[ Home ] [ Library ] [ Bookstore ] [ Contact ] [ Search ]
VORP: Value Over a Replacement at a Position
Fellow bosox-list-member Nate Calvin and I have made some progress on one of my long-standing projects -- comparing players' careers based on position-adjusted replacement-level production, rather than league-average production. The main reason for this is that comparing to league average gives no real value to an average player who plays for a long time, whereas an average player is actually quite useful. Comparing to a lower rate of production (replacement-level) is a better way of recognizing the value of playing time and durability.
We've captured most of the information from the Total Baseball CD (through 1993) into a relational database, which has allowed us to investigate some interesting questions. 1994 and 1995 stats are not included in this analysis.
We're still refining the method, and there may be a few bugs in parts of the code we're using, but I'm pretty confident that we're accurate for the vast majority of players.
Notes for hitters:
* Players are compared to league positional averages for each year.
* Replacement level is set to 35 points below the positional average in each of AVG, OBP, and SLG. This was based on an analysis I did of the 1994 AL and NL that showed that the positional averages, once you removed the 14 players with the most plate appearances per position in each league, were about 70 points below the OPS the overall positional average. I don't yet know if this is accurate across years, but its the best justification for a setting a replacement level that I've seen yet.
* Each player is considered to play only the position they appeared in most during each individual season (first position listed in Total Baseball)
* Outfielders are all grouped into one position (not split into LF, CF, RF). The only reason for this is that Total Baseball does not break out OF positions into separate categories, so its impossible to accurately classify each player as a LF, CF, or RF.
* Defense is not included in the rankings, nor are stolen bases, though I have an improvement to this method, called VORP (Value Over a Replacement at a Position), that includes both factors.
* Stats are park-adjusted and league-adjusted.
* Pitcher batting is not included (see the accompanying pitchers' analysis).
* Calculations of runs above average were done by RC/Out. At some point, I want to go back and do it using position-adjusted MLV, but RC/Out was easier to code, and is a reasonably good approximation to MLV for most players.
Marginal RC = PA * (OBPp*SLGp/(1-AVGp) - OBPl*SLGl/(1-AVGl))
AVGp, OBPp, SLGp = the player's batting average, OBP and SLG
AVGl, OBPl, SLGl = the league average (or replacement level average)
of all players who played primarily that position
during that year
* In career listings, the players position itself isn't shown because he may have played several different primary positions over his career (e.g. OF/1B). Each year is compared to his primary position for that year)
Notes for pitchers:
MRC -- the number of runs a pitcher prevented compared to a league average
pitcher in the same number of innings.
RMRC -- the number of runs a pitcher prevented compared to a replacement
level pitcher in the same number of innings.
Comparing to replacement level reflects the fact that a league average performer is actually rare enough to be valuable. Comparisons to league average gives an average player who plays for a long time no credit for that durability. Using replacement level (which is the theoretically, the level at which pitching of that talent level is easy and/or nearly costless to acquire), more properly rewards durability *and* production.
* Pitchers are compared to league averages for each year.
* Starters and relievers are considered together -- there is no adjustment for the lower number of innings (and higher leverage situations) generally pitched by closers.
* Replacement level is set to 1.00 runs worse than the league average ERA
* Defense is not included in the rankings, nor are pitcher's batting.
* Stats are park-adjusted.
* MRC is calculated as IP/9 * (LgERA - PkERA) RMRC is calculated as IP/9 * ((LgERA+RepLvl) - PkERA)
* Players are ranked in order of total runs above replacement level over the course of their careers:
Legend for pitchers:
NAME = The player's name
PLAYED = Years spanning the player's career (may include
some skipped years when he did not play)
YRS = total number of years spent in the majors
IP = Total innings pitched
ERA = ERA (unadjusted)
W = Wins
L = Losses
SV = Saves
POSMRC = Position-adjusted Marginal Runs Created
RPOSMRC = Replacement-position-adjusted Marginal Runs Created
VORP = Value Over a Replacement at a position, for pitchers,
this includes the value of his hitting.
VORPD = VORP with defense, and measured by Total Baseball's
Fielding Runs
Legend for hitters:
NAME = The player's name
PLAYED = The years spanning the player's career (though he
need not have played in every single intervening
year).
YRS = number of years spent in the majors
PA = plate appearances
AVG = batting average
OBP = on-base percentage
SLG = slugging average
MRC = Marginal Runs Created, compared to a league average hitter
POSMRC = Position-adjusted Marginal Runs Created
RPOSMRC = Replacement-position-adjusted Marginal Runs Created
VORP = Value Over a Replacement at a Position; the combination
of RPOSMRC, and the value of a player's stolen base
record (SBR = .3 * (SB - 2*CS))
VORPD = VORP with defense as measured by Total Baseball's
Fielding Runs
[ Home ] [ Library ] [ Bookstore ] [ Contact ] [ Search ]
Last Updated: Contact webmaster@stathead.com for corrections or problems
Copyright 1997-2001 by Keith Woolner. All included authors retain the copyrights to their original works.