Pitching and Fielding, The End
February 08, 2004
This has gone on long enough!
Bill James, Charlie Saeger and I have developed, somewhat independently of each other, our own systems for allocating defensive Win Shares to pitchers and fielders. A summary output of our systems for 2003 can be found in this table.
We began this Win Shares pitching/fielding split adventure in this December article, asking if pitchers are undervalued relative to fielders. The time has come to ask the question again, and to review the results of the three systems. So let’s start with a graph.
Here's a box whisker graph of the three systems, applied to all 2003 teams.
As a reminder, the horizontal blue line in the box whisker graph represents the average for each set of data. The red boxes represent the 25th and 75th quartiles, and the extended lines cover the rest of the data points.
As you can see, the original James system is very balanced, with remarkably even variances around the mean. This isn't surprising; the system was built to do this -- I'm guessing that the results for every year would look like this.
James put some strict restraints into his Win Share calculations. In fact, he put restraints on top of his restraints. As a result, the quartiles and extended lines you see in this graph will also look very similar from year to year.
Pete has pointed out to me that a team cannot achieve more than 52.4 fielding Win Shares in a 162-game season. Remember this article, in which we said that the 1999 Reds might have been the greatest fielding team of all time? They actually reached 52.8 fielding Win Shares, because they played 163 games. Should they have been credited with more? Probably.
By my count, sixteen teams have hit the maximum number of fielding Win Shares since 1960, including the 2002 Angels, and the 1969 Mets and Orioles. Some of these teams probably should have achieved more, but were almost certainly held back by the system's restraints.
Charlie's system drops all restraints. Charlie's results go further up and down than James' do, and the overall effect is to slightly increase the percent allocated to fielding. I think this is basically because there is more room to stretch down than up. You can see this in the graph.
My proposed system also stretches more than James' but not as much as Charlie's. Also, the net impact of my system is to allocate more defensive Win Shares to pitching -- about four points more than the James system, and six more than Charlie's. And my proposed system actually stretches up in the quartiles (giving more of an emphasis to pitching), though it evens out in the extended lines.
Let's talk about specific teams. In James' original system, the best fielding team in 2003 was Seattle, with 52.3 fielding Win Shares. Under Charlie's system, that total increases to 58.7, but the best fielding team turns into the San Francisco Giants, whose fielding Win Shares increase from 50.9 to 60.1. In my system, Seattle is still first with 52.3, and the Giants are second with 51.3.
Next, let's look at the teams with significant variances between our systems. I've linked each team's baseball team graphs in the title of each paragraph. The team fielding graphs should particularly help as you review the comments.
Boston: James' original formula allocates 72% of their defensive Win Shares to pitching. Charlie's system ups it to 74%, and my proposed system takes it all the way to 79%. In all three cases, Boston has the second-most shares allocated to pitching (behind the Yankees), so the systems remain relatively even. It's that the emphasis on pitching increases from system to system.
Let's look at the impact on some specific players. If you move from James' system to mine, Pedro's Win Shares increase from 22 to 23, and the player who received the most fielding Win Shares, Garciaparra, declines from 25 to 24. Fundamentally, the impact on any one player is not very big (or even statistically significant).
Arizona: Same story as the Red Sox, though the impact on some players is two Win Shares.
Atlanta: These guys are a little different. James ranked them third; Charlie dropped them to tenth; and I had them fifth (all in terms of percent allocated to pitching). The biggest reason is the ADER calculation, which favors the Braves' defense, and which is only included in Charlie's system.
This might make sense. The Braves did have a high GB/FB ratio last year (1.5 vs. a major league average of 1.25). We'll talk a bit more about ADER later.
Cubbies: Then we come to the Cubs. All three systems ranked them fourth, but the pitching allocation rose from 70%/71% to 79%. If you head on over to the Cubs Team Graphs, you'll see that the Cubs had the second-best FIP in the league, and their DER was worse than the Dodgers (on a park-adjusted basis). In general, I think this should make their allocation similar to the Dodgers'.
Dodgers: Speaking of the Dodgers, James ranked them first, Charlie second, Dave tied them for first. Both James and Charlie allocated about 72% of defensive Win Shares to pitching, but my system jumped that way up to 80%.
The Dodgers were an extreme groundball team, so Charlie's ADER calculation impacted the allocation significantly. On the other hand, my system catches pitching staffs with low FIP (the Dodgers' FIP was a park-adjusted 0.50) and gives them extra credit for non-FIP runs allowed.
Here's the thing about ADER. We think that groundball staffs have higher DERs, but we also think that flyball staffs give up more extra base hits. So, the research I've seen indicates that these two factors are primarily a "wash" when allocating run prevention.
Cincinnati: Then there's the curious case of the Reds. James pegged them at 68%, Charlie at 70% but I pumped them up to 76%.
The Reds were one of the worst defensive teams in the league. On a park-adjusted basis, they were second-worst in runs allowed, and their park-adjusted FIP was the third worst. So you might expect a slight leaning toward the pitching. This, along with the bias of my system towards pitching, seems to be the reason for the uptick.
The same thing happened with the Tigers. Bad pitching, bad fielding. What's to choose? San Diego too, though their fielding was better than Detroit's.
I have to admit that I like the "feel" of my proposed system the best. It increases the emphasis on pitching, but seems to maintain an emphasis on fielding for outstanding fielding teams.
However, this review is based on just one year's results. Bill James spent four years reviewing results from all of major league history, and I'm not going to argue with that. However, there may be an opportunity to revisit previous years or, at least, look at next years' Win Shares with this alternative approach.
In the Win Shares book, Bill James created lists of the greatest offensive, defensive and pitching teams of all time, but he pointedly did not include a list of the best fielding teams. He didn't trust his system enough to believe such a list would be valid. To quote (from page 132), "I put limits on the system to avoid results that I might not trust."
Good idea. But I think that the goal, for those of us carrying on the work, is to develop a system that can be trusted.
We think that groundball staffs have higher DERs, but we also think that flyball staffs give up more extra base hits.
Actually, we think that flyball staffs have higher DERs and give up more extra-base hits. Groundball staffs—more singles. The two do wash out, but since we don’t have XBHs for all of baseball history, we need the adjustment.
FYI—I have 8 years of results of the NL from the late 1970s, and find things to be in the same bounds as above.
Yes, sorry. I got my DER up from down wrong.
Still, I wouldn’t think you’d need ADER to correct for DER variances. DER may be lower for groundball staffs, but the impact on runs is about the same as that of a flyball staff, with its higher DER.
Posted by studes
on 02/09 at 02:59 PM
Page 1 of 1 pages
Commenting is not available in this channel entry.