Poor Man’s Super Linear Weights (PM-SLWTS)

Introduction

A few years back Mitchel Lichtman one of the co-authors of “The Book” first published a metric of his design called Super Linear Weights (SLTWs).  The metric used play by play data purchased from Stats Inc. to calculate the value of a player’s batting, defensive, and base running contributions to a team.  Since then, Mitchel went to work for a ball club and was no longer able to publish SLWTS, which left us without a really good measure of a player’s contribution.  The stats that are available are flawed and not all encompassing.  Win Shares designed by Bill James and available at the Hardball Times website uses a useless defensive system, and no base running system.  VORP and WARP published by Baseball Prospectus are each flawed in different ways – VORP uses a ridiculous fictional baseline, has no base running system and only uses a positional adjustment for defense.  WARP, also uses a ridiculous fictional baseline, has no base running system and uses a useless defensive system. 

Luckily for us, with the advent of the Bill James Baseball Handbook, and various other sources we have the components available to us to calculate a Poor Man’s Super Linear Weights (PM-SLWTs).  I have done very little heavy lifting here.  Most of the real work was done, by Bill James, Lee Sinins, Chone Smith, and the good people at Baseball Info Solutions – I neither deserve nor want any credit for their work – they are the real sabrmetricians at work on this project, I’m just a book keeper. 

Sources

Batting:  RCAA – The Complete Baseball Encyclopedia published by Lee Sinins and available here: http://www.baseball-encyclopedia.com/

Fielding: Zone Projections – Published by Chone Smith and can found here: http://lanaheimangelfan.blogspot.com/

Base Running: Baseball Info Solutions – The Bill James Handbook 2007 and can be purchased at BN or Amazon

Position Adjustments: The data used can from the Complete Baseball Encyclopedia using 2000-2006 data and was calculated by me – the factors are based on 400 outs are:

1b:  -14
2b:  +6
SS:  +9
3b:  +2
CF:  +1
LF:  -9
RF:  -8
DH:  -20 (no calculation, intuitive based on other adjustments)

Notes
Before we get into the data, a few notes on some of the data.  There are a number of defensive systems available – I chose Chone Smiths for a couple of reasons.  1) It is zone based.  At the end of the day any system that is based on primary defensive data will be virtually useless as it is nearly impossible to estimate player’s opportunities from that data. Imagine trying to calculate batting average without, having a reasonable idea if a player had 600 At-Bats or 475 At-Bats.  While Zone data isn’t nearly perfect, and has a considerable amount of noise it is still the best that we have. 2) It incorporates outfield park factors, which play a large part in determining specific player values.  A major flaw in zone data is that it measures the distance of flyballs from home plate rather from outfield wall, so you can see players get charged with opportunities when balls are not actually playable (think Green Monster) this can create large distortions from reality.  3) It is regressed – while the idea of this project is to measure how many runs a player contributed relative to an average player during last season (2006), I believe that there is so much noise surrounding defensive statistics it is appropriate to be as conservative as possible so that you get the best estimate of how many runs a player contributed last year.  While I have no doubt that I using a regressed projection rather than last years specific stat will mean that we get less players exactly right, I believe your sample as a whole get more players approximately right.  Also keep in mind that catchers are not getting any defensive credit, beyond a position adjustment, that isn’t the correct methodology and should be enhanced at some point in the future.  Zone rating isn’t applicable to them, however, at minimum a caught stealing factor should be included, I have some ideas but haven’t put enough thought into it yet.  Also outfielders don’t have an arm rating either.  Again that should be enhanced at some point in the future. On to Base Running, BIS provides data on bases gained.  I converted bases to runs assuming each base was worth 0.30 runs, which is based off traditional linear weight values.

Purpose and Uses
The purpose of this project isn’t to reinvent the wheel; it is just to get a catch all tools to evaluate how players performed last year.  It isn’t meant to be used as a forecasting device as it only uses 1 year worth data rather than multiple year’s worth of data which is needed to make reasonably accurate projections.  I don’t believe it should be used as a device to determine MVP, as this is somewhat context neutral while MVP discussions should include the context of ones production.  I believe that these rankings give us a reasonably clearer idea of who was a better player least season.  For example, let’s compare two players – Josh Willingham and Mark Ellis. 

At Bats OPS+ VORP

Josh 502 121 28
Mark 441 85 7

It would be very tough from that line to determine that Ellis was a better player, yet when you look at the players PM-SWLTs it’s tough to ignore that conclusion as Ellis is better by more than 25 runs.  While I realize that these numbers are estimates, 25 runs is a very hard number to ignore. 

Legend
POS:  The players primary position as defined by the Complete Base Ball Encylopedia
RCAA: Runs Created Above Average (Batting Runs)
POS – ADJ:  Position Adjustment
DEF: Defensive Runs
BASE RUN: Base Running Runs
PM-SLWTs: Poor Man’s Super Linear Weights
PM-SLWTS/400: Poor Mans Super Linear Weights Per 400 Outs

DATA can be found HERE

I tried to post the data here but I couldn’t get the formating right on the columns.  Sorry for having to link it.