keyboard_arrow_uptop

The Hubert H. Humphrey Metrodome was a good place to watch a baseball game.

Okay, that might not be “objectively” true, strictly speaking, but having grown up in Minnesota, I had a lot of affection for the Dome. It probably helped that it was the only major-league stadium (and that’s what it was: a stadium, not a ballpark) I’d ever been to, a fact that would remain true until I was 22 years old. Still, its lack of frills did an excellent job of keeping your focus on the game; sight lines were terrific from literally everywhere I ever sat, which was just about everywhere at one point or another; it got incredibly loud (read: incredibly fun) when the team was doing well, though that wasn’t often enough in my prime baseball-game-attending days (the late 90s). It also had its own special touches, as all stadiums/parks do. I was instantly nostalgic about it when Target Field opened in 2010, even while acknowledging Target Field was actually a better place to watch a game (I swear, I’m getting to the point soon).

The Twins carried a few of their Metrodome gimmicks with them to the new park. One in particular was a classic amongst fans, so much so that the (now-defunct) Minnesota Sports-themed apparel company DiamondCentric made a t-shirt out of it: a ghost-like face with a speech bubble saying “WALKS WILL HAUNT!” The Twins might not have been the only team with this sort of graphic, but it fit so well with the ideals the team showed (primarily on the pitching side, but also during the small-ball offensive era of the mid 2000s) that it really stuck with the fans.

Photo courtesy of @KirbysLeftEye

I couldn’t help but think about that when I saw a conversation on Twitter between Matt Goldman and Steve Martano in which they talk about how often intentional walks backfire—that is, how often a run scores after a batter gets walked to load the bases. Having gotten their permission (since it was their idea first), I’m going to look into that here today.

Before I start, I’ll note that (as is so often the case) I’m not the first person to look at this topic; in fact, I’m following up on the work of writers who I greatly admire, which is more than a little intimidating. Grant Brisbee in 2013 started the either Herculean or Sisyphean task, depending on your perspective, of examining every intentional walk to load the bases in 2013 via the box score. He was more interested in the peculiarities of each situation than anything else, though he did conclude “Walk the bases loaded to get to the pitcher. Other than that, c'mon.” He also created an acronym almost-initialism that I’ll be using throughout here, IBBTLTB. Dave Cameron wrote about what happens after an IBB, also in 2013, and although he didn’t focus on IBBTLTBs, which may change the conclusions, he did find that it’s not an absurd strategy (and might even be the right one). Matt Snyder of CBS wrote about it in an article, now only available through the Wayback Machine or Google Cache, after one in particular backfired for the Mariners (again in 2013—there was something about that year I guess), through the lens of run expectancy charts.

There are a bunch of layers to this problem, but let’s start at the very top: In general, and ignoring any possible context, runs will score more often than not after an IBBTLTB. Since 1974, which is the first year for which Retrosheet has complete play-by-play data, there have been 20,559 IBBTLTBs. In 54.06 percent of instances, a run scored after the walk; the average number of runs scored (by the three players subsequently on base) was 1.21.

Year

count

scoring_pct

runners_scoring

Total

20559

54.06%

1.21

Compare this to what happens after not issuing an intentional walk in a two-baserunner situation:

type

count

scoring_pct

runners_scoring

IBBTLTB

20559

54.06%

1.21

Non-walks

842959

42.98%

0.49

That doesn’t tell the whole story, though. The very first bit of context I want to apply is the number of outs when the walk occurs. Unsurprisingly (to me anyway), the majority of IBBTLTBs occur with one out (to set up either a force play at home or a double play). In these situations, though, the frequency of run scoring actually rises; runs score 66.37 percent of the time, and the average number of runs is 1.50. Run scoring is also more frequent when there’s an IBBTLTB with no outs—the rate jumps all the way to 83.10 percent and 1.80 runs/instance. Of course, there’s a drop when looking at two-out situations, but even then this isn’t evidence it’s a worthwhile strategy (without more context).

