Masters and Victims: Pitchers and Unearned Runs – Part 2

When it comes to allowing unearned runs, are pitchers masters of their own fate, or innocent victims of circumstance (or, perhaps, not so innocent)? That is the question I’ll look at in this second installment on unearned run prevention.

In Part 1 of this series, I looked at what evidence there might be to support the hypothesis that unearned run avoidance was a skill that some pitchers could consistently demonstrate over their careers. The major conclusion was that, since 1961, team ERA and team unearned run prevention showed positive correlation. That is, of those teams successful in limiting unearned runs, the proportion with better pitching (i.e. lower ERAs) was higher than for groups of teams having less success in preventing unearned runs. This conclusion also seems intuitively correct – pitchers who are good at preventing earned runs are probably also good at preventing unearned runs.

In Part 2, I’ll take the next step and try to identify those pitchers who seem most and least adept at the skill of preventing unearned runs. Yes, I called it a skill because, while some errors lead immediately and unavoidably to runs scoring, many times the consequences are not so dire, affording pitchers the opportunity to work out of jambs caused by their defense. Whether because of temperament or determination, some pitchers seem to do this quite well, while others … well, not so much.

After the jump, more on pitchers and unearned runs.

Unearned Runs – a “treasure” to be shared

To properly evaluate a pitcher’s ability to prevent or limit unearned runs scoring, it would be appropriate to consider the “opportunities” that a pitcher has to practice this skill – a euphemism for how frequently a pitcher is victimized by his teammates’ (or his own) errors. Unfortunately, I haven’t been able to find these data. So, I will instead use estimation techniques to derive how many unearned runs individual pitchers on a given team would be expected to give up based on the team’s total unearned runs allowed. In Part 3, I’ll look at whether a team’s unearned runs seem low or high relative to the number of errors committed, and then see if those indicators correlate with the presence or absence of pitchers skilled in limiting unearned runs.

To quantify pitchers’ abilities in preventing unearned runs, I chose to view a team’s unearned runs as a pool of runs to be shared across the pitching staff. This premise is based on the notion that a team is always trying to play at its best, but that, unfortunately, errors still occur and can happen at any time. Thus, pitchers don’t cause the defense to commit errors, and the fact that some pitchers happen to have to deal with more errors than other pitchers on the same team is as much a matter of luck as anything else.

The premise in the preceding paragraph is fine as far as it goes. However, while pitchers may not cause errors, their pitching may influence the frequency of errors committed behind them. Pitchers who limit balls in play generally stand a better chance of having fewer errors committed behind them than do pitchers who have more “exciting” innings. The strikeout pitcher who is also stingy with hits and walks is probably ideally suited to limiting the errors his defense makes. In Part 3, I’ll more into what type of pitcher is best suited to limiting unearned runs. But for now, let’s get back to that concept of unearned runs as a team commodity to be shared among the pitchers.

Unearned Run Allocation

So, how should unearned runs be shared among pitchers? I tried two different methods that I’ll describe this way:

(1) Workload-based allocation

• a team’s unearned runs are shared among the team’s pitchers according to the proportion of the team’s innings that each pitcher worked
• pitchers who work more innings are allocated more unearned runs, and those working fewer innings are allocated fewer unearned runs
• pitchers with similar workloads are allocated similar numbers of unearned runs, regardless of the relative skill of such pitchers

(2) Skill-based allocation

• a team’s unearned runs are shared among the team’s pitchers according to the proportion of the team’s earned runs that each pitcher allowed
• good pitchers are “rewarded” with a lower allocation of unearned runs; bad pitchers are “penalized” with a higher allocation of unearned runs

Unearned Runs Saved

In both allocation methods, what is of interest is not the allocation itself, but the difference between the allocation and the actual unearned runs that a pitcher did allow, which I will call unearned runs saved or URS. Thus, a negative result is good, indicating a pitcher allowed fewer unearned runs than the portion of unearned runs allocated to him. Similarly, a positive result indicates a pitcher whose actual unearned runs allowed were more than the portion of the team’s unearned runs allocated to him.

