We Blew It

OK, everybody who thought that last year’s trade of Alfonso Soriano for Brad
Wilkerson
was a good deal
for the Rangers, hold up your hand. C’mon, get ’em up—don’t worry, you’re not alone. My own mitt is sheepishly raised.
When the trade was made about a year ago, many commentators cited the
trade as another example of Washington general manager Jim Bowden’s ineptitude. I don’t want
to single out any particular blogger; if you want to go back and read
individual posts, it’s easy to find them (Google “+Soriano +Wilkerson
+trade” for a sampling). Suffice to say that most bloggers thought
Bowden had made a major mistake—and it wasn’t even close.

I didn’t
actually write anything about the trade, but I nodded knowingly
while reading those pieces and was totally in agreement with my
sabermetrically inclined colleagues. Now it’s time to admit the
mistake.

The Trade

First some background: in December of 2005, the Nationals and Rangers
made a four-player trade: the Nats obtained Soriano while the Rangers got
Wilkerson, fourth-outfielder Terrmel Sledge and a minor league pitcher
named Armando Galarraga. Sledge was 28 years old, had
missed the previous season with a torn hamstring, and even when he
played he wasn’t very good. A month later he was a throw-in in the big
trade with San Diego, which sent Adrian Gonzalez and Chris Young to
the Pads in exchange for Akinori Otsuka and Adam Eaton. Galarraga, at
the time, was a 24-year-old mid-to-low-level prospect, who had
struggled in a half-season at the Double-A level. The trade involved
money issues as well: Soriano was set to make about $10 million, while
Wilkerson was making $4 million. Also, Soriano was to be a free-agent
in 2007, while Wilkerson would remain under the Rangers’ control for
two full seasons.

So, the deal was a bit complex, but let’s just break it down as
Soriano vs. Wilkerson. If you go back and read the commentary at the
time, just about everybody was evaluating the trade by comparing these
two players straight up. That makes sense in a way, as you often hear the
“winner” of a trade is the team who gets the best player in the
deal. So who got the best player in this deal?

Here’s what the two players ended up doing in 2006:

+-----------+------+------+------+----+------+------+------+-------+-------+-------+
| Player    | PA   | R    | H    | HR | RBI  | SB   | CS   | AVG   | OBP   | SLG   |
+-----------+------+------+------+----+------+------+------+-------+-------+-------+
| Soriano   |  726 |  119 |  179 | 46 |   95 |   41 |   17 | 0.277 | 0.351 | 0.560 |
| Wilkerson |  363 |   56 |   71 | 15 |   44 |    3 |    2 | 0.222 | 0.306 | 0.422 |
+-----------+------+------+------+----+------+------+------+-------+-------+-------+

There’s really no contest, is there? Wilkerson missed half the season
because of injuries, but even when he was in there, his performance
paled beside that of Soriano. So, what happened? How did so many
people pick Wilkerson to have the better year, or at least claim that
Wilkerson is the better player?

The Almighty Walk

Well, one thing that I think happened is a lot of saber-minded folks
don’t really like players like Alfonso Soriano. Why? Because they
don’t walk much and hence they don’t excel in getting on
base. On-base-percentage is justly regarded as the single most
important contribution to run scoring and, at the time of the trade,
Wilkerson had a big edge over Soriano in career OBP: .365 to .320.
But, as the saying goes, there is more than one way to skin a cat, or
produce on a ball field. Soriano had a big edge of his own: the
ability to drive the ball for extra bases. Soriano’s career slugging
percentage (through 2005) topped Wilkerson’s, .500 to .452. Now, it’s
generally accepted that a point of OBP is more valuable than a point
of SLG, so perhaps Wilkerson has an edge here due to his higher OBP.
But, Soriano had other advantages over Wilkerson: his speed and the
fact that he played a premium defensive position.

