Friday, December 19, 2008

Jack Morris and pitching to the score

With the Hall of Fame voting just around the corner, I thought it would be a good time to revisit the Jack Morris debate. It usually goes something like this:

Jack Morris supporter: Morris had 254 lifetime victories and was the winningest pitcher of the 1980's. That should put him in the Hall of Fame.

Jack Morris skeptic: He won a lot of games because he was durable and played for great teams. Durabililty is a good thing but it doesn't make him a great pitcher. As for winning the most games in a decade, that's an arbitrary time period. It's no more meaningful that being the winningest pitcher from 1976 to 1985 or from 1985 to 1994. Morris just happened to have his peak years in a convenient period: 1980-1989.

Supporter: Jack was a big game pitcher. That 10 inning shutout in the 1991 World Series was a classic. He also won three post-season games for the Tigers 1984 championship team. He's the pitcher you want pitching the seventh game of a playoff for you.

Skeptic: But he pitched poorly during the 1987 and 1992 post-seasons.

Skeptic: Let's talk about his lifetime 3.90 ERA. He had a 105 ERA+ which means he was just 5% better than the league average pitcher during his career.

Supporter: His ERA was high because he pitched to the score. He would coast when his team had a big lead and save his best stuff for tight games. His ERA would have been a lot lower if he pitched the same way with a big lead as he did in close games.

The last part of that exchange is what the rest of this post is about. Morris has said on several occasions that he was more interested in getting wins and pitching deep into games than he was about his ERA and that he would give up a lot of his runs in games where his team had built up a big lead. I don't doubt that pitchers pitch differently according to the situation but how much does it affect their actual performance?

A few years ago, Joe Sheehan at Baseball Prospectus examined Morris' claim in "The Jack Morris Project". He tracked every inning of Morris' career and looked at his performance in different game situations - ties, one run leads, six run leads, etc. Joe did not find that Morris pitched much better in close games than he did with big leads. He did however, admit that he was not really sure what a pitcher's performance would look like if he actually did pitch to the score.

When I started examining the pitching to score theory in May, my goal was to expand upon Sheehan's project by looking at a large group of pitchers rather than just one. One limitation of my analysis at the time was that I had downloaded and prepared retrosheet data only from 1990 forward. Thus, I was comparing Jack Morris to more modern pitchers rather than his contemporaries. I now have data going back to 1977 so I can get a better idea of what pitching to the score looked like and which pitchers had such a tendency during Morris' era.


An example of how the analysis works is this: If a starting pitcher begins an inning with a two run lead, his runs allowed during that inning go into the "up by two runs" category.
Using the retrosheet databases, I looked at every inning pitched by every starting pitcher between 1977 and 1994. The results are shown in Table 1 below.

Table 1: MLB ERA by Score - 1977-1994

total

up 7+

up 6

up 5

up 4

up 3

up 2

up 1

IP

458,407

11,702

7,897

12,618

19,252

29,734

44,332

67,341

ERA

4.01

4.19

3.98

4.16

4.05

3.98

3.97

3.93


tied

down 1

down 2

down 3

down 4

down 5+

IP

167,082

50,392

27,688

13,438

5,057

1,376

ERA

3.98

3.93

4.16

4.22

4.10

4.09



The table reveals that the overall ERA for starting pitchers was 4.01 and also that the lowest ERAs did indeed come in innings where the margin was small: 3.98 in tie games, 3.93 when up or down by one run, etc. In contrast, pitchers had worse ERAs in innings where they were behind by a significant margin: 4.10 when down by four, 4.09 when down by five or more, etc. That should be expected though because a pitcher who is behind by a lot of runs must not have been pitching well prior to the inning and would tend to continue to struggle going forward.

More interestingly, the ERAs in innings where pitchers had big leads were also relatively high: 4.05 with four run leads, 4.16 with five run leads, etc. The one exception was a 3.98 ERA with a six run lead but that appears to be just a blip in the data. The overall trend points to pitchers doing better in close games.

Table Two presents data comparing the performance of pitchers in close games (tied, down by one, up by one) and blow outs (leading by five or more runs). I computed a ratio as follows:

(ERA in close games/ERA in blow outs) X 100

A ratio of 100 would mean that pitchers performed the same in innings where the margin was small as in innings where they had big leads. A ratio under 100 would indicate that they did better when the score was close, whereas a ratio of over 100 would indicate a better ERA in innings beginning with a large lead.


