
61
|
.891
|
-138
|
McCovey
|
Willie
|
L
|
550
|
148
|
.731
|
-47
|
Cepeda
|
Orlando
|
R
|
522
|
139
|
.734
|
-46
|
Buckner
|
Bill
|
L
|
581
|
161
|
.723
|
-45
|
Stuart
|
Dick
|
R
|
292
|
59
|
.798
|
-44
|
McGwire
|
Mark
|
R
|
458
|
118
|
.742
|
-44
|
Thomas
|
Frank E.
|
R
|
215
|
35
|
.837
|
-41
|
Dropo
|
Walt
|
R
|
357
|
87
|
.756
|
-39
|
Pipp
|
Wally
|
L
|
496
|
141
|
.716
|
-35
|
Burns
|
George H.
|
R
|
397
|
106
|
.733
|
-35
|
The cost to Garvey’s teams from this weakness was basically three times the cost for the next-worst player in this respect. That’s what I was trying to get to: Garvey’s unwillingness to throw, out of fear that he would throw the ball away, is not like a "common" thing. It is a completely unique thing. Nobody is comparable; nobody else is anything like that.
If you look at the list above. . .these are not famously good fielding first basemen. Dick Stuart. . .whenever Dick Stuart is on a list of the worst ever, it validates the method, because Stuart is the most famous bad fielding first baseman in baseball history. McCovey, Frank Thomas, Walt Dropo. . .these guys are not famous for their fielding.
But Garvey won four Gold Gloves in his career. Doesn’t it seem to you relevant that Garvey just. . .wouldn’t throw? I will bet you that when we document 3-2 plays and 3-5 plays and 3-4 plays, Garvey is also going to be near the bottom of those lists. I’m not denying that Garvey may have done some things well with the glove.
Well, finish up here. . . this is the list of the players who were BEST at taking out the runner at second base:
Last
|
First
|
T
|
Oppor
|
Att
|
Runner Advanced
|
Benefit
|
Kuhel
|
Joe
|
L
|
505
|
268
|
.469
|
89
|
Hernandez
|
Keith
|
L
|
688
|
332
|
.517
|
88
|
Olerud
|
John
|
L
|
574
|
290
|
.495
|
87
|
McQuinn
|
George
|
L
|
426
|
215
|
.495
|
64
|
Overbay
|
Lyle
|
L
|
367
|
192
|
.477
|
62
|
Mattingly
|
Don
|
L
|
459
|
223
|
.514
|
61
|
Vernon
|
Mickey
|
L
|
658
|
293
|
.555
|
60
|
Cooper
|
Cecil
|
L
|
425
|
205
|
.518
|
55
|
Hrbek
|
Kent
|
R
|
386
|
184
|
.523
|
47
|
Konerko
|
Paul
|
R
|
446
|
204
|
.543
|
46
|
Those are all famously good first basemen, or at least the top 9 are; not confident about Konerko. But Mattingly, Mickey Vernon, Hernandez, Olerud, Joe Kuhel. . .the greatest defensive first basemen of all time. Vic Power, in a shorter career, was +24 on 431 chances.
I could write another 10, 20 pages about where Bill Terry ranks, and where George Sisler is and where Moose Skowron fits and Willie Aikens and Lou Gehrig and Jimmie Foxx and Ted Kluszewski, etc. I’ll try to answer a few questions about that, if you are interested. I am reluctant to print the whole chart, because it is someone else’s research. But I appreciate very much his efforts, and I am very pleased to have hard information about this area of performance.