Kerry

I've said it before and I'll say it again (right now...aren't you lucky?): a pitcher's won-loss record is the most overrated stat in baseball, bar none, because it's so context-driven and, ultimately, so misplaced. After all, pitchers don't win games, teams do. It has long amazed me that people moan and groan about other context-driven pitching statistics--saves, for instance, and quality starts--as being susceptible to "cheap" occurrences but seem utterly unperturbed by the same problem plaguing W-L record. A pitcher can throw a no-hitter and lose; a pitcher can throw nine perfect innings and get no decision at all. A pitcher can allow, say, seven runs in five innings and win. A relief pitcher can blow a lead and get a win. All of these things have happened in baseball, and the latter two examples happen with stirring regularity, yet no one seems bothered by it. Fair enough. I'm increasingly tired of tilting at this particular windmill. Pitchers' wins and losses are apparently here to stay, evidently too burned into the baseball public's psyche to be susceptible to being dumped in the ash can in which they belong. Oops...sorry, I promised to stop trashing the stat.

How ingrained are wins credited to pitchers? So much so that it's the one stat regarded as a sure-fire pitcher's guarantee of enshrinement in the Hall of Fame: if a pitcher wins 300 games, he's a lock for Cooperstown, and everyone knows it. There is no other guaranteed pitching benchmark. Position players have 3000 hits or 500 home runs to shoot for, but pitchers, while not needing 300 wins for enshrinement (just as position players don't need 3000 hits or 500 home runs), know they're in if they reach the mark. And it's the only such mark that offers that kind of lock.

The central problem with a pitcher's win-loss record is that it's so heavily influenced by what a pitcher's team does on the offensive end of the game, where the pitcher's role is limited or, in the case of the designated hitter enabled American League (1973 to the present), totally irrelevant. This much we all know. A pitcher who allows five runs while his team is scoring nine will be credited with a win, as long as he throws at least the first five innings, leaves the game with the lead, and his team holds that lead to the end. On the other hand, a pitcher who allows one run in a complete game is tagged with a loss if his club gets shut out. There are plenty of ways to get around the idiosyncratic nature of W-L record in analyzing pitchers (ERA comes to mind as a good starting point), but the public is so focused on won-loss record...

For instance, does anyone seriously believe the Roger Clemens would have been the "obvious choice" as the 2001 AL Cy Young Award winner if his record hadn't been 20-3?

Enter "neutral wins" and "neutral losses," which together make for "neutral win-loss record." This stat eliminates the effect of run support and, after adjusting for ballpark effects, tells us, for a given number of decisions, what a pitcher's record could be expected to be given an average amount of run support. For instance...Roger Clemens finished 20-3 last year, and pitched well, for the most part, posting a 3.51 ERA in a league with a 4.48 earned run average. That's good. But 20-3? Clemens benefited from exceptional run support--6.6 runs per game...the Yankees team average was 5.0 runs per game. What if Clemens had received a league average amount of support (4.86 R/G)? Assuming 23 decisions, he would have gone 14-9 in neutral wins and losses terms. Clemens was clearly a better than average pitcher in 2001, but nowhere near as much as his traditional won-loss record would imply.

Those not interested in the quasi-technical gobbledy-gook should skip this section.

The measure is, loosely put, a comparison of a pitcher's ERA to the league ERA. This provides a benchmark for comparing the player to the league average--the more earned runs a pitcher allows per game compared to the league, the higher a percentage of decisions he can be expected to win. An average pitcher--someone whose park-adjusted ERA matches that of the league as a whole--can be expected to win half his decisions. The concept is, obviously, the fewer earned runs a pitcher allows, the more games he should be expected to win, and the fewer games he should be expected to lose, all other things being equal. Let's look at Clemens/2001 as an example: he posted a 3.51 ERA for the year. Park-adjusting does not significantly change his ERA, so let's go with the raw value for the sake of simplicity. The AL ERA for 2001 was 4.48. 3.51/4.48 is approximately .78, which we will refer to as the Loss Index. The average mark is 1.00. The Win Index is 1.22 (2 - LI = WI) Clemens' ERA this past year was 22 percentage points better than the league average; Clemens "should" win 14 of his 23 decisions:

There has been some discussion on the Belfry message boards over the past few months about the possibility that run support itself can be a function of who the pitcher on the mound is. The theory, loosely stated, is that some pitchers routinely get better than expected run support than others...because they're better liked by their teammates. I am, to put it mildly, extremely skeptical of this claim, which arose in part from John Eisenberg's From 33rd Street to Camden Yards. In that book, a number of ex-Orioles discuss the fact that, when he was with the club, Milt Pappas claimed that he suffered from lousy run support; some of his teammates from those years imply that this was because players on the team thought that Pappas was something of a malingerer. Without dealing with the substance of the charge itself, it goes without saying that if Pappas suffered from a chronic run support problem, it would show up by his recording an actual win-loss record that was substantially worse than his neutral win-loss record.