Table 2: MLB ERA - Up by five runs versus within one run
(1977-1994)

total ip

total ERA

IP up 5+ runs

ERA up 5+ runs

IP within 1 run

ERA within run

ratio

458,407

4.01

32,218

4.12

284,815

3.96

96



For the period 1997-1994, we get (3.96/4.12) x 100 = 96. That means that pitchers pitched 4% better in close games than in blow outs.

Next, I looked at individual pitchers and that's why I included so many years in the study. Just looking at one year or a couple years of data wasn't going to work because the sample size for innings pitching with a lead of five or more runs would be too small. So, for this part of the analysis, I selected 46 pitchers with 2,000 or more innings pitched during the period. These data are in Table 3 below.

Not all pitchers pitched better in close games. In fact, 21 0f the 46 pitched as well or better in blowouts. The ratios range from 59 (41% better in close situations) for Ron Darling to 140 (40% better in blow outs) for Jerry Reuss. This gives us some idea of what the performance of a pitcher who pitches to the score might look like - probably more like Darling than Reuss


Jack Morris had a ratio of 91 indicating that he was 9% better in close games than he was in games where he had a large lead. This is better than league average but not unusual - 12th out of 46 pitchers. What if Morris had pitched the same in blow outs as he did in close games? Well, he would have given up 12 fewer runs and finished with a career ERA of 3.87 rather than 3.90. That's hardly an earth shattering difference.