So, what am I trying to measure here? The workload-based allocation is simply measuring the degree of deviation from a uniform world where all pitchers are equal and all will be equally affected by bad defense behind them. But, all pitchers are not equal, and the results from Part 1 of the study suggest that better pitchers should be more effective at preventing unearned runs. Therefore, the difference between actual and allocated unearned runs will measure how much better or worse pitchers are in unearned run prevention.

Responding to Misfortune

The second allocation is trying to measure something a bit more subtle. In a sense, it’s trying to get into the mind of the pitcher and gauge how he reacts to misfortune, specifically the misfortune of errors committed behind him. For example:

1. One extreme is the pitcher who is upset with the mistakes of his teammates and loses his cool, affecting his performance in pitching to subsequent batters. Another variant of this extreme is the pitcher who feels sorry for himself and allows his concentration and effort to waver, secure in the self-fulfilling knowledge that “it just isn’t my day”.
2. The other extreme is the pitcher who empathizes with the teammate who made the error, and wants to “pick him up” by exerting greater effort and determination to overcome the misfortune, and make even better pitches to subsequent hitters.
3. In the middle are the professionals who just go about their business. Can’t do anything to change the past, so let’s get on with business and get the next out. The only pitch that matters is the next one. Insert your favorite hackneyed aphorism here. 

So, you’re saying “Okay, I know what you’re talking about, but how the heck are you going to measure that?”. Well, as it turns out, the second skill-based allocation, while obviously not directly measuring subjective qualities, does provide some indication of which of those three types a given pitcher might be. That indication can then be “tested” according to anecdotal or even observational evidence for given pitchers

What the skill-based allocation will tell you is, when faced with adversity caused by his team’s defense, does a given pitcher perform better than he normally does, about the same, or worse? Thus, I’ve called this metric Clutch Unearned Runs Saved or CURS, indicating that overcoming the adversity of poor defense is indeed a form of “clutch” pitching.

You will recall that I said that, under skill-based allocation, a pitcher is rewarded by having a lower unearned run allocation. Except, I enclosed “rewarded” in quotations because it isn’t really a reward since what we’re looking at is the difference between actual unearned runs and allocated unearned runs. So, the smaller the allocation, the harder it will be to go even lower. But, that’s reality – if you’re a good pitcher, it’s tough to be even better. In practice, the pitchers who have the best opportunity to do well here are the inferior pitchers of the type 2 variety – with a big unearned run allocation, they have lots of room to improve.

Results

Most pitchers have a small number of unearned runs charged to them each year, though occasionally there can be quite a few more in a given season. The point is that you aren’t going to learn much from looking at a season’s worth of data – the likelihood is too high that what you’re seeing is just good or bad luck. However, over time, luck tends to balance out so that what remains provides an indication of a pitcher’s skill in preventing unearned runs. Thus, the data I’ll be showing you are for entire careers, or portions of careers, for the 589 pitchers with at least 1000 IP since 1961. All data have been extracted from the Fangraphs database.

Here are the moments for our URS and CURS metrics. I’ve converted URS and CURS into per 9 IP quantities, but the distribution is for unweighted values. Thus, all pitchers are equally weighted, regardless of innings pitched.

As you can see, there are very similar distributions for both the workload-based and skill-based allocation methods. Also, the mean for the subset of pitchers with 1000+ IP is almost bang on the true mean of zero for the entire population of all pitchers.

Below are charts visually depicting the two distributions.

The table below shows you all 589 of these pitchers. You can search for a pitcher by entering his name in the Search box, or sort by any of the listed fields. These are:

Unearned Runs Saved (URS) - these are the career totals since 1961 of the difference between a pitcher’s actual unearned runs and the unearned runs allocated to him via the workload-based allocation described above. Negative results are good, indicating a pitcher who allowed fewer unearned runs than those allocated to him.

