In an interesting move, Roger Clemens has fired back at Brian McNamee’s steroid allegations in the Mitchell Report. Using a report of his own, his representatives examine his late-career performance as a pitcher. After reading through the report, I became inspired to do my own investigation. I present my analysis below. I have received no financial compensation, but I will disclose that I do own a Clemens rookie card. 🙂
Clemens’s study looks at his performance in reference to the changing playing environment of the league. For many potential reasons (e.g., new ballparks, expansion, juiced balls, etc.) the run environment dramatically changed over the course of Clemens’s career. To combat any bias from this, the authors of the study compare Clemens’s ERA to the average of the league, looking at the difference. It is no surprise that his ERA difference fluctuated quite a bit over the course of his career (p. 6). ERA is inherently a volatile statistic. In fact, in my book I use Clemens as an example of how drastically a pitcher’s ERA can change during a career, and I show that his variation in performance was well within the normal range (p. 165).
What I would like to do is to take this analysis one step further. One of the reasons that ERA is volatile is that it includes the performance of fielders and random bounces of the ball. If we use the novel idea of Voros McCracken’s DIPS, then we can more closely examine the components of Clemens’s performance over his career using factors that were solely in his control. The major defense-independent components of pitcher performance are strikeouts, walks, and home runs. In particular, I would expect Clemens’s strikeout numbers to influence and reflect his decision to use performance-enhancing drugs. Because these statistics occur at different rates, I compare how Clemens performed relative to the league average as percent of the league average.
First, I look at ERA, as percent above league average, using a connected scatter plot and a curve that represents the best quadratic fit of the data.
Though Clemens remains well above average during his career, there is a rise and decline in his performance. Two peaks jump right out: 1997–1998 (ages 34–35) and 2004–2006 (ages 41–43). McNamee claims that Clemens began using steroids in mid-season 1998, and during parts of the 2000 and 2001 seasons. I think the 2004–2006 period is more indicative of steroid use: a period during which McNamee was working with Clemens but claims not to have given him steroids. This is also a period when MLB was testing for steroids.
In terms of his strikeout rate, Clemens shows a trend of slow rise and decline during his career.
Again, there is a spike in 1996—1998; however, his performance drops considerably in 1999, and only in 2002 does he strike out batters at a similar rate again. Clemens shows clear signs of aging in his ability to strikeout batters.
Clemens’s age profile for walks is strange, and it is the only area in which he improved dramatically as he grew older.
During the first half of his career his walk prevention decined—with the exception of 1997, when he was outstanding in every aspect of the game on his way to winning the Cy Young—but after bottoming out in 1999, he showed slow improvement. Could this be the sign of an aging veteran who is using knowledge and finesse to make-up for the physical effects of aging? Walks are something that tend peak later for all pitchers. This could be due to pitchers’ choosing to compensate for physical deterioration or just a product of learning. I find it interesting that as Clemens’s strikeout power waned he improved in this area—the area in which I expect steroids would be least helpful.
Clemens’s home run prevention fluctuates wildly throughout his career, with a barely discernible rise-and-fall aging trend .
As we saw with ERA, there are two spikes in 1997–1998 and in 2004–2006. I’m not sure there is much to glean from this area of performance.
In summary of the data, there is no doubt that Clemens is a good pitcher who had some excellent years late in his career. Only the spike in 1998 fits with McNamee’s testimony; and, this doesn’t seem like a time that Clemens would consider using drugs. Also, the more remarkable late-career spike occurs when Clemens was in his 40s—after McNamee was not giving steroids and MLB had implemented random drug testing. Next, I want to consider the possible motives at the time McNamee claims Clemens used the drugs.
In the Mitchell Report, McNamee claims that Clemens first expressed interest in using steroids on, or just after, a road trip to Florida, during which Clemens pitched on June 8. Here are Clemens’s performance statistics from his first start until when he allegedly indicated to McNamee that he wanted to start using steroids. As a reference point I include how much better or worse than he was than league average at the time. While these are excellent numbers, Clemens may have demanded more of himself, so I include his performance from 1997 (overall and relative to the league) as another benchmark.