Pappas pitched for the Orioles from 1957 to 1965, inclusive. During those years, he compiled a record of 110-74--his ERA was 3.24 over those years while the league fashioned a 3.76 composite mark. Pappas' neutral W-L mark over that time was 102-82. In other words, Pappas won eight more games during his time with the Orioles than one would expect, given the number of decisions he had, which stands the run support theory on its head. Now, it's possible that Pappas suffered from a larger than expected number of non-decisions over that time, but since Pappas earned decisions in roughly 80% of his starts during those years--approximately equivalent to, if not higher than, the league and team average during that time frame, it's difficult to see any evidence in support of the claim that Pappas received less run support overall than expected.

One of the interesting things about the whole run support thing is that it really does appear to play out according to Hoyle...that is to say, it does seem to even out, as evidenced by neutral win-loss records, over a lengthy career. There certainly are significant seasonal blips here and there--witness the Clemens example above--but over the course of many years, things tend to even out.

Take Clemens as an example again. He was six wins better than expected in 2001. But for his career, Clemens is 280-145. His neutral record? 282-143. He was eight games under for his career until last year; now he's two under. That's pretty close.

Remember 1987? Sure you do. One thing that happened that year was Nolan Ryan's winning the National League ERA title as a member of the Houston Astros. Nothing all that remarkable about that...but he posted a W-L record that year of 8-16. His neutral W-L that season was 15-9, nearly the reverse of his actual mark and a substantial single-season difference of seven wins. Ryan had a reputation as a guy who pitched better than his W-L record all through his career, but is it so? His real life record was 324-292. His neutral mark was 336-280. So, for his career, Ryan was down 12 wins. That's a fair number...but keep in mind that seven of them came in one season. For the rest of his career, he was only down by five wins--not so large a number.

What about our buddy Milt Pappas, who finished his Orioles career eight wins down? Pappas pitched eight seasons after he was traded away from Baltimore and posted a career mark of 209-164. His neutral mark was 206-167, so he finished +3 wins over his lengthy career and actually made up 11 wins after moving over to the NL.

What follows is a list of the top 50 pitchers in terms of career wins, and how their neutral win totals compare.

RANK	Player	W	NW	DIFF	% of W
1	Cy Young	511	533	-22	-4.3%
2	Walter Johnson	417	470	-53	-12.7%
T3	Grover Alexander	373	374	-1	-0.3%
T3	Christy Mathewson	373	361	12	3.2%
5	Warren Spahn	363	353	10	2.8%
6	Kid Nichols	361	373	-12	-3.3%
7	Pud Galvin	360	359	1	0.3%
8	Tim Keefe	341	346	-5	-1.5%
9	Steve Carlton	329	327	2	0.6%
10	John Clarkson	328	323	5	1.5%
11	Eddie Plank	326	311	15	4.6%
T12	Don Sutton	324	310	14	4.3%
T12	Nolan Ryan	324	336	-12	-3.7%
14	Phil Niekro	318	337	-19	-6.0%
15	Gaylord Perry	314	336	-22	-7.0%
16	Tom Seaver	311	312	-1	-0.3%
T17	Old Hoss Radbourn	309	300	9	2.9%
T17	Mickey Welch	309	291	18	5.8%
T19	Lefty Grove	300	298	2	0.7%
T19	Early Wynn	300	288	12	4.0%
21	Tommy John	288	284	4	1.4%
22	Bert Blyleven	287	313	-26	-9.1%
23	Robin Roberts	286	296	-10	-3.5%
T24	Tony Mullane	284	289	-5	-1.8%
T24	Ferguson Jenkins	284	291	-7	-2.5%
26	Jim Kaat	283	282	1	0.4%
27	Roger Clemens	280	282	-2	-0.7%
28	Red Ruffing	273	273	0	0.0%
29	Burleigh Grimes	270	259	11	4.1%
30	Jim Palmer	268	251	17	6.3%
T31	Bob Feller	266	252	14	5.3%
T31	Eppa Rixey	266	291	-25	-9.4%
33	Jim McCormick	265	275	-10	-3.8%
34	Gus Weyhing	264	262	2	0.8%
35	Ted Lyons	260	284	-24	-9.2%
36	Greg Maddux	257	268	-11	-4.3%
T37	Red Faber	254	269	-15	-5.9%
T37	Jack Morris	254	232	22	8.7%
39	Carl Hubbell	253	253	0	0.0%
40	Bob Gibson	251	264	-13	-5.2%
41	Vic Willis	249	258	-9	-3.6%
42	Jack Quinn	247	265	-18	-7.3%
43	Joe McGinnity	246	230	16	6.5%
T44	Amos Rusie	245	257	-12	-4.9%
T44	Jack Powell	245	265	-20	-8.2%
46	Juan Marichal	243	228	15	6.2%
47	Dennis Martinez	241	229	12	5.0%
T48	Herb Pennock	240	212	28	11.7%
T48	Frank Tanana	240	252	-12	-5.0%
50	Three Finger Brown	239	244	-5	-2.1%