Table 3: Individual pitcher ERA - Up by five runs versus within one run

first

last

total ip

total ERA

IP up 5+ runs

ERA up 5+ runs

IP within 1 run

ERA within run

ratio

Ron

Darling

2,219

3.76

122

6.11

1,401

3.61

59

Bill

Gullickson

2,544

3.91

208

5.58

1,519

3.57

64

Mark

Langston

2,442

3.74

134

5.05

1,502

3.49

69

Frank

Viola

2,791

3.67

189

4.86

1,707

3.41

70

Tommy

John

2,300

3.71

176

4.90

1,350

3.51

72

Rick

Sutcliffe

2,565

4.10

203

5.13

1,466

4.00

78

Ron

Guidry

2,302

3.30

219

3.99

1,294

3.12

78

Rick

Rhoden

2,276

3.68

163

4.13

1,410

3.52

85

David

Stewart

2,238

3.85

144

4.44

1,299

3.87

87

Bob

Forsch

2,148

3.84

182

4.11

1,298

3.61

88

Mike

Krukow

2,164

3.90

138

4.25

1,386

3.84

90

Jack

Morris

3,746

3.90

320

4.11

2,178

3.76

91

Fernando

Valenzuela

2,417

3.42

149

3.50

1,510

3.19

91

Frank

Tanana

3,308

3.91

305

4.39

1,952

4.02

92

Dennis

Eckersley

2,121

3.81

160

4.16

1,277

3.86

93

Mike

Moore

2,687

4.25

161

4.59

1,639

4.27

93

Steve

Carlton

2,488

3.30

251

3.19

1,510

2.99

94

Bret

Saberhagen

2,004

3.20

143

3.41

1,239

3.23

95

Doyle

Alexander

2,493

3.88

177

3.91

1,577

3.73

96

Kevin

Gross

2,001

3.96

133

4.06

1,250

3.90

96

Dennis

Martinez

3,341

3.59

244

3.77

2,006

3.63

96

Tom

Seaver

2,062

3.37

162

3.49

1,252

3.38

97

Rick

Reuschel

2,417

3.28

167

3.35

1,535

3.31

99

Jimmy

Key

2,032

3.34

200

3.51

1,230

3.48

99

Don

Sutton

2,465

3.52

157

3.50

1,525

3.48

99

Joe

Niekro

2,269

3.51

206

3.53

1,381

3.56

101

Orel

Hershiser

2,089

2.99

159

2.89

1,288

2.92

101

Scott

Sanderson

2,380

3.84

171

3.78

1,452

3.86

102

Charlie

Hough

2,999

3.82

185

3.79

1,797

3.90

103

Roger

Clemens

2,391

2.94

203

2.88

1,368

2.97

103

Floyd

Bannister

2,298

4.04

143

4.09

1,416

4.24

104

Mike

Witt

2,021

3.89

189

3.62

1,196

3.84

106

Bob

Welch

3,013

3.45

173

3.17

1,918

3.38

107

Mike

Boddicker

2,041

3.81

172

3.41

1,198

3.64

107

Mike

Flanagan

2,537

3.89

197

3.65

1,608

3.96

108

Bob

Knepper

2,611

3.71

151

3.33

1,616

3.62

109

Jim

Clancy

2,373

4.19

167

4.00

1,473

4.37

109

Mike

Scott

2,026

3.53

119

3.09

1,342

3.43

111

Dwight

Gooden

2,163

3.10

187

2.69

1,328

3.01

112

Phil

Niekro

2,626

3.76

206

3.27

1,555

3.84

118

David

Stieb

2,809

3.42

191

3.02

1,746

3.56

118

Bruce

Hurst

2,371

3.83

228

3.20

1,459

3.79

118

Bert

Blyleven

3,055

3.63

268

3.02

1,778

3.73

123

Scott

McGregor

2,003

3.99

188

3.02

1,199

4.11

136

Nolan

Ryan

3,451

3.24

203

2.35

2,177

3.26

139

Jerry

Reuss

2,032

3.58

126

2.50

1,210

3.51

140



Now, let's try a different criteria for a big lead - up by 4 or more runs. The major league results for 1977-1994 are displayed in Table 4. Pitchers pitched 3 percent better in close games (3.96 ERA) than they did when they had a lead of four runs or more (4.10).

Table 4: MLB ERA - Up by four runs versus within one run

IP up 4+ runs

ERA up 4+ runs

IP within 1 run

ERA within run

ratio

51,469

4.10

284,815

3.96

97



Table 5 lists the individual pitchers and reveals that Jack Morris had a 3.92 ERA in games where he had a big lead. That yielded a ratio of 96 which was 23rd out of 46 pitchers. So, by this measure, Jack was right around league average in terms of pitching to the score.