Clutch Unearned Runs Saved (CURS) – this may not be the best name for this metric, but these are the career totals since 1961 of the difference between a pitcher’s actual unearned runs and the unearned runs allocated to him via the skill-based allocation described above. Negative results, again, are good, indicating a pitcher who allowed a lower proportion of his team’s unearned runs than of his team’s earned runs. A pitcher doing this consistently is probably pitching better in these “clutch” situations (i.e. after an error committed behind him) than he does overall.

As indicated above, pitchers who do well here tend to be of the inferior variety who have lots of room to improve whereas a good pitcher is probably doing well with a score near zero, indicating his normal level of solid performance is maintained in these clutch situations. Probably the most interesting result are the good pitchers who do poorly here – this may be an indicator of those pitchers who do not respond well to errors committed behind them.

URS/162 and CURS/162 - the same metrics as above, but shown per 162 IP as an approximation of runs saved (or not) over the course of a season

URS Rank, CURS Rank - ranking among these 589 pitchers for URS/162 and CURS/162.

So, what are these numbers telling us? Because all of the URS totals are calculated for each player, season by season (and within a season, by team), you are seeing how well a pitcher did relative to other pitchers on the same team. So, comparing pitchers on different teams to each other is comparing how well each did in his own context. Which, when you come down to it, is what really matters. For example, in preventing unearned runs, this metric would show whether Andy Pettitte helped the Yankees more than Roy Halladay helped the Phillies, not whether or by how much either was better or worse  than a major-league average pitcher, whatever that might be. It’s important to bear this type of comparison in mind when looking at scores for different pitchers.

Thus, the runs saved or not saved are indicative of how much a pitcher contributed to his team(s) over the course of his career. What these numbers are NOT telling you is how well a pitcher would have done had he played on another team. To compare pitchers to each other directly is probably most meaningful among pitchers who played a large portion of their careers on the same team, in front of the same defense. Thus, Smoltz, Glavine and Maddux could be a good group, or Koufax and Drysdale, Marichal and Gaylord Perry, etc.

Some observations on this measurement approach:

• Large URS numbers (20 or more, either positive or negative) are meaningful. Remember this is a zero sum method. All pitchers on a team sum to zero in every year. For a pitcher to compile numbers of that magnitude means they are consistently posting positive or negative scores in most seasons, something quite unlikely if this was all about luck. This is more especially true, still, if a pitcher played for different teams of varying quality through the course of his career.
• While the yearly run totals are, as explained, very context-sensitive, the summation of these run totals is not normalized. Thus, the win value of a run at different points in a player’s career is not factored into the career totals, nor are differences in win values of a run within a season owing to park effects. These would be useful enhancements to this approach.
• Remember that an average CURS score (i.e. near zero) is not bad. It just means the pitcher performed similarly with regard to allowing both earned or unearned runs. So, for example, Sandy Koufax was -24.3 URS, but only 1.2 CURS. That means he did very well (top 10%) in saving unearned runs, and allowed about equal shares of his teams’ earned and unearned runs (which, for a player with a 156 ERA+ during the years of this study, is very, very good indeed).
• URS does NOT love knuckleballers. Which kind of makes sense. Knuckleballs tend to result in more errors and more passed balls, both of which can lead to unearned runs. Thus, Phil Niekro, Tom Candiotti, Wilbur Wood, Charlie Hough and Tim Wakefield are all near the bottom in URS, in or very close to the lowest 10%. Tim Wakefield is dead last in URS among these 589 pitchers and his 64.3 URS total is 10 runs (okay, it’s 9.9) clear of number 588. I think you definitely want to look at RA more than ERA when evaluating a knuckleball pitcher.

The King of URS

So what about the guy at the top of our list? Not only did Curt Schilling rank first in career URS, he also ranked first in URS/162, despite pitching over 3000 innings, many more than the majority of pitchers in this study (number 2 in URS/162 is reliever Don McMahon, with just 1061 IP) . Thus, Schilling truly was consistent in compiling good URS scores just about every year (he had negative URS in 18 of 20 seasons), regardless of what team he was on. In addition, Schilling also ranked 13th of 589 in CURS, meaning he allowed a much smaller proportion of his teams’ unearned runs than of his teams’ earned runs – no mean feat considering Schilling’s career 127 ERA+. It would seem Schilling took errors as a challenge, determined not to let his opponents get “cheap” runs as a result. More conventional metrics also agree – Schilling has the lowest career % of runs allowed that are unearned (less than 5%) among all pitchers with 1200 IP since 1901.

