Since 1900 there have been only three seasons by a pitcher in which the pitcher had 25 wins, 300 strikeouts, an ERA under 2.50 and a winning percentage of .750.   Those three seasons were by Sandy Koufax, 1963, Sandy Koufax, 1965, and Sandy Koufax, 1966.    In a recent article I referred to Hal Newhouser in 1946 as having a Sandy Koufax season, meaning that he had almost Koufax-like numbers: 26-9 with 275 strikeouts and a 1.94 ERA.   It’s a hell of a year, although actually he misses two of Koufax’ standards, his winning percentage being less than .750 and his strikeouts less than 300.   Still . ..a good season.

                Better than a good season; better than a Cy Young season—not that Sandy Koufax was better than Cy Young, but a Sandy Koufax season is above the standard that is required to win a Cy Young Award.   You can win a Cy Young Award with 20 wins, a .700 winning percentage, 200 strikeouts and an ERA of 3.00.   A Koufax season is a step up from that.

                I suppose I should warn you all, before you waste your time reading this article, that this is not serious research.   Most of what I do is chasing a question with data.    There’s no clear question here; I’m just farting around with the data.   I’m doing this because I like doing this.   It’s not real research.

                So anyway, I’ve always been interested in these occasional seasons that pitchers have, two or three times each decade, that are above the standard of your usual Cy Young season.   There were three of them in the 1970s—Vida Blue, 1971, Steve Carlton, 1972, and Ron Guidry, 1978.    Vida was 24-8, 301 strikeouts, 1.82 ERA; one more win and he’d be with Sandy on that list in the first paragraph.    Carlton was 27-10, 310 strikeouts, 1.97 ERA; one less loss and he’d be with Koufax on the list.    Guidry was 25-3, 1.74 ERA, although he struck out only 248.   There was only one comparable season in the 1980s:  Doc Gooden in 1985. 

                Suppose that we start with a lower set of standards, and work toward the Sandy Koufax standards in steps.   Some of this is fairly complicated; it started out simple, but I’ve been playing around with this for several days, and things get complicated.    One thing doesn’t quite work, so you back off and try something else.    I started with a simple idea, but we won’t exactly go back to the start because that would create an unnecessarily long and winding trail, and might even involve going to Cleveland.   

                I’m trying to keep it as simple as I can.   We start with these standards:

                15 or more wins,

                a .500 winning percentage,

                100 strikeouts,

                an ERA no higher than 4.50, and

                a league average strikeout to walk ratio.     

                This is what I call a Bronson Arroyo season, although the pitcher who best draws the line here is Jake Westbrook in 2005.    Westbrook went 15-15, a .500 winning percentage.    He struck out 119 batters, had a 4.49 ERA and a strikeout to walk ratio just a tiny bit better than the league average.   He just barely reaches the standard in all five areas. 

                There are 7,995 seasons in baseball history that meet the Bronson Arroyo set of standards.    That’s not exactly true, or rather, it requires more explanation.    First of all, by "baseball history", I do not mean the 19^th century, since 19^th century baseball was not, in reality, major league baseball.   Second, a couple of the stats are adjusted; I’ll explain that later.   And third, the pitcher doesn’t actually have to meet all five standards.  

                One of the first things I learned, messing around with the data, is that a rigid, no-tolerance "standards" approach leads to poor results.    Jim Palmer in 1973 went 22-9, 2.40 ERA, and won the Cy Young Award, but his strikeout/walk ratio was just a tiny bit below the league norm.   Warren Spahn in 1963 went 23-7, 2.60 ERA, but his strikeout/walk ratio (102 to 49) was just a tiny, tiny bit worse than the league norm.   Bob Welch in 1990 was 27-6, 2.95 ERA, won the Cy Young Award, but his strikeout/walk ratio is a tiny bit worse than the league norm.    When their strikeout/walk ratio is stated as a winning percentage, Spahn’s is .498 and Welch’s is .499—below average, but not really.

                Maybe Palmer didn’t deserve the Cy Young Award; probably Welch didn’t deserve the Cy Young Award—but they were certainly better than Jake Westbrook is 2005, 15-15 with a 4.49 ERA.    They were better than your typical Bronson Arroyo season. It isn’t logical to include a pitcher who just skims by all five of the standards, but exclude a pitcher whose overall performance is far better, but who misses on one or two standards.