Through 2005, Soriano had stolen 169 bases at an 80% success rate,
while Wilkerson had swiped only 43 bases at a decidedly uninspiring
success rate of 55%. And, although Soriano was generally considered a
poor second baseman, there is value in getting a big bat in the lineup
at a traditionally weak-hitting position. Now, whether Soriano’s bat
was actually compensating for his defense—that is not an easy
question to answer with any degree of certainty. But we can try. Using
one defensive metric, Baseball Prospectus’ FRAA (that’s Fielding Runs
Above Average), Soriano averaged -9 runs for 2001-2005. The offensive
difference between an average-hitting second baseman and an
average-hitting corner outfielder is generally larger than nine runs,
so despite the weak glove, Soriano’s defensive position is a net
gain. Wilkerson is an OF/1B type.

So, Wilkerson looked to be a slightly better hitter perhaps, but one could
argue that Soriano compensated with his speed and by playing a premium
position on defense. In any case, it’s seems hard to make the case,
at this point, that Wilkerson was clearly the better player,
which is what we were saying a year ago.

On The Road

I haven’t mentioned park effects at at all and the fact that the trade
had Soriano moving from a hitter’s park in Texas to a pitcher’s park
in Washington had many predicting hard times for Alfonso in
2006. Some pundits at the time sought to shed light on the home park
issue by comparing only the road stats of the players. And when you do
that, a red flag certainly goes up for Soriano:

Home/Road Splits, 2004-2005
                Home             Road
Soriano    .316/.357/.593   .234/.278/.409
Wilkerson  .250/.371/.456   .253/.355/.449

The problem is, just looking a player’s road performance and giving it
preference to his full record is just a bad idea. For one, doing this
throws out half the information we know about the hitter, how is that
going to increase our knowledge? It’s much better to correct
the home stats with a multi-year park factor (we’ll get to park
adjustments in a minute). Secondly, many players play better at home
than on the road. We know this because teams win 54% of their home
games and only 46% of their road games. So, you don’t expect a
player’s road stats to reflect his overall ability. A third problem
with focusing only on road stats is that, because of the unbalanced
schedule, a team’s road ballparks are not
necessarily “average”. All three of Texas’ AL West rivals play in
parks that have tended to suppress hitting in recent years.
(This last point is also true when “correctly” applying
park adjustments, but the error is compounded when you only consider
the road stats.) And if you don’t believe me on this road stats issue,
here’s what noted sabermetrician Mitchel Licthman thinks about the practice:

For one, or even several, years of data only, using road
stats only is a HORRIBLE method of estimating player neutral
performance or talent. Absolutely horrible. To simply ignore half of a
player’s performance because you are worried about properly
neutralizing those home stats (which is admittedly problematic) is the
epitome of throwing out the baby with the bath water.

The Perils of Park Factors

OK, so considering a player’s road stats while tossing his home stats out
the window is not the way to go. A better way is to apply
park adjustments to a player’s complete stats to try to take the effect of the
home park into account. This is actually a tricky subject, because
there are different park factors on the market and if you apply one
blindly, you may not be doing the right thing or even doing what you
want.

Park factors in their
basic form are fairly simple: you calculate the number of runs scored
in, for example, Rangers home games and divide that by the runs scored in
Rangers away games. That gives a reasonable result, but you can improve things if you
actually consider the number of innings played in the home and away
games: the innings count can vary quite a bit, especially for teams
with very good home records: the winning team does not usually bat in the
bottom of the ninth. There is also quite a bit of
statistical fluctuation in park factors—analysts attempt to
minimize those by 1) using regression to the mean and/or 2) combining
data over a period of several years.

OK, I don’t want to give a treatise on park factors, but I wanted to
point out that they aren’t as simple as they seem. I’m wondering if
some confusion about park factors contributed to the pessimistic
judgments of Soriano a year ago. Ameriquest Field in Texas is rightly regarded as
a hitter’s park, but it’s not Coors Field by any stretch of the
imagination. The five-year park adjustment for Ameriquest from
2001-2005 is 107 (without any regression), which is a good hitter’s park, but we’re talking
about a 7% correction to a Ranger’s stat line. This can in no way
explain Soriano’s home/road stats that we saw above.