Year

outs_ct

Count

scoring_pct

runners_scoring

Total

0

1408

0.831

1.7997

Total

1

11510

0.6637

1.4981

Total

2

7641

0.3017

0.6787

There’s not a lot of meaning in cross-year run expectancy charts, so it’s hard to compare these multi-year average numbers I’m listing with expected scoring for a specific year, but using just 2016 as an example, in every case but one the team issuing the walk is worse off than before the walk by both overall RE and expectation of scoring at least one run. I’ll show the two-out situations as an example (within the example, I suppose).

year

base_state

two-out RE

two-out scoring_pct

2016

023

0.5775

24.86%

2016

103

0.4794

38.12%

2016

120

0.4345

26.43%

post-IBBTLTB

0.6787

30.17%

It’s also potentially interesting to look at how scoring is affected by which lineup spot was walked. Looking at each lineup spot individually and split by the number of outs, in only one out of 27 situations did the frequency of scoring a run decrease, and that was (maybe you can guess?) when the eight batter gets IBBTLTB’d with two outs. The frequency of scoring drops all the way from 22.52 percent of the time when the batter isn’t walked to 22.28 percent of the time when he intentionally is. Also unsurprising is that the middle of the lineup is the group most often issued an IBBTLTB, except with two outs, in which case the top spot belongs to the eight batter.

outs

spot

IBBTLTB

non-walks

scoring_diff

re_diff

count

scoring_pct

runners_scoring

count

scoring_pct

runners_scoring

0

1

70

80.00%

1.6571

10917

72.60%

1.3044

7.40%

0.3527

0

2

62

80.65%

1.871

13862

72.57%

1.342

8.08%

0.529

0

3

226

85.84%

1.8673

31155

75.07%

1.2182

10.77%

0.6491

0

4

304

83.22%

1.8651

19029

73.45%

1.1457

9.77%

0.7194

0

5

246

83.74%

1.6911

16240

70.89%

1.1009

12.85%

0.5902

0

6

190

82.63%

1.8368

19338

68.41%

1.0614

14.22%

0.7754

0

7

160

84.38%

1.8125

17800

68.07%

1.0675

16.31%

0.745

0

8

109

78.90%

1.7706

15525

67.49%

1.1085

11.41%

0.6621

0

9

41

80.49%

1.5854

14902

68.53%

1.2677

11.96%

0.3177

1

1

823

68.65%

1.6294

26571

53.83%

0.6377

14.82%

0.9917

1

2

459

70.59%

1.5882

29049

53.61%

0.6939

16.98%

0.8943

1

3

1657

68.50%

1.6584

28041

55.79%

0.665

12.71%

0.9934

1

4

2087

69.72%

1.5961

53394

53.79%

0.5962

15.93%

0.9999

1

5

1721

66.01%

1.473

37180

52.50%

0.5604

13.51%

0.9126

1

6

1562

65.81%

1.4622

31193

50.24%

0.5413

15.57%

0.9209

1

7

1582

64.10%

1.3635

35122

48.49%

0.502

15.61%

0.8615

1

8

1293

59.78%

1.2846

33427

47.55%

0.4992

12.23%

0.7854

1

9

326

64.11%

1.4018

31212

47.24%

0.6019

16.87%

0.7999

2

1

625

32.16%

0.6976

36440

25.26%

0.1409

6.90%

0.5567

2

2

208

31.25%

0.7404

37500

25.16%

0.1693

6.09%

0.5711

2

3

944

34.43%

0.7977

35875

27.78%

0.1633

6.65%

0.6344

2

4

1112

32.55%

0.7419

35879

26.79%

0.1453

5.76%

0.5966

2

5

1086

33.33%

0.7541

60897

25.16%

0.1273

8.17%

0.6268

2

6

890

33.60%

0.7539

46695

24.22%

0.114

9.38%

0.6399

2

7

767

30.12%

0.6597

39490

23.87%

0.1065

6.25%

0.5532

2

8

1854

22.28%

0.4924

41358

22.52%

0.1025

-0.24%

0.3899

2

9

155

30.32%

0.7032

44868

19.37%

0.1046

10.95%

0.5986

Although I think it’d be interesting to further break that chart out by starting base state, at that point I’d need to include 81 rows, and the sample sizes for certain configurations would be so small as to be meaningless. I don’t even feel that confident saying the quarter-percentage-point drop seen with two outs for the eighth batter is meaningful.