                I’m not trying to be strictly logical here; it’s more of an intuitive process than organized, formal research.    But it is too illogical, too far off the reservation, to exclude these outstanding seasons by Jim Palmer, Warren Spahn and Bob Welch, while including Mike Smithson in 1985 (15-14, 4.34 ERA.   Smithson’s strikeout to walk ratio was 127-78; Welch’s was 127 to 77.)    

                I modified the Five Standards approach by adding this rule:   that a player who misses one or more standards may remain within the group, but he must "trade off" for missed standards on a two-for-one basis.   In other words, if a player misses one standard, he must meet a higher standard in some other area—not one higher standard, but two.

                A small rule, but in practice it profoundly changes the process of identifying the pitchers we want.    In the end, this rule will enable us to include Ron Guidry (1978) in the Sandy Koufax group, even though Guidry fell well short of 300 strikeouts.   For now, though, it makes the process a lot harder, and a lot harder to explain.

                I’ll explain more later, but let’s move forward to the next step.   The first step is 15 wins, a .500 winning percentage, 100 strikeouts, an ERA no higher than 4.50, and a league-average strikeout to walk ratio.   The second step is 16 wins, a .525 winning percentage, 120 strikeouts, an ERA no higher than 4.25, and a strike zone winning percentage of .520.

                Huh?

                I’ve explained this before, but I know you don’t all memorize everything I write.   A pitcher’s strikeout to walk ratio can be stated as a winning percentage by the following method:

Strikeouts Times League Walks + Walks Times League Strikeouts

                In 2013 Cliff Lee had a strikeout to walk ratio of 222 to 32, while the league totals were 18174 strikeouts, 7219 walks.   That makes a strike zone winning percentage of .734:

222 * 7219  = 1 602 618

32 * 18174  =    581 568

1 602 618 / (1 602 618 + 581 568) = .734

                This was the best strike zone winning percentage in the majors among pitchers with 100 or more inning pitched.    Jake Westbrook, on the other hand, had a strikeout to walk ratio of 44 to 50, which makes a strike zone winning percentage of .259, which was the worst in the majors, 100 or more inning pitched.    In this way, strikeout-to-walk ratios can be stated on a league-normalized scale, consistent with winning percentages.    And I promise not to pick on Jake Westbrook any more.

                The strikeout to walk ratio is actually the only one of the five standards which is fully adjusted for the league norms, except that the winning percentage, of course, is on the same scale every year.   Anyway, our "Second Step" standards are 16 wins, a .525 winning percentage, 120 strikeouts, a 4.25 ERA, and a strike zone winning percentage of .520.     I call these the Milt Pappas standards.   Too old?     How about the Dan Haren standards?

                OK, now that you have the idea, these are the ten sets of standards that take us from Bronson Arroyo to Sandy Koufax:

                Even the lowest standard here is pretty good.    Bronson Arroyo is a good pitcher.    He’s been paid tens of millions of dollars, and he has earned it.   We didn’t start at the bottom of the barrel here.   Only about 20% of pitchers even meet the Bronson Arroyo standard.

                But with each step up the ladder, the number of pitchers who qualify is essentially divided in half:

                It forms a stairway, leading to Sandy Koufax:

                In the Bronson Arroyo group, the number of seasons by rookie pitchers outnumbers the number of seasons by pitchers who are in the Hall of Fame, 871 to 734.     But with just one step up the ladder, up to the Dan Haren group, there are far more seasons in the study by Hall of Fame pitchers than by rookies, 602 to 375:

#	Cognomen	QUALIFIERS	Rookies	Hall of Famers
Step 2	The Dan Haren Standards	4462	375	602
Step 1	The Bronson Arroyo Standards	7995	871	734

 

                By the fourth step on the staircase, seasons by Hall of Famers outnumber rookies almost five to one; by the sixth step, more than ten to one.   By the eighth step on the ladder (Tom Seaver), most of the seasons in the data are by pitchers now in the Hall of Fame, and there are no seasons at all by rookie pitchers:

 

