On pages 153 and 154 of his 1988 Baseball Abstract, Bill James talks about pitchers and run support. He compares two men of the era—Danny Jackson and Walt Terrell—and examines why one (Jackson) consistently has a better ERA whereas the other (Terrell) consistently has a better winning percentage.
Essentially James looks at their statistical records over the course of a three-year stretch and breaks down their numbers according to how many runs were scored on their behalf in a given game. For example, he finds that from 1985 to 1987, Terrell outpitched Jackson when their respective teams scored three runs, but the opposite held true when four or five runs were scored.
This may not seem all that exciting in 2008, but 20 years ago, it was pretty groundbreaking stuff. For one thing, the type of data needed to conduct such analysis wasn’t readily available. For another (and this only stands to reason given the paucity of information), very few people were analyzing baseball with anything resembling intelligence. It is thanks in large part to these earlier, “simple” (I use quotes because they seem simple only in retrospect) efforts that the rest of us have access to data that we now take for granted in measuring… well, pretty much whatever we want.
There is another way in which James’ study of Jackson and Terrell is simple: It is elegant in design and execution. It asks a question (Can anyone be expected to win games consistently with two or three runs?), presents evidence, and leaves us with a little better understanding of the game we love.
My original intent here was to find pairs of pitchers with similar ERAs over a decade but widely disparate won-loss records. For example, from 1998 to 2007, David Wells parlayed a 4.23 ERA (107 ERA+) into a 133-72 record (.649 Pct); on the other hand, Javier Vazquez‘s 4.28 ERA (105 ERA+) netted him only a 115-113 record (.504 Pct). I then planned to examine how each pitcher fared at different levels of run support over that stretch.
I decided against this because (a) going through 10 years of starts seemed like an awful lot of work and (b) I was less interested in learning about Wells and Vazquez than I was in deploying James’ strategy. Why am I telling you all this? Well, maybe you’ll want to compare Wells and Vazquez… or Aaron Sele and Steve Trachsel (another matched set). Or, more practically, maybe you’ll be interested to see how, say, Tim Lincecum and Brandon Webb stack up this year.
Whatever the case, I wanted to share the thought process—the motivation—behind this little exercise. As a certain favorite lyricist of mine once noted, “The point of the journey is not to arrive.” We must pay attention to how we reached our destination as well, especially if we intend to return.
So, if I didn’t compare Vazquez and Wells, what the heck did I do? Good question. In the course of identifying matched pairs of pitchers, I found two guys from 1998 to 2007 in different ERA groups who fell on opposite ends of the won-loss spectrum. In other words, one pitcher had a good ERA (4.07; 110 ERA+) but a mediocre (93-86; .520) record. The other, despite a lousy ERA (4.90; 91 ERA+), achieved similar results (84-80; .512). Even better, these two men pitched in the same rotation for parts of six seasons.
With laziness again kicking in (you want me to comb through six years worth of starts?), I narrowed my focus to 2003. Neither Washburn nor Ortiz pitched particularly well that year, but one was rewarded for his “efforts” whereas the other was not. First, here are their final lines:
Both pitchers made the same number of starts. Washburn worked about one more inning per start and gave up about one less run, yet he ended up with six fewer wins and two more losses. That hardly seems fair, but there it is. The obvious assumption would be that Ortiz received better run support. In fact, he did: The Angels scored 5.50 runs in his starts but only 4.26 in Washburn’s.
That’s a good start, but let’s dig deeper. Here’s the first chart that James presented in his comparison of Jackson and Terrell, this time applied to our two heroes:
This is kind of a mess, so we’ll break it into smaller chunks:
Now maybe we’re getting somewhere. The Angels scored two runs or fewer in 15 of Washburn’s 32 starts (nearly 50 percent). Ortiz received such paltry support in only 25 percent of his starts. In the 3-to-5-run range, it is Ortiz who benefits much more than his teammate. The 6-to-8-run range is roughly even, and Ortiz again gains the advantage at 9 runs or more.
Let’s look at how each pitcher did with various amounts of run support. James was working with three years worth of data, which provided pretty good sample sizes for most of these categories; since we’re looking at one year, we’ll stick with the “chunks” of data, i.e., 0-2 runs, 3-5, etc.
Both pitchers won exactly one game when the Angels scored two runs or fewer, though it took Washburn seven additional starts to accomplish the feat. The Angels won his two no-decisions. Ortiz’s ERA is higher by more than a run, but this is deceptive: He also coughed up nine unearned runs in these starts. (Washburn, incidentally, allowed just four unearned runs over all his starts.) If you go by straight runs allowed, Washburn checks in at 4.67, whereas Ortiz is at 7.36.
As in our first case, Washburn pitched better, albeit in a smaller sample. Again with Ortiz, ERA doesn’t tell the whole story: His RA when the Angels scored 3-5 runs was 6.55.
For grins, here is what our two heroes did when the Angels scored five runs or fewer:
And of course, Ortiz gave up a bazillion unearned runs. If we compare RA, we have Washburn at 4.49 and Ortiz at 6.91. So basically, in these scenarios, Ortiz compiled a better won-loss record despite allowing more than two runs more per game than did Washburn. The guess here is that Ortiz faced weaker pitchers, but that is beyond the scope of our current exercise. (However, if we were truly interested in the case of Washburn and Ortiz, and not just in using them as an example to demonstrate a technique, we would want to continue excavating.)
Again, we see Ortiz with the inferior numbers and better record. The guy cleaned up when his team put runs on the board!
This is the only breakdown where Ortiz actually outpitched his teammate. Ortiz kept kicking while his opponent was down, whereas Washburn performed with relative indifference, even losing a July 23 contest to the Rangers despite being handed nine runs.
For the sake of completeness, here is how these guys did with six or more runs:
Washburn’s record is nice, but Ortiz’s is insane. Again, neither pitched particularly well in these circumstances, but Ortiz at least managed to reap some rewards for, uh, being there. (If we lower an already arbitrary cut-off and look at records when the Angels scored five runs or more, we see that Ortiz went an almost unfathomable 14-0 with a 4.13 ERA.)
Who cares, and what more can we do with this?
This is all very interesting, if a bit esoteric. Who would care about such things? Well, I can think of a few people who might be interested. Washburn and Ortiz, for example. Their agents. Their employers. Their prospective employers. You know, anyone who has a stake in how much money they get paid for their services.
Can Ortiz be counted on to win 16 games? Or is he someone who needs all the offensive help he can get? Is Washburn really a sub-.500 pitcher, or could he thrive with better run support (I’m thinking of 1984 Milt Wilcox, or maybe 1990 Scott Sanderson)? These are important questions to ask (and answer correctly) when millions of dollars are on the line.
As for next steps, that’s up to you. Every answer leads to more questions. We might wonder why Washburn received so much less run support. Did he face superior pitchers? Did he pitch a lot on days when his manager rested the starters? Did more of his starts come in pitcher-friendly environments? Really, there’s almost no limit to the way that you can parse this stuff; it all depends on your angle of interest.
The example we’ve used throughout of Washburn and Ortiz in 2003 doesn’t have a great deal of practical application today. However, maybe you’ll want to apply some of these concepts when tackling Webb vs. Lincecum. Or maybe you’re curious to learn why Bartolo Colon has a much higher career winning percentage (.607) than does Brad Radke (.516) despite the fact that their career ERAs (4.09 for Colon, 4.22 for Radke) and ERA+ (112 for each) are nearly identical.
Better still, maybe you’ve thought of ways to use this that I haven’t considered. Well, that is my hope anyway…
References & Resources
1988 Bill James Baseball Abstract, Baseball-Reference.