Soriano’s new park in Washington, RFK Stadium, had a park
adjustment of 93 in 2005, the only year of data available for that
park. I’m using Baseball Reference as a source for park adjustments, because
I’m guessing that many people do the same. That’s a good pitcher’s
park, but it’s only based on one year of data, and hence is “polluted”
by statistical noise. In fact, the park adjustment for RFK in 2006 was
much higher, at 97. Where the number settles down is anybody’s guess,
but the point is that putting too much faith in a one year park factor
is rather dangerous.

A Hardball Times Update
Goodbye for now.

What about Wilkerson’s home parks? Wilkerson had been a regular from
2002-2005 at the time of the trade and his one year park adjustments
from that period are 101, 118, 95 and 93. The last number is the
already-mentioned RFK 2005 value and the 2003-2004 numbers (118 and
95) represent “combined” park factors when the Nats played a bunch of home games in
Puerto Rico. I’m rather skeptical of these Monteal/Puerto Rico park
adjustments, especially since the two years show such a wide variation—a 23% change in park adjustment in one year is very unusual.
I’m not at all confident that those games in Puerto Rico did not
add in even more “pollution” to the one-year numbers. To accentuate
this issue, Wilkerson had his very best season in 2004, with an OPS of
.892. The low park adjustment for 2004 gives him on OPS+ of 128, which to me
looks like it’s overestimated.

A last note on park factors: I actually don’t like the Baseball Reference park
factors, mainly because (since 1999) they are based on only a year’s
worth of data (although the description of the park factor calculation
at bb-ref.com says the park factor is based on 3 years worth of data—this is not correct). Here at the Hardball Times we use the

park factors calculated by the sabermetrician known as USPatriot
. USPat uses five years of
data (when available) and also includes some regression to the
mean. His results can differ quite a bit from the Baseball Reference numbers:
for example the 118, 95 that bb-ref.com finds for Montreal in
2003-2004 are calculated by USPat to be 102 and 104. Note that using
this improved park factor, Wilkerson’s 2004 OPS+ goes from 128 to
117. Oops.

Age and Durability

Is there any reason to think that Soriano’s projection going forward
from 2005 should have been downgraded due to age and/or durability issues? I
don’t think so. Soriano’s 2006 seasonal age was 29, Wilkerson is one
year younger. It’s true that Soriano was past the canonical peak age
of 27, but the downward slope of the aging curve is very shallow at
age 29 and projecting a steep dropoff due to age would have been
foolhardy. Furthermore, research has shown that fast, athletic players
like Soriano tend to age better than average.

On the question of durability, there’s not much difference between
Soriano and Wilkerson, at least there wasn’t before the 2006
season. From 2001-2005 Soriano was very durable, averaging 154 games
per season. Wilkerson was pretty tough, as well, averaging 152 games
per season from 2002-2005.

The Last Word

Just to recap—it seems to me that the rush to condemn Jim
Bowden for the Soriano-Wilkerson trade a year ago may have been a bit
hasty and overly hostile. I think a careful analysis at the time would
have shown Soriano and Wilkerson to be players of comparable ability.
I’m speculating that a superficial treatment of park effects and
perhaps an ingrained prejudice against players with below-average
plate discipline led many commentators astray when evaluating this
trade.

Or maybe not. Maybe Wilkerson really was and still is a better player than
Soriano. Perhaps 2006 was just an aberration—a career-year for
Soriano and just an injury-marred bad-luck campaign for
Wilkerson. It may be that another year or two of data will show Wilkerson’s
superiority. Perhaps, but I don’t think so. I think we just blew this
one.

References & Resources
Despite writing the the first person plural, I hope it is clear that I’m only speaking for myself
in this piece. I’m sure many who weighed in on the
Soriano-Wilkerson trade last year will not agree what you’ve read
here.


Comments are closed.