Whether aptly named or not, since I had a metric called Clutch URS, I have to tell you about Jack Morris. Actually, Jack does pretty well with scores of -18.1 URS and -16.0 CURS, the result of posting consistent negative scores most of his career. Jack’s totals would have been considerably better (he was -30.3 URS and -22.6 CURS through the 1989 season) had he not posted some high positive scores in his final years.

For those with the inclination to dig deeper, this link will take you to a spreadsheet where you can select pitchers of interest and see their URS and CURS scores, on a season by season basis. To use the tool:

1. Give Google a minute or two to load the data (there’s a lot)
2. In the Report Editor on the right side, scroll to the bottom to the Filter box. Under Filter: Name will be shown the number of items (players) currently being displayed (should be just 1 for best results)
3. Click on the little down arrow beside the number of items, and a selection pop up will appear
4. Click on Clear to clear the current selection
5. Enter the last name of the player of interest in the Search box, and select from the displayed search results
6. Wait a moment or two, and the season by season results for that pitcher will be displayed (looks like below)

I’ll close with this listing of the 100 pitchers with the most innings pitched since 1961 (which just happens to be all pitchers over 2500 IP), sorted by URS.

Name	                     IP	  URS	 CURS  URS/162 CURS/162 URS Rk CURS Rk Combined Rk
Curt Schilling		3261.00   -77.3	 -50.1	-3.84	-2.49	  1	 13	  14
Tom Seaver		4782.67	  -63.2	 -24.8	-2.14	-0.84	 41	166	 207
Don Sutton		5282.00	  -61.1	 -55.4	-1.87	-1.70	 61	 67	 128
Bert Blyleven		4970.00	  -60.9	 -24.0	-1.98	-0.78	 52	174	 226
Roger Clemens		4916.67	  -57.1	  -1.0	-1.88	-0.03	 60	305	 365
Dennis Eckersley	3285.67	  -55.1	 -34.6	-2.72	-1.71	 14	 65	  79
Luis Tiant		3486.33	  -53.5	 -37.9	-2.49	-1.76    22	 60	  82
Fergie Jenkins		4500.67	  -52.6	 -29.8	-1.89	-1.07	 59	141	 200
John Smoltz		3473.00	  -48.5	 -33.3	-2.26	-1.55	 33	 81	 114
Bob Gibson		3722.00	  -48.2	 -18.1	-2.10	-0.79	 46	172	 218
Bret Saberhagen	        2562.67	  -44.9	 -26.5	-2.84	-1.67	 11	 69	  80
Pedro Martinez		2827.33	  -43.1	  -6.3	-2.47	-0.36	 24	250	 274
Javier Vazquez		2840.00	  -36.6	 -30.9	-2.09	-1.76	 47	 59	 106
Doug Drabek		2535.00	  -33.6	 -28.6	-2.14	-1.83	 39	 50	  89
Livan Hernandez	        3189.00	  -33.4	 -37.9	-1.70	-1.92	 73	 44	 117
Tom Glavine		4413.33	  -31.0	 -20.3	-1.14	-0.75	148	179	 327
Catfish Hunter		3449.33	  -30.9	 -28.2	-1.45	-1.33	106	103	 209