This has turned into a bit of a data dump at this point, so the time feels right to return to my original question: How often does an IBBTLTB backfire? The answer appears to be “most of the time." In nearly all situations at least one run is more likely to score, and in general more runs are likely to score, after an intentional walk as compared to letting the batter hit. This is an intuitive result that anyone can see just by looking at the appropriate run expectancy and one-run expectancy tables here at BP, but hopefully you agree it was still worth examining in greater detail.

I’ll close on a mildly uplifting note: the instance of IBBTLTBs have been dropping fairly steadily (from a long-term view) across the entire data set I was working with, and last year, for the first time since 1975, the frequency of scoring after an IBBTLTB dropped below 50 percent, so maybe in another hundred years no one will worry about this at all.

year

Count

scoring_pct

runners_scoring

1974

584

0.5719

1.2466

1975

559

0.4991

1.0823

1976

490

0.6

1.2939

1977

558

0.5341

1.1254

1978

587

0.5554

1.2947

1979

585

0.547

1.3333

1980

545

0.5211

1.1982

1981

346

0.5318

1.1387

1982

554

0.5433

1.1606

1983

569

0.5448

1.2847

1984

540

0.5519

1.3167

1985

532

0.5226

1.2086

1986

500

0.544

1.154

1987

520

0.5212

1.1904

1988

562

0.5605

1.2598

1989

556

0.5558

1.2014

1990

557

0.5189

1.1329

1991

471

0.5478

1.1465

1992

484

0.5434

1.124

1993

521

0.5547

1.2015

1994

396

0.5657

1.2677

1995

424

0.5448

1.217

1996

555

0.5532

1.3063

1997

425

0.5412

1.2024

1998

395

0.5696

1.2557

1999

427

0.5738

1.3208

2000

420

0.569

1.2881

2001

540

0.5259

1.2148

2002

544

0.5184

1.171

2003

520

0.5442

1.2904

2004

491

0.554

1.3136

2005

477

0.5073

1.2138

2006

523

0.5392

1.2256

2007

488

0.5574

1.2049

2008

511

0.5127

1.1468

2009

444

0.5631

1.3356

2010

479

0.5115

1.1378

2011

441

0.5125

1.1882

2012

365

0.5041

1.1562

2013

398

0.5075

1.1231

2014

347

0.5504

1.1585

2015

329

0.4985

1.0608

You need to be logged in to comment. Login or Subscribe
Chiefsnark
9/01
We went to the Dome during a SABR convention in 1988 and yes, it seemed charmless and plastic...someone once said the worst thing he could imagine in baseball was a night game in a dome, using the designated hitter rule, on artificial turf, with wild car implications. But my friend remarked that although she was used to Fenway Park and I was a Tiger Stadium habitue, the kids we saw coming into the Dome would decades from then be telling their own kids about how Grandpa took them to the Twins game and what they saw Kirby Puckett do.
JohnChoiniere
9/02
There seems to have been some trouble with chart headers that were supposed to span multiple columns... let me know if anything's unclear.

Baseball Prospectus uses cookies on this website. They help us to understand how you use our website, which allows us to provide an improved browsing experience. Cookies are stored locally on your computer or mobile device and not by BP. To accept cookies continue browsing as normal. You will see this message only once. Privacy Policy

The cookie settings on this website are set to "allow cookies" to give you the best browsing experience possible. See the BP Cookie Policy for more information. If you continue to use this website without changing your cookie settings or you click "Accept" below then you are consenting to this.

Close