Waging WAR: MLB Player Valuation Using Advanced Metrics

Waging WAR:
MLB Player Valuation Using Advanced Metrics
By Alexander Knorr, Sebastian Kadamany and Nick Vedder
Who we are and what motivates us
 UCD Economic Master’s Students
 Curious about how goods are valued and priced
(Microeconomics)
 Using the idea of market value to identify disparity
between current prices and competitive pricing
Value in the setting of MLB players
 Identifying players who most outperform their current
salaries
 We estimate what players should be earning to identify
price disparity
 We calculate a straightforward index to identify who
outperforms their contract the most in (relative terms)
What makes a good Baseball Player?
Who is the best?
 This year’s Cy Young Award is three headed monster:
 Zack Greinke; The lowest ERA
 Clayton Kershaw; The most strikeouts.
 Jake Arrieta; The most wins, two shut-outs and a no hitter.
 All arguably could be deserving of the award
depending on which metric is evaluated.
 Adding FIB and WHIP Metrics makes the decision even
harder.
Wins Above Replacement (WAR)
 WAR isolates the individual effect a player has on
winning, holding team, season and even stadium factors
constant
 Identifies relative performance to an “average”
replacement player
 Superior to standard metrics such as RBI
 At its core WAR is about runs scored
 Why reinvent the wheel?
Data and WAR
 Used Baseball Reference for all variables: WAR, age and
salary
 Unfortunately was not available for Minor Leagues
 Used recent career statistics (2010 – 2014) which has 145,109
player – season observations
 Collapsed and summarized Data
 Salary as of 2014
 Median WAR statistic
 Age as of 2014
Model Specification (1)
 Stratified players by position (Fielder or Pitcher)
 Transformed Salary (dependent variable) using logarithms to
normalize and provide relative interpretations
 Removed players who were earning exactly the 2010 rookie
minimum to reduce the skew of salary
Log Transform
for Fielders
Model Specification (2)
 Used Linear Regression for estimating “true” salary
 WAR will drive true value of a player
 Experience terms will control for player experience as that is not
explicitly accounted for in WAR
 Age start will attempt to temper the power of experience in
predicting salary. Controlling for players “past their prime”
Aside: Controlling for High Correlation
Between Experience and Salary
 By using the squared experience term, we allowed for
players to lose predicted salary as they “passed their
prime”
Results of Salary Estimation
Fielders
Pitchers
Ranking
 Simply calculated the ratio of predicted player salary to
actual salary as of 2014
 Creates the Salary Arbitrage Index (SAI)
 Rank by SAI
Conclusions
 Ranking the top 100 players by traditional statistics or
even WAR does not identify opportunity
 The very best players are typically already paid
appropriately
 Management always wants to find efficiency
 “Biggest bang for the buck”
 SAI was directly developed to identify opportunity in player
prospecting