Jim Perry		2871.33	  -30.7	 -29.5	-1.73	-1.67	 71	 70	 141
Vida Blue		3343.33	  -29.4	 -21.6	-1.42	-1.05	110	143	 253
Rick Reuschel		3548.33	  -28.1	  -3.7	-1.28	-0.17	135	285	 420
Dave McNally		2730.00	  -27.5	 -32.7	-1.63	-1.94	 80	 39	 119
Mike Mussina		3562.67	  -25.9	  -6.8	-1.18	-0.31	143	259	 402
Kevin Appier		2595.33	  -25.6	 -11.9	-1.60	-0.74	 86	180	 266
Dwight Gooden		2800.67	  -25.0	 -16.5	-1.44	-0.95	108	152	 260
Bob Welch		3092.33	  -24.5	 -18.6	-1.28	-0.98	134	150	 284
Doyle Alexander	        3367.67	  -24.0	 -28.5	-1.15	-1.37	146	 94	 240
David Wells		3439.00	  -23.9	 -16.7	-1.12	-0.79	150	173	 323
Andy Benes		2505.33	  -23.6	 -21.8	-1.53	-1.41	 96	 91	 187
Steve Rogers		2837.67	  -23.1	  -6.7	-1.32	-0.39	127	241	 368
Frank Tanana		4188.33	  -22.9	  -0.4	-0.89	-0.02	192	309	 501
Jim Bunning		2667.33	  -22.5	  -9.8	-1.37	-0.59	122	206	 328
John Candelaria	        2525.67	  -22.5	 -11.9	-1.44	-0.77	109	175	 284
Nolan Ryan		5386.00	  -22.1	   6.7	-0.67	 0.20	230	338	 568
Rick Sutcliffe		2697.67	  -21.6	 -20.0	-1.30	-1.20	132	124	 256
Gaylord Perry		5350.33	  -20.9	  17.9	-0.63	 0.54	235	391	 626
Mickey Lolich		3638.33	  -20.6	 -13.3	-0.92	-0.59	181	207	 388
Tim Hudson		2682.33	  -19.0	  -3.4	-1.14	-0.21	147	277	 424
Mark Langston		2963.00	  -18.2	  -6.4	-0.99	-0.35	169	254	 423
Jack Morris		3824.00	  -18.1	 -16.0	-0.77	-0.68	214	190	 404
Dave Stewart		2629.67	  -16.6	 -18.9	-1.02	-1.16	163	127	 290
David Cone		2898.67	  -16.4	  -2.8	-0.92	-0.15	182	287	 469
Danny Darwin		3016.67	  -15.7	  -8.6	-0.84	-0.46	197	232	 429
Stan Bahnsen		2529.00	  -14.9	 -20.3	-0.96	-1.30	175	106	 281
Mel Stottlemyre	        2661.33	  -14.9	  -4.2	-0.91	-0.26	186	266	 452
Burt Hooton		2652.00	  -14.7	 -16.2	-0.90	-0.99	189	149	 338
Mike Flanagan		2770.00	  -14.2	 -17.8	-0.83	-1.04	198	144	 342
Dave Stieb		2895.33	  -12.4	   3.6	-0.70	 0.20	227	339	 566
Mike Moore		2831.67	  -12.0	 -16.5	-0.69	-0.94	229	153	 382
CC Sabathia		2564.33	  -11.8	   6.5	-0.75	 0.41	219	367	 586
Milt Pappas		2626.67	  -11.7	  -9.5	-0.72	-0.59	223	210	 433
Jim Palmer		3948.00	  -10.9	   7.4	-0.45	 0.31	272	354	 626
Claude Osteen		3400.33	  -10.7	  -7.8	-0.51	-0.37	258	247	 505
John Burkett		2648.33	   -9.4	 -13.9	-0.58	-0.85	247	165	 412
Jerry Koosman		3839.33	   -9.0	  -3.6	-0.38	-0.15	281	288	 569
Bill Gullickson	        2560.00	   -8.9	  -9.5	-0.56	-0.60	250	205	 455
Rick Rhoden		2593.67	   -7.7	  -8.0	-0.48	-0.50	264	228	 492
Frank Viola		2836.33	   -6.9	   4.6	-0.39	 0.26	277	346	 623
Mike Morgan		2772.33	   -6.8	  -9.8	-0.40	-0.57	276	211	 487
