Fielding and Pitching, Part Two

January 25, 2004

A look at Charlie Saeger’s system for splitting fielding and pitching Win Shares.

Charlie Saeger, who was one of the contributors to the Big Bad Baseball Annual, has done a ton of Win Shares research, particularly into the fielding side of the system.  He has developed his own methodology for assigning pitching and fielding Win Shares to players, and I think it represents a big step forward.

Charlie explained his methodology in an e-mail to me, and the detail is mind-boggling.  I’m not going to go over the whole thing, but here are the highlights:

- Charlie approaches this in the same basic way James did, with a numerator and denominator representing the percent of defensive Win Shares to be allocated to pitching.

- Charlie sets the initial denominator equal to innings pitched divided by four.  Charlie does this in order to have a baseline that can incorporate other events equal to their linear weights.  I don’t fully understand the math, though the basic logic makes a lot of sense.

By doing this, Charlie makes his methodology a lot more transparent than James did.  This is good.

- Next, the numerator changes based on the number of strikeouts achieved by the pitchers.  Charlie’s methodology is set so that a team that averages about 0.5 strikeouts an inning will have 67% of its shares allocated to pitching, and a team that averages 3 strikeouts an inning will have 100% allocated to pitching.

So, you see, Charlie assumes the same baseline ratio of Win Shares allocated to pitching as James did. 

However Charlie’s system differs from the original at this point.  He subsequently adjusts the pitching allocation in the same manner, but his adjustments have more weight.

- For instance, the percent allocated to pitching changes according to the number of home runs allowed by the pitching staff.  There is a change of 1.6% points for every ten home runs allowed vs. the league average (vs. 0.9% points change in James’ system).

- Walks and HBP also impact the allocation up and down, about 0.5% points for every ten (vs. 0.2% points in James’ system) vs. the league average.

- Charlie also calls for similar impact allocations based on errors, passed balls and caught stealing, outfield assists (which James doesn’t include) and double plays.  These are all similar to James’ approach, though the impact of each is larger.

Charlie does something really interesting with Defense Efficiency Ratio (DER), using a two-step process:

- First, Charlie calculates park-adjusted DER vs. the league average (exactly like James does), and allocates 100% responsibility for this to the pitching.

- Then he calculates a second park-adjusted DER factor, called ADER.  ADER is calculated in the same manner as DER, except that assists are included in both the numerator and denominator.  Charlie’s system then allocates 100% responsibility for this to the fielders.

These two separate DER calculations are included to offset the fact that groundball staffs tend to have lower DERs, which is no fault of the fielders.  Ground ball staffs tend to have more assists, so including assists in both sides of the DER equation helps balance this out.

Charlie explained it to me in this way: “I ran a sample of NL teams for the late 1970s (I run a lot of things on this sample, actually) and found that DER correlated with PO-SO-A at about r=0.41 or so. This jibed with the idea that folks had about the DIPS H$ being higher for a groundball pitcher. Adding assists to both sides made this go away.”

Here’s an example of how it works.  The 2003 Los Angeles Dodgers’ DER was actually a tiny below league average, once adjusted for ballpark.  In Charlie’s system, the pitching is given 100% responsibility for this, and so the pitching allocation decreases from 75.9% to 75.6%.

However, the Dodgers’ staff was an extreme groundball staff, so there were a lot of assists among the fielders.  When you add assists to both sides of the DER ratio, yielding ADER, you find that the Dodger fielders performed about 23 runs above average.

100% of this credit is allocated to the fielders, decreasing the pitching allocation to 72.0%.

I’m not 100% sure of the validity of this approach, but I give Charlie an A+ for inventiveness.

The net effect of Charlie’s approach, vs. the original James approach, is a few more defensive Win Shares allocated to fielding overall, and a wider variance in results by team.

Attached is a list of defensive Win Shares allocated to pitching in the original system and Charlie’s system (assuming I interpreted Charlie’s directions correctly).

Team James Charlie Diff
Arizona 71% 74% 3%
Atlanta 70% 67% -3%
Chicago 70% 71% 1%
Cincinnati 68% 70% 2%
Colorado 68% 69% 1%
Florida 70% 68% -2%
Houston 68% 64% -4%
Los Angeles 72% 73% 0%
Milwaukee 66% 68% 2%
Montreal 69% 68% -1%
New York 67% 66% -1%
Philladelphia 70% 72% 2%
Pittsburgh 68% 67% 0%
San Diego 63% 51% -12%
San Francisco 68% 62% -6%
St. Louis 64% 57% -7%
Anaheim 66% 65% -1%
Baltimore 67% 68% 1%
Boston 72% 74% 2%
Cleveland 67% 65% -2%
Chicago 70% 68% -2%
Detroit 63% 60% -3%
Kansas City 68% 62% -5%
Minnesota 69% 70% 1%
New York 74% 80% 5%
Oakland 71% 71% -1%
Seattle 66% 62% -4%
Tampa Bay 61% 52% -9%
Texas 66% 58% -8%
Toronto 69% 71% 1%
Average 68% 66% -2%


San Diego? Well, their pitchers gave up 208 home runs in a very tough home run park, making their adjusted total equal to 226. League average was 168. They also gave up 70 more walks/HBP. With Charlie's larger weights on these events, the pitching gets much less credit for holding down the opposition.

Comments?

Can you, at the team level, show how much Win Shares the pitchers and fielders get in this new system?

And, if possible, the number of Win Shares an average team of pitchers and an average team of fielders would have each gotten, if they had played for each of the 30 teams?

Posted by tangotiger  on  01/26  at  09:23 AM

Yes.  I’ll post it sometime Tuesday.

But, the key point is that total defensive Win Shares, across the league, would still equal 52% of all Win Shares.

So an average team would have about 126 defensive Win Shares (81 times 3 times 52%).  Right now, an average team would allocate 85.5 of those to pitching, and Charlie’s system would allocate 83 of those to pitching.

I’m working on a third approach, which I’ll post in a couple of days.

Posted by studes  on  01/26  at  09:39 AM

I’m interested to see the differences in Win Shares between James/Saeger, and how those would compare to UZR.

Posted by tangotiger  on  01/26  at  02:52 PM

You might really, really like his third approach, Tom. grin

Incidentally, I make sure I can have adjustable baselines. I also sent Dave an alternate way of handling strikeouts, which would keep things working in a system where 0.61/1.61 are the baselines for offensive/defensive claim points (as you proposed). This has strikeouts as a purely linear variable, and keeps the fielder portion at about 25%.

Posted by .(JavaScript must be enabled to view this email address)  on  01/26  at  04:10 PM

I should also note I keep things on a very adjustable basis overall, so even if I have erred on a value or something (and I’m sure someone like MGL or DS will find that I have), I wanted the framework to be correctable and extendable. Dave has an additional “theory” e-mail wherein I give how I derive at values and how someone could use something like BaseRuns to make custom values for each team or something like that.

Posted by .(JavaScript must be enabled to view this email address)  on  01/26  at  04:13 PM
Page 1 of 1 pages
Commenting is not available in this channel entry.
<<Previous Article:  FIP and DER through History Next Article:  AL West Win Shares>>