Table 5: Individual pitcher ERA - Up by four runs versus within one run

first

last

total ip

total ERA

ip_up4_

era_up4_

IP within 1 run

ERA within run

ratio4

Ron

Darling

2,219

3.76

214

5.37

1,401

3.61

67

Bill

Gullickson

2,544

3.91

351

4.85

1,519

3.57

74

Tommy

John

2,300

3.71

278

4.76

1,350

3.51

74

Frank

Viola

2,791

3.67

329

4.30

1,707

3.41

79

Mark

Langston

2,442

3.74

236

4.35

1,502

3.49

80

Dennis

Martinez

3,341

3.59

389

4.47

2,006

3.63

81

Fernando

Valenzuela

2,417

3.42

253

3.88

1,510

3.19

82

Kevin

Gross

2,001

3.96

210

4.58

1,250

3.90

85

Bob

Forsch

2,148

3.84

277

4.20

1,298

3.61

86

Rick

Sutcliffe

2,565

4.10

343

4.61

1,466

4.00

87

Ron

Guidry

2,302

3.30

346

3.59

1,294

3.12

87

Mike

Krukow

2,164

3.90

224

4.38

1,386

3.84

88

David

Stewart

2,238

3.85

267

4.35

1,299

3.87

89

Mike

Scott

2,026

3.53

201

3.72

1,342

3.43

92

Rick

Rhoden

2,276

3.68

268

3.79

1,410

3.52

93

Rick

Reuschel

2,417

3.28

240

3.56

1,535

3.31

93

Dwight

Gooden

2,163

3.10

285

3.19

1,328

3.01

94

Scott

Sanderson

2,380

3.84

289

4.08

1,452

3.86

95

Mike

Moore

2,687

4.25

277

4.51

1,639

4.27

95

Bret

Saberhagen

2,004

3.20

235

3.41

1,239

3.23

95

Don

Sutton

2,465

3.52

290

3.66

1,525

3.48

95

Doyle

Alexander

2,493

3.88

297

3.91

1,577

3.73

95

Jack

Morris

3,746

3.90

507

3.92

2,178

3.76

96

Jim

Clancy

2,373

4.19

237

4.51

1,473

4.37

97

Bob

Knepper

2,611

3.71

279

3.68

1,616

3.62

98

Frank

Tanana

3,308

3.91

450

4.08

1,952

4.02

99

Joe

Niekro

2,269

3.51

289

3.58

1,381

3.56

99

Jimmy

Key

2,032

3.34

284

3.49

1,230

3.48

100

Steve

Carlton

2,488

3.30

374

2.98

1,510

2.99

100

Bruce

Hurst

2,371

3.83

306

3.71

1,459

3.79

102

Floyd

Bannister

2,298

4.04

244

4.13

1,416

4.24

103

Orel

Hershiser

2,089

2.99

229

2.83

1,288

2.92

103

Dennis

Eckersley

2,121

3.81

262

3.71

1,277

3.86

104

Mike

Witt

2,021

3.89

259

3.68

1,196

3.84

104

Scott

McGregor

2,003

3.99

298

3.93

1,199

4.11

105

Charlie

Hough

2,999

3.82

336

3.69

1,797

3.90

106

Roger

Clemens

2,391

2.94

358

2.79

1,368

2.97

107

Mike

Boddicker

2,041

3.81

277

3.42

1,198

3.64

107

Tom

Seaver

2,062

3.37

256

3.13

1,252

3.38

108

Phil

Niekro

2,626

3.76

298

3.54

1,555

3.84

109

Bob

Welch

3,013

3.45

299

3.04

1,918

3.38

111

Mike

Flanagan

2,537

3.89

318

3.54

1,608

3.96

112

Nolan

Ryan

3,451

3.24

329

2.65

2,177

3.26

123

Jerry

Reuss

2,032

3.58

234

2.81

1,210

3.51

125

Bert

Blyleven

3,055

3.63

402

2.98

1,778

3.73

125

David

Stieb

2,809

3.42

313

2.59

1,746

3.56

137



For those that need more convincing, I tried one more cutoff - 3 or more runs. Table 6 shows that major league pitchers had an ERA of 4.05 when up by three or more runs. That gave a ratio of 98. I have too many tables already so I won't add another big one but Morris had a 4.03 ERA in 798 innings where he had a lead of three runs or more. That comes out to a ratio of 96 which is 16th out of 46 pitchers.

If Morris had pitched the same way in those 798 innings as he did in close games, he would have allowed 23 fewer runs and had a career ERA of 3.85. Again, not a radical difference from 3.90.

Table 6: MLB ERA - Up by three runs versus within one run

IP up 3+ runs

ERA up 3+ runs

IP within 1 run

ERA within run

ratio

81,204

4.05

284,815

3.96

98



In conclusion, the typical pitcher did pitch better in close games than he did when he had a big lead. That is, pitchers did have a tendency to pitch to the score. There is some evidence that Jack Morris pitched to the score a little more than the typical pitcher but he was not extraordinary in that respect. He came nowhere close to pitchers like Ron Darling, Bill Gullickson and Tommy John who pitched at 25%-40% better with large leads than they did in close games. There is also no evidence that Morris would have had a significantly better career ERA had he not pitched to the score.
The information used here was obtained free of charge from and is copyrighted by
Retrosheet. Interested parties may contact Retrosheet at "www.retrosheet.org".

10 comments:

  1. Love you work. Thanks for all the effort. I have an off topic question. Do you plan on doing an averageing of all the different defensive metrics again this year? If so, when can we expect to see it? I for one very much appriciateed it the last two years and hope you plan on continuing to do it and refine it.

    ReplyDelete
  2. Yes, I plan to do that. I'm currently collecting the data and then I need to merge it. Merging is the most time consuming part because different databases use different player names. I'll start posting the results some time in January.

    The main refinement this year will be that the UZR data is now fully available on fan graphs. Last year, I was only able to include leaders and trailers.

    Lee

    ReplyDelete
  3. I was thinking about that Sheehan study a couple of weeks ago when the HOF ballot was in the news, but I wouldn't have known where to find it. Thanks for the reprise and for takng it to the next level with the comparison to all the other starters.

    I'm wondering if WHIP might be a good test of "pitching to the score"? If you have a big lead and just throw fastballs for strikes, perhaps you give up more baserunners but not significantly more runs. But wait, you should give up fewer BB's if you are not trying to hit the corners. Maybe your BB/9 actually goes down with a big lead but your BA against goes up? I have to stop thinking about this now.

    ReplyDelete
  4. Thanks! I look forward to reading them.

    By the way, I understand the UZRs on Fangraphs are with BIPS Data rather then STATS Data. I Think there are still UZRs available somewhere with the STATS Data I just have no idea where or how to find them. I would be very interested in how these two compare if you have access to both. And of course I would like to see both BIPS UZRs and STATS UZRs included in you analysis if you have access to both.