Jamie Moyer		4074.00	   -6.3	  -1.2	-0.25	-0.05	299	302	 601
Jimmy Key		2591.67	   -4.8	   6.1	-0.30	 0.38	292	362	 654
Roy Halladay		2687.33	   -3.6	  17.0	-0.22	 1.02	306	460	 766
Juan Marichal		3426.00	   -2.8	  29.9	-0.13	 1.41	317	496	 813
Steve Trachsel		2501.00	   -1.0	   0.4	-0.06	 0.03	321	310	 631
Scott Sanderson	        2561.67	   -0.3	  -2.0	-0.02	-0.13	329	290	 619
Kevin Millwood		2720.33	    0.7	  -0.8	 0.04	-0.05	338	303	 641
Fernando Valenzuela	2930.00	    0.9	  -3.4	 0.05	-0.19	340	281	 621
Jim Slaton		2683.67	    1.4	  -1.7	 0.08	-0.10	341	295	 636
Randy Johnson		4135.33	    2.4	  43.4	 0.09	 1.70	343	515	 858
Jim Clancy		2517.33	    4.2	  -5.5	 0.27	-0.35	362	253	 615
Chuck Finley		3197.67	    5.2	  24.5	 0.26	 1.24	361	485	 846
Bob Forsch		2794.67	    6.9	   2.5	 0.40	 0.14	379	330	 709
Dennis Martinez	        3999.67	    8.0	   7.8	 0.32	 0.32	372	356	 728
Orel Hershiser		3130.33	    9.0	  18.2	 0.47	 0.94	393	450	 843
Greg Maddux		5008.33	    9.2	  46.0	 0.30	 1.49	370	500	 870
Bob Knepper		2708.00	   10.1	   4.4	 0.60	 0.27	417	347	 764
Joe Coleman		2569.33	   10.4	  10.1	 0.66	 0.64	426	405	 831
Steve Carlton		5217.33	   10.8	  37.7	 0.33	 1.17	373	475	 848
Joe Niekro		3584.00	   11.2	  10.8	 0.50	 0.49	400	381	 781
Mark Buehrle		2679.00	   11.2	  24.9	 0.68	 1.51	429	503	 932
Mike Torrez		3044.00	   11.3	  -1.0	 0.60	-0.06	415	301	 716
Terry Mulholland	2576.00	   11.9	   6.6	 0.75	 0.42	440	368	 808
Jeff Suppan		2542.67	   11.9	   7.4	 0.76	 0.47	442	377	 819
Kenny Rogers		3302.67	   12.6	  28.4	 0.62	 1.39	420	494	 914
Phil Niekro		5404.33	   15.2	  61.9	 0.45	 1.85	388	534	 922
Tom Candiotti		2725.00	   17.3	  30.0	 1.03	 1.78	482	525	1007
Ken Holtzman		2867.33	   17.9	  15.8	 1.01	 0.89	480	445	 925
Rick Wise		3127.00	   18.8	  21.3	 0.97	 1.11	476	464	 940
Kevin Brown		3256.00	   19.5	  50.3	 0.97	 2.50	474	559	1033
Rudy May		2622.00	   21.1	  18.9	 1.30	 1.17	503	474	 977
Derek Lowe		2658.33	   22.2	  24.0	 1.35	 1.46	510	498	1008
Wilbur Wood		2684.00	   23.9	  42.1	 1.44	 2.54	519	561	1080
Jerry Reuss		3670.00	   25.0	  20.4	 1.10	 0.90	489	446	 935  
Mike Cuellar		2804.00	   25.3	  31.2	 1.46	 1.80	522	531	1053
Charlie Hough		3801.00	   25.4	  31.0	 1.08	 1.32	487	491	 978
Paul Splittorff	        2554.67	   26.4	  22.9	 1.68	 1.45	535	497	1032
Andy Pettitte		3130.67	   32.1	  40.4	 1.66	 2.09	534	543	1077
Tommy John		4710.33	   41.0	  51.7	 1.41	 1.78	513	522	1035
Jim Kaat		4475.33	   49.9	  56.9	 1.81	 2.06	543	541	1084
Tim Wakefield		3226.33	   64.3	  61.7	 3.23	 3.10	586	577	1163

Subscribe to: RSS feed