Does this conversion thing work? Joakim Soria (Part 2)

Sean “Rally” Smith has updated his Baseball Projections web site with a brand new feature—historical pitching Wins Above Replacement (WAR). This made my plans to integrate park factors and leverage values into my Retrosheet database a slightly lower priority. Back to the topic at hand….

Should the Royals convert Joakim Soria into a starting pitcher?

Last week, I explored the pre-MLB history and PITCHf/x data that are available on Soria. Given his experience and repertoire, it seems like a valid point to ponder—would Soria and the Royals benefit by moving him to their rotation?

There are a lot of angles in play here, from Soria’s health (he’s shelved for a few days due to shoulder soreness) to the other moves necessitated (move one starter somewhere and shuffle the bullpen) and even the “chemistry” component. In today’s game, how much of an edge would be lost by removing a proven and reliable closer from his role?

So, I don’t think I can answer the question as stated—hence the title of this article.

Does this thing work?

While we all can think of teams converting a closer to a starter, I was a little surprised by how rarely something like that happens. Plenty of pitchers break in as a relievers and become starters, but the closer to starter conversion is fairly uncommon.

To find comps for the hypothetical swap of Soria, I hit the Retrosheet database with a query (complete code below) to find pitchers who had at least 10 saves in 50 games, and fewer than five starts, in “year 1” and had at least 10 appearances in “year 2” where they made at least nine more starts than relief appearances. In other words, to qualify with one relief appearance, you needed 10 or more starts. Twenty relief appearances, and you’ll need 29 starts to go with them to qualify. The cut-offs are arbitrary; you can play with them on your own if you so choose. Data covers the Retrosheet era, 1953 onward.

Based on the parameters selected, I ended up with 16 pitchers:

Card Corner Plus: Gene Michael and High Intelligence on 1972 Topps
Three smart players devoted their lives to baseball.

Pitcher y1 y2
Rick Aguilera 1995 1996
Miguel Batista 2005 2006
Gary Bell 1965 1966
Shawn Chacon 2004 2005
Ryan Dempster 2007 2008
Scott Garrelts 1988 1989
Rich Gossage 1975 1976
Danny Graves 2002 2003
Jimmy Key 1984 1985
Craig Lefferts 1991 1992
Derek Lowe 2001 2002
Bret Myers 2007 2008
Bob Shaw 1964 1965
John Smoltz 2004 2005
Bill Swift 1991 1992
Wilbur Wood 1970 1971

Now, it isn’t fair to base this comparison on just these two seasons. Each player has some history coming in and some future history to make going out. Well, almost each player. Key was a rookie in “year 1”, Dempster and Myers are just now living their post-“year 2” future and Soria hasn’t even had “year 2.” Let’s compare the before and after phases for each of our 16 comps, plus Soria. in terms of games started.

Click any of the charts below for a larger version


* rookie in “year 1,” no “before”
** “year 2” was 2008, no “after”
*** no “year 2,” no “after”

Aguilera had experience as a starter and a reliever before entering his “year 1.” After one season as a pure starter, he moved back to the bullpen for good.

Batista and Bell had similar usage patterns, mostly sticking as a starter after the switch.

Chacon went from starter to reliever abruptly, but from “year 2” on he played both roles.

Dempster had a long record as a starter before the Cubs made him their closer. He seems settled back into the middle of the rotation role now.

Garrelts and Gossage both had minimal experience as starters before becoming relievers and then starters in earnest. Garrelts stayed in the starting role, Gossage went back to the bullpen. I think that worked out all right for the Goose.

Graves had one year as a starter, aside from a few starts in “year 1.”

Key barely makes the list, with 10 saves as a rookie. He rarely made relief appearances beyond that first season.

Lefferts had made some starts about a decade before he went from a reliable 20-save guy to the rotation. It didn’t last; he finished his career in the bullpen.

Lowe has made the switch to starter very well, judging by his ground ball rates and paycheck.

Myers lasted one year in relief. It’s too soon to know if that’s forever, but it seems that he’ll be a starter for the foreseeable future.

Shaw barely qualified, with 11 saves, and spent most of his career, before and after, as a starter.

Smoltz had a long career as a starter, a short one as a closer, and another short one as a starter. The latter remains in progress, however.

Swift’s time in the bullpen was brief; he earned more starts as his career progressed.

Wood, not Kerry, made an impressive leap (see below) and never looked back.

The numbers

This shows WAR during both of the sample years, and you’ll see that Gossage was an amazing reliever and Wood was a brilliant starter.


Now the difference from “year 1” to “year 2,” with positive numbers indicating an improvement.


For the most part, the moves worked out well for the teams. The average gain was 2.2 wins, with winners beating losers 11-3 (Myers was a wash). But look closely, and you’ll see only two pitchers had a “year 1” WAR better than Soria’s: Gossage, who cost his team a few wins by moving to the rotation, and Swift who barely improved, in terms of WAR, as a starter. Swift also had more experience in the major leagues as a starter.

Don’t do it

Not often does a successful reliever become a successful starter. Soria has a history—and a present—of arm problems. His best pitch, in terms of throwing strikes, is a fastball that doesn’t average over 92 mph in short spurts. Drop him a couple mph to imagine him starting. Throwing high-80s without secondary pitches that can find the zone could be a deadly combination. It’s just my opinion, there’s no definitive answer, but I’d have Soria drop one of his pitches and stay in the bullpen. It probably will be easier for him to throw strikes by keeping fewer pitches sharper.

References & Resources
Comp Filter:
seasons.* from
select namefirst, namelast, pid1, year1, ip1, g1, gs1, sv1, year2, ip2, g2, gs2, sv2
(select yearid as year1,playerid as pid1,sum(ipouts)/3 as IP1,sum(GS) as GS1, sum(G) as G1, sum(SV) as SV1
from pitching
group by 2,1 having sum(gs)<5 and sum(G)>=50 and sv1>=10) as pl1
(select yearid as year2,playerid as pid2,sum(ipouts)/3 as IP2,sum(GS) as GS2, sum(G) as G2, sum(SV) as SV2
from pitching
group by 2,1 having sum(gs)-sum(g) > -10 and sum(G)>=10) as pl2,

where pl1.year1 + 1 = pl2.year2 and pl1.pid1=pl2.pid2 and year1 >= 1953
and master.playerid=pid1
) as seasons


WAR Data

Print This Post

Comments are closed.