#	Cognomen	Rookies	Pct	HOF	Pct
Step 10	The Sandy Koufax Standards	0	0%	12	57%
Step 9	The Randy Johnson Standards	0	0%	22	52%
Step 8	The Tom Seaver Standards	0	0%	50	57%
Step 7	The Roger Clemens Standards	1	1%	87	50%
Step 6	The Roy Halladay Standards	11	3%	143	41%
Step 5	The Jim Bunning Standards	29	4%	237	33%
Step 4	The Kevin Brown Standards	73	5%	334	25%
Step 3	The Kevin Appier Standards	179	7%	465	18%
Step 2	The Dan Haren Standards	375	8%	602	13%
Step 1	The Bronson Arroyo Standards	871	11%	734	9%
#	Cognomen	Rookies	Pct	HOF	Pct

 

                Does it need to be said that, of course, the true Hall of Fame percentages are higher than this, since pitchers like Randy Johnson and Pedro Martinez have not yet seen the glories of Cooperstown?   The one rookie pitcher who survived into the Roger Clemens group was Russ Ford, 1910; he was 26-6 with a 1.65 ERA, 209 strikeouts.

                Speaking of that season, there is something else I needed to explain.   Russ Ford won 26 games, but I didn’t treat him like a 26-game winner for purposes of this exercise.    In scoring him, I treated him as if he had won 19 games, rather than 26—an .813 winning percentage, as he actually had, but only 19 wins. 

                Why?

                It was necessary to do this to prevent the list from being overrun with pitchers from the Dead Ball era.    It was a very different game; standards were different.    Two pitchers in that era won 40 games in a season, and many pitchers in that era won 30 and 30+ games.   Cy Young won 30+ games in 1901 and 1902; Pete Alexander won 30+ in 1915, 1916 and 1917.    Christy Mathewson won 30 games three times.    These numbers are not comparable to modern baseball.   To decide into which bracket to put a pitcher from the 1900 to 1919 era, I multiplied his wins times .7, and then rounded up; 41 wins becomes 29, 35 wins becomes 25, 30 wins becomes 21, 25 wins becomes 18.

                The pitchers from that era still do very, very well in this competition; they just don’t dominate as much as they would if we didn’t adjust these things to scale.   Also, I partially adjusted Earned Run Averages for the league ERAs.      What I did is, I normalized each pitcher’s ERA to a league norm of 3.92, since 3.92 was the average league ERA for the pitchers in my data, then I moved the pitcher’s actual ERA to a point half-way between his "raw" ERA and his season-adjusted ERA.    For illustration, Chris Short in 1965 and David Cone in 1997 both had ERAs of 2.82.   In the National League in 1965, however (Short’s league), the league ERA was 3.54; in the American League in 1997 the league ERA was 4.57.  

                If you normalize Cone’s ERA to a 3.92 league ERA, it drops from 2.82 to 2.42; if you normalize Short’s ERA to a 3.92 league norm it goes up from 2.82 to 3.12.   So what I did is, I moved each pitcher half-way to his normalized ERA, creating a "compromise" ERA between his raw data and his normalized data.   Cone’s ERA goes from 2.82 to 2.62; Short goes from 2.82 to 2.97.

                It’s an accommodation, rather than a full adjustment.    This isn’t a literal, serious effort to rank pitchers; it’s just messing around, having fun with the data.   The accommodation doesn’t make much difference in most cases.   Bob Gibson’s 1.12 ERA in 1968 goes to 1.30; Three Finger Brown’s 1.04 in 1909 also goes to 1.30.    Mike Mussina’s 4.81 ERA in 1996 adjusts to 4.29.    About 50% of pitchers wind up with ERAs within 10 points of their raw ERA.

                Without these accommodations, there would be long stretches of baseball history where no pitchers at all reach the standard of a Roy Halladay season or better, while other eras would be saturated with them.   With these adjustments, we have a reasonable balance across time:

 

                The systems are designed to identify rotation starters, and then to identify increasingly better starting pitchers.  Sometimes, however, two other classes of pitchers sneak into the lower groups:  pitchers with a limited amount of work, and relievers.    In the Bronson Arroyo group, more than 15% of the pitchers who qualify are actually relievers with good enough numbers in the other categories to overpower their lack of wins.    The last four relievers who are able to stay in the group are Phil Regan, 1966, Bruce Sutter, 1977, Mark Eichorn, 1986, and Mariano Rivera, 1996.   All four of them pitched 107 or more innings, had winning percentages over .700 and ERAs under 2.00, and three of whom also struck out well over 100 batters.  These four relievers are able to meet the Roy Halladay Standards.