Note how clustered these pitchers are around their real-life totals in a more or less normal manner. The biggest percentage differences--indeed the only two of 50 on the list to reach double digits--are Walter Johnson, who comes in a staggering 53 wins (12.7%) below his neutral total, and Herb Pennock who finishes 28 wins (11.7%) on the upside. These are the outlying cases--something one expects in virtually every sample, even non-random ones like that above.

The implication is that while there is a fair amount of season-to-season fluctuation in wins/losses from expected totals, it appears that the career won-loss marks are usually a reasonably accurate reflection of a pitcher's ability. Perhaps I should rethink that rant at the beginning of this article. Nah. <g>

RANK	Player	W	NW	DIFF	% of W
1	Jim Palmer	268	251	17	6.3%
2	Dave McNally	181	158	23	12.7%
3	Mike Mussina	147	141	6	4.1%
4	Mike Cuellar	143	125	18	12.6%
5	Mike Flanagan	141	130	11	7.8%
6	Scott McGregor	138	121	17	12.3%
7	Milt Pappas	110	102	8	7.3%
8	Dennis Martinez	108	94	14	13.0%
9	Steve Barber	95	95	0	0.0%
10	Mike Boddicker	79	84	-5	-6.3%
11	Scott Erickson	74	67	7	9.5%
12	Dick Hall	65	59	6	9.2%
13	Hal Brown	62	58	4	6.5%
14	Storm Davis	61	57	4	6.6%
15	Ben McDonald	58	62	-4	-6.9%
16	Tippy Martinez	52	52	0	0.0%
T17	Ross Grimsley	51	43	8	15.7%
T17	Sammy Stewart	51	53	-2	-3.9%
19	Tom Phoebus	50	48	2	4.0%
20	Chuck Estrada	48	45	3	6.3%
21	Mark Williamson	46	42	4	8.7%
22	Wally Bunker	44	36	8	18.2%
T23	Hoyt Wilhelm	43	57	-14	-32.6%
T23	Arthur Rhodes	43	41	2	4.7%
25	Robin Roberts	42	44	-2	-4.8%
26	Steve Stone	40	32	8	20.0%
T27	Stu Miller	38	50	-12	-31.6%
T27	Jim Hardin	38	37	1	2.6%
T29	Bob Milacki	37	35	2	5.4%
T29	Eddie Watt	37	41	-4	-10.8%
T31	Pat Dobson	36	35	1	2.8%
T31	Jeff Ballard	36	37	-1	-2.8%
33	Doyle Alexander	35	36	-1	-2.9%
34	Sidney Ponson	34	36	-2	-5.9%
35	Ray Moore	33	29	4	12.1%
36	Alan Mills	32	29	3	9.4%
T37	Jack Fisher	30	34	-4	-13.3%
T37	Pete Richert	30	32	-2	-6.7%
T39	Connie Johnson	29	31	-2	-6.9%
T39	Billy O'Dell	29	37	-8	-27.6%
T39	Rudy May	29	24	5	17.2%
42	Dave Schmidt	28	24	4	14.3%
43	Wayne Garland	27	25	2	7.4%
T44	Rick Sutcliffe	26	22	4	15.4%
T44	Ken Dixon	26	24	2	7.7%
46	Jamie Moyer	25	25	0	0.0%
47	Grant Jackson	24	21	3	12.5%
48	George Zuverink	23	23	0	0.0%
T49	Jimmy Key	22	21	1	4.5%
T49	Jose Mercedes	22	21	1	4.5%

Just one note on the above table...take note of the fact that this is not a normally distributed sample. Only 14 of the players on the list are on the minus side, and the top eight players are all on the plus side, four of them by double-digit percentages. This is at least partly function of how good the Orioles were--offensively--during the careers of most of the above pitchers. The run support that these pitchers received, relative to the league, most of the time gave them a larger than average margin for error. One of the great myths of Orioles history is how the team based its success on pitching and defense, with relatively little attention paid to offense. That's simply not true--the Orioles were routinely among the upper half of the American League in runs scored from the mid-1960s into the 1980s.