The Pitching/Fielding Split
January 19, 2004
A closer look at how Win Shares splits responsibility for runs allowed between pitching and fielding.
We’ve spent the offseason picking apart Win Shares, understanding it better, making some suggestions for improvement. Up to now, however, we’ve primarily dealt with offensive Win Shares, which are pretty much based on runs scored and runs created. A veritable piece of cake compared to the next step: defensive Win Shares.
I’ve spent some time circling around defensive Win Shares, checking them out, trying to figure how to approach them. And I still haven’t figured it out. But let’s at least start with the first step: allocating defensive Win Shares between pitching and fielding.
Bill James fully intended for others to play with and improve upon Win Shares, but he made defensive Win Shares maddeningly hard to analyze, frustrating those of us who would like to follow. See, he developed a Claims Points system that feels very logical at times, but downright arbitrary other times. Let’s review his pitching/fielding split as an example.
Here are the eight steps James established to allocate total defensive Win Shares between pitchers and fielders.
First, he established a ratio of 650/1097.5, or 59%, as a baseline for the percent of Win Shares that should be allocated to pitching. Why? I have no clue.
Second, he added a winning percentage factor to both the numerator and denominator of the ratio. The factor is 450 times team winning percent.
The impact of adding this factor to both the numerator and denominator is to increase the percent of defensive Win Shares to pitching. If the winning percent is.400, the pitching allocation increases to 64.5%. If it’s .500, the allocation increases to 65.6%. And if it’s .600, the allocation increases to 66.6%. So the formula assumes that pitchers are more responsible for overall team success than fielders.
This actually is a good assumption, given our previous work on FIP and DER. I’m not sure the use of a won/loss record to account for this phenomenon makes sense, however.
The following steps are more logical (or, at least, more understandable). I’m going to review the steps and isolate the impact of each one, using a baseline of .500 ball (65.6% allocation to pitching).
Third, the percent allocation is adjusted according to the team DER. If DER equals the league average, the pitching allocation decreases to 63.5%. If it’s ten points higher (say, .710 vs. .700), the allocation decreases to 63.1%. If it’s ten points lower (.690), the allocation decreases to 64.0%.
DER has to be forty points lower than league average in order for the pitching allocation to remain at 65.6%. So, okay, I don’t totally understand this calculation either.
Fourth, the pitching allocation increases for every batter struck out by the pitchers. Actually, the allocation increases to 68.5% even if the pitchers strike out no one. And then it increases 0.5 to 0.6 points for every batter per nine innings that the pitchers strike out. The exact allocation actually decreases somewhat as the strikeout rate rises.
If the pitchers were to strike out every single batter, the pitching allocation would be 80%. That’s not very likely, but I thought I’d check it out. Really, this step makes a lot of sense, but I have no idea if the proportions are correct.
Fifth, the pitching allocation changes based on the number of walks issued. If the pitchers give up the league-average number of walks, the pitching allocation actually increases to 71.1%. If the walk rate is .1 walk/inning lower, the allocation rises to 73.6%. And if it’s .1 walk/inning higher, the allocation decreases to 68.1%.
That’s roughly 0.2% points for every ten walks difference from the league average over a full season.
Sixth, you’ve got your adjustment for home runs. If the pitchers give up the league-average number of home runs, the pitching allocation increases to 71.1%. If they give up ten more home runs (over a full season), the allocation decreases to 70.2%. If they give up ten less, it increases to 72.0%. So, 0.9% points for every ten home runs.
Seventh, is the adjustment for errors and passed balls. If the team’s fielders make the league-average number of errors and passed balls, the pitching allocation declines to 62.2% (remember, our baseline is 65.6%). If the fielders make ten less errors/passed balls, the allocation further decreases to 61.8%. If they make ten more, the allocation increases to 62.7%.
That’s 0.4% points for every ten errors or passed balls.
Eighth, (the last step!) is the adjustment for double plays, based on the expected number of double plays. If the fielders turn the expected number of double plays (based on opportunities and league average), the pitching allocation decreases to 62.2%. If they turn ten less double plays than expected , it increases to 62.8% and if they turn ten more than expected, it decreases to 61.7%.
That’s roughly 0.6% points for every ten double plays.
So what does all this mean? Well, I dunno. But the essential point is that the formula starts with an assumption that about 65% of all defensive Win Shares should be allocated to pitching, and it then refines the exact allocation from there. James designed the system so that an average team would have 67.5% of its defensive Win Shares allocated to pitching.
In the majors last year, the pitching allocation ranged from a high of 74.3% for the Yankees, vs. a low of 61.3% for the Devil Rays. It makes sense that these two teams are at the extremes, but do the exact allocations make sense? Dunno.
So I’m going to end this article on that note (seems appropriate somehow). Let’s pull a Mike and call it Part One. In the next few articles, I plan to introduce a couple of other approaches that should help us close the circle on this particular issue.
In one of my experiments with DIPS, I looked at how runs allowed related to changes in performance.
My recollection is that the ratio of the standard deviations of pitching and fielding was very close to the win share ratio. I don’t believe I kept the spreadsheet, though.
I wouldn’t be surprised if the ratio kept working right, VOU. (Voux?) Things have a way of working out in the James system, though no one seems to know how at times.
Voice, if you can recall some of your general thoughts and approaches, I’d appreciate it. I, too, believe that James’ ratios are about right overall, but I think they can probably be improved when applied to specific teams. And I think that having a system that makes the fundamentals more explicit would make Win Shares more accepted in the baseball community.
Posted by studes
on 01/20 at 03:25 PM
Page 1 of 1 pages
Commenting is not available in this channel entry.