To limit the number of relievers and partial-season guys who worm their way into what are supposed to be lists of starting pitchers, I put in a couple of special rules.

1)   The strike zone winning percentage doesn’t count unless the pitcher has 100 strikeouts,

2)  The number of credits given (standards met) for winning percentage cannot exceed the pitcher’s number of decisions.  

The percentages still count as category qualifications; they just don’t count for "offsets" if the pitcher fails to meet standards. ..don’t know if that makes sense.   If a pitcher goes 1-0, that’s a 1.000 winning percentage, but he can’t count that as an offset against a failure to meet one of the other standards.    I don’t even know why I’m explaining this; I can’t imagine that any of you actually care.  

OK, let’s head toward the finish line.    In the Bronson Arroyo group, there are 5,332 pitcher/seasons since 1956.   Of those 5,332, 104 won the Cy Young Award, or about 2%.    As the standards tighten and pitchers drop out, however, the Cy Young Award winners tend to stay in:

		Count
		of	Cy
		Pitchers	Young
		Since	Award
#	Cognomen	1956	Winners	Percentage
Step 10	The Sandy Koufax Standards	13	12	92%
Step 9	The Randy Johnson Standards	27	21	78%
Step 8	The Tom Seaver Standards	50	35	70%
Step 7	The Roger Clemens Standards	99	55	56%
Step 6	The Roy Halladay Standards	212	72	34%
Step 5	The Jim Bunning Standards	435	86	20%
Step 4	The Kevin Brown Standards	824	96	12%
Step 3	The Kevin Appier Standards	1593	101	6%
Step 2	The Dan Haren Standards	2924	102	3%
Step 1	The Bronson Arroyo Standards	5332	104	2%

 

Up to Step 6, the Roy Halladay standards, the standards of the group are clearly lower than the standards of a Cy Young season.   Most pitchers who meet the standards do not win the Cy Young Award; a third of them do, but two-thirds do not.     On Step 7, however, the standards move past the standards of the Cy Young Award.    A season that meets the Roger Clemens standards normally wins the Cy Young Award; one can argue the issue at Step 7, but not at Step 8.   At Step 8, the Tom Seaver Standards, the pitchers in the group are clearly much better than the list of  Cy Young winners.    35 of the 50 are the same.   The 15 pitchers in this group who didn’t win the Cy Young Award include, for example, Juan Marichal in 1963 (25-8, 248 strikeouts, 2.41 ERA), Juan Marichal in 1966 (25-6, 2.23 ERA, 222-36 strikeout/walk ratio), Ferguson Jenkins in 1974 (25-12, 2.82 ERA, 225 strikeouts), Roger Clemens in 1990 (21-6, 209 strikeouts, 1.93 ERA), and Curt Schilling in 2002 (23-7, 316-33 strikeout to walk ratio).   The Cy Young Award winners who don’t meet the Roger Clemens standard, on the other hand, are like Pete Vuckovich, 1982 and Pat Hentgen in 1996—guys who had good years in leagues in which nobody was having a really good year.    The only pitcher in the Cy Young era who has met the Sandy Koufax standards, but who didn’t win the award, was Curt Schilling in 2002.

So that was what I was trying to do; I was trying to create a standard of a pitcher’s season that is ABOVE the standard of a Cy Young Award.   These are the pitchers, since 1900, who meet the Sandy Koufax standards:

 