ERA K9 BB9 HR9 1998 Clemens 3.27 9.18 4.32 0.32 4/1--6/8 Lg Avg. 4.66 6.39 3.43 1.10 % > Avg 29.77% 43.57% -25.95% 71.36% 1997 Clemens 2.05 9.95 2.32 0.31 Lg Avg. 4.57 6.43 3.50 1.10 % > Avg 55.26% 54.71% 33.80% 72.10%
His ERA was higher than it was in 1997, and he was walking more batters; but, his strikeouts and home run prevention were similar. It looks like Clemens’s problems had more to do with control than power. After June 8, he cut his walk rate to 2.83 per nine innings and increased his strikeout rate to 11.09 per nine innings. I don’t see Clemens—a 15-year veteran at that point—looking at a .500 record and a slightly higher walk rate and thinking, “I need some steroids to get back on track.” It is possible that the walks were a function of fatigue and possibly he wanted steroids to help him recover quicker and go deeper into games. This is what McNamee claims, but I think we would observe performance declines in strikeouts and homers in the early part of the season if this was the case. Clemens appeared to get his command back, which lowered his walks and increased his strikeouts. Steroids could have also helped, but the issue is explaining why he would choose to start using when he seems to be performing up to recent standards.
Clemens’s second supposed use of steroids occurred during the 2000 season as a member the New York Yankees. According to the Mitchell Report, McNamee alleged that “during the middle of the 2000 season Clemens made it clear that he was ready to use steroids again.” He claims to have administered steroids and human growth hormone, each four to six times, during the second half of the season. Though we do not have the exact timeline, looking at the first-half performance should be sufficient to see how Clemens was performing up to that point.
Half ERA K9 BB9 HR9 First 4.33 8.47 3.76 1.22 Second 3.15 8.12 3.64 1.08
Clemens’s first-half ERA was better than the league average and his performance in 1999. His strikeout and walk rates (7.81 and 4.31) were also better than the previous season. It was his worst year at preventing home runs, but I doubt this one thing would be enough to motivate Clemens to use. Again, this seems like an odd time for Clemens to begin using performance-enhancers. And even if he did take performance-enhancing drugs, they did not appear to have much of an effect.
The third instance at which McNamee accuses Clemens of steroid use occurs in August 2001. This would have been an odd time to ask for help considering that he was performing better than the previous season in all areas, and he would ultimately win the Cy Young award.
ERA BB9 K9 HR9 2001 (Apr-July) 3.58 2.79 8.92 0.91
That is the last time McNamee would document Clemens’s steroid use, even though he would continue to train him; thus, McNamee claims that the second late-career bump was clean. In 2002 and 2004, Clemens would post strikeout rates of over one per inning—something he had not achieved since 1998. His walks steadily improved, and 2004–2006 were excellent years for his ERA and home run prevention. He would win the Cy Young in 2004, and he came in third in 2005. Some of this may be due to his travel arrangements and short seasons, but he would pitch over 200 innings in 2004 and 2005. This is also a period of drug testing; thus, all signs point to Clemens’s performance being natural.
Another tool is used in the Clemens report is a comparison to other great pitchers: Nolan Ryan, Randy Johnson, and Curt Schilling. Again, this doesn’t prove that he’s clean, but it does show that such continued excellent performance is common enough to be expected by natural means. I was surprised to see one comparison not included, John Smoltz. Since returning to the Atlanta rotation in 2005, he has been one of the most dominant pitchers in the game, ranking among the league’s ten most valuable pitchers from 2005–2007 (ages 38–40). This is another example of an exceptionally talented ballplayer who loves playing the game and has a strong work-ethic.
There is also the possibility that Clemens used steroids throughout his career, and that other pitchers having late-career success have all been aided by performance-enhancing drugs. I find both to be unlikely, especially the former. If Clemens continued employing McNamee several years beyond the last time the alleged steroid injections occurred, why would he go somewhere else for his steroids? Or, why would McNamee lie to say that Clemens was clean during this period?
This in no way proves that Clemens didn’t use steroids; he is right when says he can’t prove a negative. Others may interpret the data differently than I do. But, the more I look into these numbers, the more I am convinced that McNamee is the one who is lying. Until I see some other corroborating evidence—I wonder what is in those seized 2004 drug tests—I have to conclude that Clemens pitched without the aid of steroids.