The first thing writers learn is to start off with a good lead. Don’t jump in assuming everyone knows what you’re talking. Lead them into your story, and go slowly. Well that’s just boring, I say, so I’m deleting my first three paragraphs and getting to the good stuff. Hang in and enjoy the ride.

Let’s say a team allows one run per game while scoring five. How often are they expected to win? About 93.4% of the time, actually. See how quickly I came up with that answer? It just inspires confidence, doesn’t it? Okay, so here’s the follow-up: How many runs would a team that *allows* five runs a game need to score to win 93.4% of the time? The answer is about 15.4.

And here’s the point of that: A run saved is not equal to a run scored. Keep that in mind, because it’s important. In fact, that motto is what forms the rest of my article.

***

Astute baseball people like to use runs above some baseline—average, replacement—to evaluate players. It’s simple, and it also allows us to compare pitchers and hitters. If we’re just comparing hitting performance, we can look at Runs Created. But if we need to compare hitters to pitchers, we need to introduce some kind of baseline—like runs above average—to have meaningful numbers for the two. Or so goes traditional statistical thinking. So it goes.

The thing is, as shown in the example above, that baseline isn’t *really* the same thing. A pitcher who allows one run per game is exactly as valuable as a lineup that scores 15.4, but if we looked at runs above average (assuming that an average team scores/allows five runs a game), the pitcher would only be four Runs Above Average, while the lineup would be 9.4 RAA. The offense comes out looking 235% better, despite actually having an equivalent output. Obviously, that’s a problem.

There may not be a lot of major league pitchers that allow one run per game, but there are those that allow two, three or some other number low enough to render a baseline such as average useless.

This isn’t a problem that I’ve seen discussed much in the baseball community, even though I think it’s a fairly serious one, especially when we’re talking about comparing players’ value, both in a monetary sense, and in a historical sense. It’s a problem I noticed almost a year ago, and one that I’ve worked on since.

Well, about half a year ago, I came up with a solution, I’ve made many revisions since, and improved my model but the basic idea is the same: Convert runs allowed into runs scored. Taking the example from the beginning, I’d say that the pitcher who allows one run a game is equivalent to a line up that scores 15.4. Radical, yes, but also mathematically equivalent.

I’d make a list of the cool things about this idea, but The Hardball Times doesn’t give me enough space (I’m channeling my inner-Maddox here). Here’s what I do have space to include:

1. It’s easy to use for comparison purposes. Want to know how Roger Clemens compared to Derrek Lee? Well, take Clemens’ Pitching Runs Created (PRC) and Lee’s Runs Created, and there you have it: A comparison! There are no hoops to jump through, no 52/48 splits to make (that sounds painful doesn’t it?), nothing. If you want to adjust for defense or position, just add Lee’s fielding runs above average (using Range , preferably), or add a positional adjustment straight to his Runs Created. That doesn’t sound right, but because of the way that Pitching Runs Created work, it is. Don’t believe me? Stick around for a few more paragraphs.

2. It tells you what a pitcher’s *absolute* value was. No more messing around with replacement levels or whatnot; Pitching Runs Created are just like Runs Created for hitters. Yeah, you can put in a replacement level if you’d like, but there isn’t really a need to.

3. It can be used for all kinds of Win Shares fun. See, in Win Shares, Bill James looks at offense and pitching using the concept of marginal runs. The idea is that the marginal level (52% of an average player for hitters; 152% worse than an average player for pitchers) “is not a replacement level; it’s assumed to be a zero-win level.” That’s all fine and dandy except for the small fact that it’s not. A team composed of “marginal” players is going to win 11.3% of its games. Since when does .113 = 0?

The reason that James does this is twofold: One is that he wants to dole out more credit to star players, which I won’t touch upon right now, and the other is that I don’t think he could figure out how to put pitchers on an absolute scale. Bill James is a smart, great man, but I remain convinced that he couldn’t figure this out, and that’s why he resorted to using the marginal run method. Win Shares is supposed to be an absolute value method, but it’s not really, and I believe that this is the case because James could not figure out absolute value for pitchers. Well, using Pitching Runs Created, we can scrap marginal runs and do Win Shares using absolute value.

Anyway, just as you thought this article couldn’t get more boring, I’m going to shock you all, and go into the nitty-gritty of my system. I’ll try to make this as painless as possible.

The first and most important part is simple, and I’ve already explained it. I simply convert a pitcher’s runs allowed into a Runs Created figure. Before doing so, I do make the first of two adjustments for fielding, adjusting the pitcher’s runs allowed for his Batting Average on Balls in Play (BABIP). Using the research of those much smarter than me, I’ve found that the best split is about 80/20 for fielders/pitcher. So if a pitcher allows 10 runs less than expected based on his BABIP, and he’s allowed 50 runs on the season, I say that he’s allowed 58 runs (given that batters put 700 balls into play against him). The amount of credit (or demerit) a pitcher gets for his BABIP is dependent on the number of balls put into play against him: The more balls in play, the more credit.

I could use DIPS 3.0 instead, and maybe in the future I will, but for now this method will suffice.

Anyway, I make one more adjustment for defense that’s actually kind of cool. This one is more of a “how much credit should the pitcher get?” type of thing. Using a fitted curve extrapolated from empirical data, I find what percentage of those runs created should be allocated to the pitcher. This is done solely based on what percentage of a pitcher’s outs are generated by strikeout; the higher the strikeouts per nine innings, the more credit the pitcher will get. This is based on the fact that fielders have less impact on runs allowed as strikeout-rate goes up, and is based on the standard deviations of runs allowed at various strikeout per nine innings levels. The higher the strikeout per nine innings, the lower the standard deviation. In today’s run environment, the average pitcher will get 69% of the credit for his Runs Created.

Okay, it took me 1200 words but I’m done with the boring part, I swear. Let’s see some actual Pitching Runs Created results from 2005. T he AL leaders first:

Last First year Team PRC Santana Johan 2005 MIN 145 Lackey John 2005 LAA 120 Buehrle Mark 2005 CHA 110 Johnson Randy 2005 NYA 106 Colon Bartolo 2005 LAA 104

Johan Santana clearly deserved the Cy Young award, and maybe even some MVP consideration. What’s sad is that the actual Cy Young winner, Bartolo Colon, was not even the best pitcher on his team. John Lackey takes that prize. Oy. In the National League, we saw more bad decisions:

Last First year team PRC Clemens Roger 2005 HOU 147 CarpenteChris 2005 SLN 129 Pettite Andy 2005 HOU 127 Oswalt Roy 2005 HOU 126 Willis Dontrelle 2005 FLO 122

I guess you can say that things balanced out given that Randy Johnson was more deserving of winning last year when Clemens got the Cy Young, and that karma decided to repay him this year, as Clemens lost due to the same exact reason: No run support resulting in a mediocre won/loss record. What’s really amazing is that the Astros had three of the top four pitchers in the National League. With Brad Lidge closing out games, they didn’t need an offense. Well, not until they met the White Sox, at least.

But this is just a small taste of what I’m offering you. Attached at the bottom here is a spreadsheet with Pitching Runs Created figures for every player in the major leagues last year. Whatever you do with this, have fun and be safe, and remember: A run saved is not equal to a run scored.

Pitching_Runs_Created_2005.xls

**References & Resources**

Walk Like A Sabermetrician

## Leave a Reply