First	Last	Year	W	L	WPct	SO	BB	ERA
Christy	Mathewson	1908	37	11	.771	259	42	1.43
Ed	Walsh	1908	40	15	.727	269	56	1.42
Walter	Johnson	1912	33	12	.733	303	76	1.39
Joe	Wood	1912	34	5	.872	258	82	1.91
Walter	Johnson	1913	36	7	.837	243	38	1.14
Dazzy	Vance	1924	28	6	.824	262	77	2.16
Lefty	Grove	1930	28	5	.848	209	60	2.54
Lefty	Grove	1931	31	4	.886	175	62	2.06
Sandy	Koufax	1963	25	5	.833	306	58	1.88
Sandy	Koufax	1965	26	8	.765	382	71	2.04
Sandy	Koufax	1966	27	9	.750	317	77	1.73
Denny	McLain	1968	31	6	.838	280	63	1.96
Vida	Blue	1971	24	8	.750	301	88	1.82
Steve	Carlton	1972	27	10	.730	310	87	1.97
Ron	Guidry	1978	25	3	.893	248	72	1.74
Dwight	Gooden	1985	24	4	.857	268	69	1.53
Greg	Maddux	1995	19	2	.905	181	23	1.63
Pedro	Martinez	1999	23	4	.852	313	37	2.07
Pedro	Martinez	2000	18	6	.750	284	32	1.74
Randy	Johnson	2002	24	5	.828	334	71	2.32
Curt	Schilling	2002	23	7	.767	316	33	3.23

 

That’s essentially the list I wanted; those are the "special seasons" I was hoping to single out.   There are a handful of seasons that I guess could be on the list:  Newhouser in 1946, one Bob Feller season somewhere, Gibson in ’68, Verlander in 2011.    But there are always going to be close calls; the highest standard has to exclude very, very good pitchers.

These are the pitchers who meet the Randy Johnson standard, but don’t meet the Sandy Koufax standards:

 

First	Last	Year	W	L	WPct	SO	BB	ERA
Cy	Young	1901	33	10	.767	158	37	1.62
Christy	Mathewson	1905	31	9	.775	206	64	1.28
Grover Cleveland	Alexander	1915	31	10	.756	241	64	1.22
Dizzy	Dean	1934	30	7	.811	195	75	2.65
Bob	Feller	1940	27	11	.711	261	118	2.62
Hal	Newhouser	1946	26	9	.743	275	98	1.94
Robin	Roberts	1952	28	7	.800	148	45	2.59
Juan	Marichal	1966	25	6	.806	222	36	2.23
Bob	Gibson	1968	22	9	.710	268	62	1.12
Roger	Clemens	1986	24	4	.857	238	67	2.48
Bret	Saberhagen	1989	23	6	.793	193	43	2.16
Randy	Johnson	1995	18	2	.900	294	65	2.48
John	Smoltz	1996	24	8	.750	276	55	2.94
Roger	Clemens	1997	21	7	.750	292	68	2.05
Greg	Maddux	1997	19	4	.826	177	20	2.20
Randy	Johnson	1997	20	4	.833	291	77	2.28
Randy	Johnson	2000	19	7	.731	347	76	2.64
Curt	Schilling	2001	22	6	.786	293	39	2.98
Randy	Johnson	2001	21	6	.778	372	71	2.49
Pedro	Martinez	2002	20	4	.833	239	40	2.26
Justin	Verlander	2011	24	5	.828	250	57	2.40

 

One more thing:  One can go higher than the Sandy Koufax standard.     Extend the process for one more step, and you get a list of seven pitchers: 

First	Last	Year	W	L	WPct	SO	BB	ERA
Christy	Mathewson	1908	37	11	.771	259	42	1.43
Walter	Johnson	1913	36	7	.837	243	38	1.14
Dazzy	Vance	1924	28	6	.824	262	77	2.16
Lefty	Grove	1931	31	4	.886	175	62	2.06
Sandy	Koufax	1965	26	8	.765	382	71	2.04
Denny	McLain	1968	31	6	.838	280	63	1.96
Pedro	Martinez	1999	23	4	.852	313	37	2.07

 

Extend it one more step after that, and you get a list of one:  Walter Johnson in 1913.   Because Walter pitched before 1920, we treat his 36 wins as if they were 26.   Because the league ERA in 1913 was 2.92, we treat Walter as if his ERA was 1.40, rather than 1.14.   And the season still reaches the highest rung on the ladder.  

But that’s not really what I was trying to do; I don’t care what the one greatest or the seven greatest starting pitcher seasons are.    I was trying to develop a protocol to make a list of the seasons worthy of the Sandy Koufax label, and I’m happy with my list.