    Thanks again for all the hard work. I think it does have real value in better understanding the defensive abilities of the MLB players we all love so much.

    ReplyDelete
  5. The problem with this study is that a larger run differential is more likely to happen in a league/park/etc. where run scoring is more common. (It's also more likely to be facing a high run differential in later innings, when a starting pitcher typically does not pitch as well.) You have to compare the runs a pitcher would allow that year, in that league, and in that park ignoring the run differential, compared to how he gave up runs given the run differential.

    ReplyDelete
  6. Good points Charles. Walk rate and strikeout rate would be good next steps.

    Colin, I will keep your points in mind as well, although they may go beyond what I have time to do. I can't spend a lifetime denigrating one of my childhood heroes. :-)

    Lee

    ReplyDelete
  7. Ditto what Colin said. You have to control for park, inning, and general run environment to determine whether all pitchers as a group pitch to the score.

    I would guess that once you do that, most of the difference will disappear. I would also think to some extent all pitchers DO pitch to the score. Why would you want to expend a lot of energy or risk injury when you are up or down by a lot of runs?

    And if you are up by a lot of runs, you pitch differently, such that you avoid a big inning and sacrifice a few runs (your ERA goes up a little, but the variance shrinks).

    On the other hand, you would NOT expect the "pitching to the score" to be symmetrical. You would expect that when a pitcher's team is up by a lot of runs that his ERA climbs, but that when down by a lot of runs, it should actually go down, similar to when the score is close. The fact that we see a lot of symmetry (ERA climbing when up OR down by a lot of runs) suggests that something else is going on that is not in the pitcher's control, like the run environment.

    Even if we find that pitchers as a whole do or do not tend to pitch to the score, what is the point at looking at individual pitchers, like Morris?

    For one thing, it might help us to reconcile ERA with win/loss record. Win/loss record is principally a function of four things: Whether and by how much a pitcher pitches to the score, the offense of his team, how many IP per game he pitches, and his pen. (And random fluctuation of course.)

    If a pitcher's w/l record is not congruent with his ERA, even given his team's offense (rpg) AND we find that he does indeed pitch to the score more so than the average pitcher, then we can rightfully give him credit for that is an MVP or HOF type of evaluation. If it turns out that he did not pitch to the score over and above the average pitchers, and we already account for his run support, then there is a better chance that something else beyond his control was going on, like random variation or his bullpen.

    So I suppose it helps to know whether a pitcher like Morris did or did not pitch to the score any better or worse than the average starter, especially if someone is claiming that he did or did not.

    Now, just because we find that Morris or anyone else did or did not pitcher to the score more or less than the average pitcher, or just because we show an interesting spread or range of pitching to the score, as lee did in this blog post, that does NOT mean that it is a skill. You will always find variation in any performance measure whether there is an element of skill involved in the measure or not (I should add whether is a spread of true talent with respect to that skill or not, if there is skill involved).

    That would be one of the next interesting steps - to take a list like Lee has generated and to see whether the distribution is significantly different than that expected by chance or see if individual players' "ratio" persists (correlates) from one time period to another.

    As with most of these things, I would guess that there is very little correlation even with fairly large samples of data used for each observation (like year to year).

    MGL

    ReplyDelete
  8. MGL, Thanks for the long critique. Everything you said makes sense. There is just one point where I disagree. For starting pitchers, I don't think you would expect ERA to go down when they are behind by a lot of runs. If a pitcher is behind by a lot of runs, it probably means he hasn't been pitching well that day and might be expected to continue having problems.

    ReplyDelete
  9. I enjoyed reading this. Like other readers I was surprised by the symmetry.

    Sorry to confound things further, but let me play the skeptic.

    Suppose as a manager I have a pitcher having a so-so, or bad day. If the run differential is large I might leave the pitcher out there to rest the pen if they need it. In a close game I would probably replace the pitcher regardless of the pen's shape. So, pitchers might only be pitching in close games when they are having a good (low ERA) day.

    ReplyDelete
  10. Jeff,I agree that bullpen management could play a role in this as well. There are so many variables. I do think I'm going to try this again in a few weeks controlling for run environment and inning and see what kind of results I get.

    ReplyDelete

Twitter

Blog Archive

Subscribe

My Sabermetrics Book

My Sabermetrics Book
One of Baseball America's top ten books of 2010

Other Sabermetrics Books

Stat Counter