Clemens 1

            The general points that I want to make here are very limited, and can be enumerated briefly:

            1)  Until this issue stepped to the center of the stage, I had always assumed that Roger Clemens had used steroids.  I assumed that because that’s what all the people “in the know” assumed—the writers and front office executives who gossip about these things in bars and in the break room.

            But once Clemens steps up, in this very dramatic way, and asserts that he never used anything, that changes the ground rules.   He is entitled to a reasonable presumption of innocence.

            2)  I am certainly not a fanatic about the presumption of innocence.  The legal presumption of innocence is a meaningful construct only where a person stands in jeopardy of punishment.   It is a protection against the careless and unwarranted use of the powers of the state.  As long as it is just you and me talking and neither of us is in a position to punish the accused, we can say whatever we think is true.

            That’s not what I’m talking about, the legal presumption of innocence.  What I am referring to is more like the benefit of the doubt that one decent person owes another in difficult circumstances.

            If you have two friends who go through a nasty divorce and he is saying horrible things about her and she is saying horrible things about him, you treat him as if the things she was saying were untrue, and her as if the things he was saying were untrue.  You give them both the benefit of the doubt.

            Decent people do not say unkind and defamatory things about others, at this level, unless they know those things to be true.   Unless and until we have clear evidence that Clemens used something, in my mind, he didn’t. 

            3)  The “Andy Pettitte” evidence seems to me to have a very low level of significance.  If Pettitte was saying that he and Roger regularly shot up steroids together, that would be one thing.  That’s not what he is saying.  He is saying that he had two conversations with Clemens about the use of HGH, five years apart, in the second of which Clemens told him that he had misunderstood the first.

            That conversation, the one in which Clemens denied HGH use, doesn’t count against Clemens; it counts in favor of him—not very much, but it is on his side.  That leaves us with one conversation, in 1999, which Clemens later told Pettitte that he had misunderstood.   I don’t see that as weighty evidence.

            4)  At an earlier stage of this dispute, some people argued that McNamee’s testimony was more credible because McNamee was in a position where he would go to jail if he lied.  But at this point, either Clemens or McNamee, whichever is found to be lying, is very likely to go to prison. 

The difference is this:  McNamee found himself in that position because he was caught selling drugs, and “legal jeopardy” was a step forward from a certain prison term.   Clemens is in that position because he very deliberately stepped in front of the bullet and demanded an opportunity to clear his name. 

            How, in that circumstance, can anyone think that makes McNamee more credible than Clemens?

Clemens 2

            This portion of the article focuses on the statistical analysis which has been mixed in with the Clemens discussion.

            On the Colbert Report on February 14, 2008, John Feinstein asserted baldly that Clemens was guilty, and said “his ERA dropped by three runs in his forties” (quote is from memory; my apologies if the wording is inexact.) 

            But first, the difference between the highest Earned Run Average of Clemens’ career and the lowest is less than three runs, so I don’t understand how anyone can interpret that as a three-run drop in his Earned Run Average, in his forties or at any other age. 

            Second, let’s set aside for a moment the issue of age.  Clemens had a 4.35 ERA in 2002, a 1.87 ERA in 2005.  But setting aside the issue of age, a 2.48 run fluctuation in ERA over a period of three years is an entirely routine event.   ERAs go up and down all the time for many reasons.

            Tim Belcher in 1991 had an ERA of 2.62.  Three years later, aged 32, his ERA was up to 5.89.   But two years after that, it was back down to 3.92.

            Vida Blue had a 5.01 ERA in 1979.  The next year it dropped to 2.97.

            Bert  Blyleven had a 5.43 ERA in 1988, at age 37.  The next year he cut it to 2.73.

            Kevin Brown had a 4.82 ERA in 1994.  In 1997 he cut it to 1.89.

            John Burkett had a 5.68 ERA in 1998 (aged 33), and, to prove it wasn’t a fluke, a 5.62 ERA in 1999.   But in 2001 he cut it to 3.04. 

