Sunday, November 06, 2005

General Manager Statistics

The Theo Epstein brouhaha has prompted me to think more about general managers, their importance, and, mostly, how to measure their impact. I was surprised to discover that Won-Loss records for general managers are not readily available on the Internet. I started to build my own database of such records, but then thought, we can do more than just Won/Loss records -- there are good databases of transactions available at Retrosheet and Baseball Reference, and we could assign a GM (or 2 in the case of trades) to each transaction. Then, using Bill James's Win Shares , we could assign a value of Win Shares Gained or Lost for each transaction, based on how many win shares the player earned for the signing team, subtracting the Win Shares that a player the GM gave up earned for another team. Ideally, we could have an entry for each GM that would show how many win shares he gained/lost in free agency, in trades, in the June draft, in waivers, etc. A GM stats year could run from October 1 to September 30.

I ran the numbers for Jim Bowden's past year. Here are the Win Shares for each of his transactions (I've omitted most of those that had zero Win Share impact):

Signed Vinny Castilla (+12 WS)
Signed Cristian Guzman (+3 WS)
Traded for Jose Guillen (+18 WS) by seding Juan Rivera (-9 WS) and Maicer Izturis (-6 WS)
Released Chad Bentz (+2 WS -- He actually had -2 WS for the Marlins, so getting rid of him was a good thing)
Drafted Tony Blanco under Rule 5 (1 WS)
Signed Wil Cordero (-2 WS)
Signed Rick Short (2 WS)
Signed Esteban Loaiza (11 WS)
Traded for Marlon Byrd (7 WS) by sending Endy Chavez (0 WS)
Released Claudio Vargas (-6 WS)
Traded for Junior Spivey (3 WS) by sending Tomo Ohka (-4 WS)
Claimed Ryan Drese (0 WS)
Signed Mike Stanton (2 WS)
Traded for Preston Wilson (10 WS) by sending Zach Day (0 WS) and JJ Davis (0 WS)
Lost Sunny Kim to waivers (-4 WS)
Trade for Deivi Cruz (1 WS) by sending minor leaguer
Drafted Ryan Zimmerman (3 WS)

Bottom Line: Bowden's moves earned a net positive 44 Win Shares this year. Each WS is 1/3 of a win, so his moves were worth about 15 wins. Plus, the book isn't closed on Bowden's moves yet -- he should also get credit for whatever Ryan Zimmerman (and other players he acquired) do this year and beyond for the Nats.

For comparison purposes, I ran the numbers for Theo Epstein's 2004 transactions, and he came out with 84 Win Shares, but that's a bit unfair, since I included the 2005 stats for the players involved (I think for just 2004 he would have gotten something like 75 WS).

Anyway, in poking around the various databases, I've determined that it would not be too hard to build WS into existing transaction databases and add GM information to them so that you could generate these stats for history. The one big problem is that a historical database of Win Share information is not readily available. I'm suprised that someone hasn't done this already.


At 8:22 PM, Blogger SuperNoVa said...

I really like this, DM, but have some comments.

I think we'd have to figure out what the marginal salary cost of a Win Share is to truly judge Bowden's performance. Spending $4.2 million on Guzman for 2 win shares seems like a bad deal. But if you pick up a bench player for the minimum and he gives you 2 win shares, that's saying a little bit more.

I'm shocked that Guzman had positive win shares - he was actually below a replacement player in terms of offensive production. I think he was 8 runs above replacement in the field and 8 runs below replacement in the field - meaning that he really was a replacement level player.

At 9:03 PM, Blogger DM said...

You raise two good points which the Hardball Times is already focused on: Win Shares Above Replacement Player (which Bill James refers to a Pandora's Box with three lids) and working salary into this. I think these are worth pursuing, but we should start simply, in the same way that Baseball Reference gives us the basic counting stats, which we assimilate to others by using our knowledge of what is good/bad. Yes, I don't really know what 44 Win Shares means either, but I'd love to just see the history of GM years and start getting an idea.

On Guzman, he was negative until September, and I think his batting win shares still ended up negative. But you can start putting that into context now -- he had the same number of Win Shares as Ryan Zimmerman.

At 9:08 AM, Blogger El Gran Color Naranja said...

Normally I don't plug myself but dammit this took a lot of time, so plug plug plug!

Before the year started I tried to do a historical WS analysis of Bowden as a GM, the results were interesting but fairly inconclusive. PLUG!



The biggest issue I have with the WS idea is a losing WS deal may in fact be a positive for your team (and vice versa) depending on the circumstances of the replacement players you have available. Like Nick Johnson for Vernon Wells might get us a gain of +3 WS, but if the real consequences of the deal is Nick Johnson and Brad Wilkerson starting versus Carlos Baerga and Vernon Wells starting, it would be a net loss for the team. Sort of the same thing supernova is saying.

I know that getting the raw WS numbers we got is not meaningless, but I think it could lead to poor conclusions, if things aren't judged in context. Of course that creates tons more work for a historical researcher, but looking at this year is probably possible.

At 9:24 AM, Blogger DM said...


Thanks for the links (and the hard work). We don't mind plugs here if they link to something good.

Sure, a straight WS analysis leaves out a lot of context. The Castilla-Lawrence trade is good because it frees up room for Zimmerman, something that might not be caught in this type of stat (but if Zimm does well Bowden would get credit). But so do most counting stats, since they don't tell you the context of the home run, double, strikeout, etc.

I think we can't tell how useful these would be without seeing where each GM fell within the universe of GMs. Then we might start getting a sense of what the numbers mean.

At 9:56 AM, Blogger Basil said...

Calculating GM Win Shares? You guys clearly need to get out more. ;-)

At 9:59 AM, Blogger DM said...

Ummm ... I guess I shouldn't mention that I was doing most of this work on Saturday night.

But you're right Basil. I'm turning over a new leaf. This weekend I'm going to write Advisory Opinions on Carlos Baerga ;)

At 10:14 AM, Blogger El Gran Color Naranja said...

"We don't mind plugs here if they link to something good"

Then I apologize.

I agree that we'd need to see how all GMs are doing to get a feel for these numbers, I just think there may be a lot of misclassification without context. I think the correlation between overall WS and "contextual" WS would be strong over the course of a GMs career, but as a yearly measure I'm not so sure.

At 10:14 AM, Blogger Basil said...

Okay, but don't take the responsibility lightly. It takes a lot of care and dedication to spend 15-20 paragraphs cultivating twin conclusions that "How should I know whether this guy is a 'clutch hitter'?" and "An old guy who's a free agent and can't play defense: maybe he shouldn't be brought back." ;-)

At 7:20 PM, Anonymous Will said...

Hardball times had some trade analysis by win shares a while back, but I can't find it now.

At 9:50 PM, Blogger Basil said...

That's a good point. HT is really into the Win Shares . . .


Post a Comment

<< Home