by Keith Glab, BaseballEvolution.com
June 10, 2007
You'll often see Sabermetric player rankings based on how players perform
versus how a player's likely replacement would perform. Baseball
Prospectus in particular uses Value over Replacement Player (VORP), Wins Above
Replacement Player (WARP), and Batting Runs Above Replacement (BRAR). The
Hardball Times modified Bill James' Win Shares into Win Shares Above Bench (WSAB).
All of these metrics run under the premise that if a star player is injured,
traded, or lost to free agency, that his replacement would not post league
average numbers, as often this replacement player is simply a callup from
But a replacement level player can mean several different things. Take
Phillies first basemen as a perfect example. In 2005, Ryan Howard replaced
the injured Jim Thome by slugging .567 over 88 games. The previous year,
Howard hit 48 combined homers over three levels of play. And you all know
what Howard did last year. So in 2004, when BP calculated Thome as having
7.1 Wins Above Replacement Player, shouldn't it be closer to 0? Although
Thome banged out 42 homers and slugged nearly .600 that year, it is reasonable
to assume that the 24-year old Howard would have put up nearly comparable
In a cycle of George Lucas proportions, Ryan Howard went down with an injury
this year. But this time there was no Young Apprentice to cover for Master
Howard. He had been "replaced" by a platoon of Wes Helms and Greg Dobbs.
Worse yet, when Wes Helms made a start at first base, it usually meant that
Abraham Nunez was manning the hot corner. What kind of a Value Over
Replacement Player does Ryan Howard have when compared to the .600 OPS of Wes
Helms or the .650 OPS of Abraham Nunez? More than a Jim Thome has when a
Ryan Howard is waiting in the wings, that is for sure, even though the raw
numbers of Thome in 2004 are comparable to those of Ryan Howard last year.
Payroll can also impact the way we think about the replacement level of
players. For a team like the Yankees, Roger Clemens is essentially a
replacement for various injuries to their starting rotation. For the
Nationals, Jason Bergmann's trip to the DL means starts for the likes of Levale
Speigner. Thinking about replacement-level players in this fashion, you
could easily argue that Matt Chico has a Value Over Replacement Player equal to
Also, when rating players, most analysts would agree that players should be
awarded extra consideration for playing in a highly competitive league.
Certainly, few would penalize a player for it. But if you think about
replacement-level players in these terms, that's exactly what happens. A
fill-in player today is generally someone with years of professional experience
and prime physical conditioning. The final man on your roster in 1910 was
most likely either an illiterate 18-year old or a pot-bellied old man.
Were star players from 1910 really more valuable than
Of course, there isn't a replacement-level metric out there that takes these
sorts of factors into account. They all assume a replacement level player
to be a certain percentage worse than a league average player, and apply that
arbitrary weight to every player in a given league.
As a result, you have very misleading statistics based on made up figures.
We know that Matt Chico really isn't as valuable as Andy Pettitte is, but we
also know that the replacement level for Yankees players is generally quite
high. Either way you look at it - replacement level as concept or as pure
math - you're not going to get the whole story. It is generally much
better, in my opinion, to compare a player to the league average when ranking
There are one or two exceptions, however. It's no secret that I favor
Batting Runs, which are essentially a player's
Linear Weights value over a
league average player, when evaluating players from different eras. The
biggest problem in my reliance on Batting Runs occurs in my evaluations of
catchers and shortstops. Players who man these positions often have
negative values for batting runs while still contributing, since Batting Runs
compares a player to the league average rather than the league average at a
certain position. In other words, a shortstop may hit below the league's
average, but still be above average offensively for a shortstop.
The best example for how this can muddle things is the case of Joe tinker
versus Ozzie Smith. I rate the two shortstops about the same defensively
(an incendiary comment that will no doubt require a separate article soon, I
know). Offensively, Joe Tinker has -68 Batting Runs while the Wizard
amassed -143. A cursory appreciation of these statistics leads one to
interpret that Ozzie was twice as detrimental to his team with the stick than
Examined more closely, we realize that both of these players were actually
boons offensively. Tinker was arguably the best offensive shortstop of the
dead ball era not named Honus Wagner, and Ozzie Smith was probably in the upper
tier of offensive shortstops for his era, at least after his first six
disastrous seasons. The important thing to note is that Smith had a longer
career by nearly 30%, and therefore continued to rack up more negative Batting
Runs as he continued to contribute on offense. Where a player with a
longer career usually receives a boost to his Batting Runs totals, the Ozeroo
took a hit because of his longevity.
Baseball Prospectus' Batting Runs Above Replacement Player (which is actually
EQA and not Linear Weights) shows
Smith to have 363 BRAR while Tinker comes in at 249. This isn't a
definitive answer, but it's a lot closer than what we saw with Batting Runs. Add
Smith's 580 stolen bases swiped at a near-80% clip, and it becomes clearer why
the Wizard was in fact more valuable on offense than the Sad Lexicologist was.
Hopefully, this essay elucidated some of my reasoning for favoring Batting
Runs over Sabermetrical measures involving Value Over Replacement Player, while
also exposing some of the weaknesses of Batting Runs as well. Batting Runs
- not a perfect measure of a player's hitting ability, but almost certainly the
best that we've got.
Disagree with something? Got something to add? Wanna bring up something totally new? Keith resides in Chicago, Illinois and can be reached at firstname.lastname@example.org.