            There are only 112 pitchers since 1950 who pitched 150 innings ten times.  Belcher, Blue, Blyleven, Brown and Burkett are all on that list, and we’re not out of the B’s yet.   And I’m only using ERAs from seasons when they pitched enough innings to make the ERA meaningful.  It’s just a very normal thing for a pitcher’s ERA to bounce up and down like that.   It doesn’t indicate a profound change in the pitcher’s stuff.   It just indicates. . .one year you have a little injury, you have a little bad luck, your slider flattens out on you, your ERA goes up.   The next year you’re more healthy, you’ve got the slider back, you’ve got Adam Everett playing shortstop behind you, things go your way.  

            That was setting aside the issue of age; now embracing the issue of age.

            If you take a 35-year-old pitcher and a 25-year-old pitcher of the same age and ability, same strikeout level, there is very little difference in their expected future performance.  If you don’t study it really, really well, you can reach the conclusion that there is no difference in their expected future performance.  There is a difference, but it’s tiny.   Aging for a pitcher is not like aging for a position player, which is general and progressive.  Pitchers pitch until they get hurt and can’t pitch anymore.

            Why, then are there (almost) no 45-year-old pitchers?  Simple attrition.  If you take a thousand 25-year-old pitchers and twenty to twenty-five percent of them get hurt each year, by age 35 more than 90% of them will be gone.  By age 40 98% of them are gone. 

            That simple paradigm describes the aging of pitchers.  There are about one thousand good 25-year-old pitchers in baseball history, counting the one-year-wonders who had their good year at age 24 or 26 or 23 or 27.   A substantial percentage of them get hurt every year.  By age 40, they’re almost all gone.

            Clemens beat the attrition lottery, and stayed in the game longer than almost any other pitcher.  Whether he beat the odds by luck, hard work or drugs. . .I don’t know.  My point is that, as long as he kept pitching 100+ innings a year he was in the game, and as long as he was in the game, he was subject to the normal year-to-year fluctuations in ERA.

            Clemens’ agents in late January released a weighty analysis of Clemens’ career, attempting to undercut the notion that there was something suspicious in Clemens’ accomplishments by pointing out similar up and down movements in the careers of other pitchers.   In response to that, three professors variously associated with the Wharton School and the Freakanomics group issued a rebuttal insisting that Clemens’ career path is highly unusual, and bears little resemblance to the career progressions of the pitchers discussed by Clemens’ advocates.

            In the debris flying from this controversy there are arguments about whether it is better to measure a pitcher’s effectiveness by ERA, ERA plus, WHIP, Runs Saved Against Average, or some other measure.  In following this discussion I find numerous assertions that Clemens had his best seasons after age 40.   The advocates of the various metrics seem to be engaged in a kind of proxy fight over whether or not that is true.

            Well, it isn’t true.  Roger Clemens did not have his best seasons in his forties.  I have two methods that are germane to this issue, either of which is much more directly on point than any of the metrics being cited in the debate.   One is the Win Shares method, which is a complicated way of asking the question “How many games did this player win for his team?”   By the Win Shares method, Roger Clemens best seasons are 1997 (aged 34), 1986 (aged 23) and 1987 (aged 24).   No season in his forties ranks among Clemens’ seven best seasons.

            The other method is the Season Scores method, which is a fairly simple, seat-of-the-pants method for combining Wins, Losses, Saves, Strikeouts, Walks, Innings Pitched and ERA into one number, used to routinely indicate the best season within a group of seasons.  By the Season Scores method, Clemens’ best seasons were 1997, 1986 and 1990.  No season in his forties ranks among Clemens’ eight best seasons.

Clemens 3

            It is difficult to say for certain what Feinstein is referring to with the allegation that Clemens cut three runs off his ERA in his forties, because

a)      Clemens has had a long career with several up-and-down movements in his ERA, and

b)      There are no two years with a three-run separation in ERA. 

But we have to assume he is referring to the difference between Clemens’ 2001 and 2004 seasons, since this is the only thing that sort-of makes sense.  So, focusing yet more narrowly on the difference between those two seasons. . .  

            A pitcher’s ERA can drop for any of five reasons:

            1)  Improved control,

            2)  Increased strikeouts, leading to fewer balls in play, thus fewer hits allowed,

            3)  Fewer home runs allowed,

            4)  A better relationship between ERA and the components of ERA, or

            5)  A lower ball-in-play average.

            On point one, Clemens did have a poor-control year by his own standards in 2002 (3.15 walks per nine innings) and a better control year in 2005 (2.64 walks per nine innings).   This difference does not account for a significant percentage of the distance between the two seasons.

            On point two, Clemens’ strikeout rate, relative to the league, did not improve between 2002 and 2005; in fact, it took a substantial step backward.   This is a critical point, and I’ll come back to it in a moment and fling a string of numbers at you.

            On point three, Clemens did reduce his home runs allowed between 2002 and 2005, despite an increase in innings pitched, from 18 to 11.  

            Clemens’ 4.35 ERA in 2002 is misleading, because his ERA components were better than that.   He had a season when

a)      hitters were able to bunch their hits against him, and/or

b)      his relievers allowed the runners he left on to score.

Clemens had a 4.35 ERA in 2002, but a component ERA of 3.74—a fairly normal discrepancy, but my point is that 3.74 is a better indicator of how he really pitched than 4.35.

            But the vast majority of the improvement in Clemens’ ERA between 2002 and 2005 comes from a dramatic improvement in his ball-in-play average.   In 2002 the in-play average against Clemens was .323.   In 2005 it was .248.   Since he had about 500 balls per season in play against him, that’s a savings of roughly 35 hits.   That is by far the largest cause of the difference in his ERA.

            What follows is the key sentence in this three-part Clemens article.   Can someone explain to me how it is possible for steroid use to cause the in-play average against a pitcher to drop by 74 points?   Modern students of the baseball almost universally believe that a pitcher has very minimal control of the in-play average against him.   The in-play average can be influenced by the park.   It can be influenced by the defensive quality of the team.  It is influenced, to a huge extent, by luck.   It can be influenced, to a small extent, by the pitcher himself.   But it is impossible to understand how any meaningful portion of that improvement could be caused by Clemens use of any performance-enhancing drug. 

            What steroid/HGH use could cause—and I am certainly not saying that Clemens did not use steroids, HGH, Rogain, Ritalin, Viagra or Formula 44—but what that could reasonably cause is

a)      an increase in durability, or

b)      an increase in the strikeouts, resulting from an increase in velocity.

Neither of which occurred.   Clemens’ Batters Faced, by three-year groups starting in 1986, show a slow decline with a dropoff in his forties:

Ages 23-25      1986-1988       3217

Ages 26-28      1989-1991       3041

Ages 29-31      1992-1994       2489

Ages 32-34      1995-1997       2699

Ages 35-37      1998-2000       2661

Ages 38-40      2001-2003       2564

Ages 41-43      2004-2006       2167

As for his strikeouts, Clemens moved in 2004 to the National League, where pitchers attempt to hit, and consequently the strikeout rate is higher.  In spite of this his strikeouts decreased sharply between 2002 and 2005.   Clemens relative strikeout rate was 1.81 (81% above the league norm) in 1988.  It was over 1.50 from 1996 to 1998, and it was 1.53 in 2002.   In 2005, when he posted a 1.87 ERA at age 42, it was 1.20—the lowest of his career up to that point. 

So both of the key indicators that would appear to be plausibly related to steroid or HGH use—strikeouts and durability—declined normally for Clemens as he aged.  His “improvement” from 2002 to 2005 resulted from small, normal fluctuations in control and home runs allowed, and from a large improvement in his in-play batting average.  None of this can be reasonably connected to steroid use.  

Why, then, did Clemens last so long?

First, as noted before, Clemens won the attrition lottery that strikes down 20 to 25% of pitchers every year.   Second, Clemens strikeout rate

1)      was declining from a high point, and

2)      declined at a very slow rate until he reached 40.

Whether that rate declined slowly because of hard work or drug assistance, again, I have no way of knowing.