Wednesday, August 06, 2008

How to Fix a Bullpen, Part 2

I'm going to change gears a bit in this series. Before, I was going to focus on improved bullpens. Upon a deeper look at those pens, I found that most of those teams just had God-awful pens the year before and simply attained a bulk of their improvement through simple addition by subtraction. Instead, we will focus on successful bullpens since 2004. Here are the top 20, by Fair Run Average:

Team FRA Year
CLE 3.10 2005
LAN 3.23 2008
MIN 3.26 2006
TBA 3.29 2004
TOR 3.38 2008
LAN 3.47 2004
OAK 3.48 2008
BOS 3.49 2007
TBA 3.50 2008
PHI 3.60 2008
MIN 3.62 2005
SDN 3.63 2007
CHA 3.63 2005
NYN 3.64 2006
SLN 3.65 2005
OAK 3.81 2006
ANA 3.82 2005
SDN 3.86 2006
NYA 3.87 2008
CHA 3.88 2008

So those are the bullpens we have at our disposal. The goal here will be to find common threads between those pens that have enabled them to be successful. In order to do this, we will have to use objective measures to determine how these bullpens were built. Bullpens, as a general rule, deal with a lot of flux over the course of the year. Injuries, ineffectiveness, and trades tend to create shakeup. For the purpose of this project, I am going to gather all data (at least initially) for the opening day bullpen. The opening day bullpen will include all players who make the opening day roster and start out in relief, not including a bumped fifth starter. I will also include players who start on the disabled list who were at least strong candidates for the bullpen. Now, let's look specifically at some data that can be taken.

Number of Active Players: Typically 6 or 7 and occasionally 5 or 8. Maybe there is some correlation between number of players in the pen and success they have.
Salary: Due to variation in the number of players, this will be the total base salary of all players divided by the amount of active plus disabled players counted. This will put to the test whether good relievers truly cost money.
Number of Pitchers who can be sent to the minors: This will be all players under five years of service with at least one option year left. Again, this will be divided by the number of active and inactive players. I suspect that roster flexibility enables more pitchers to get a shot, and therefore enables more pitchers to stick as a successful reliever. In fact, I would go so far as to say that this has been the biggest problem with the Tigers' bullpen construction the last two seasons.
Statistics (ERA, K/9, BB/9, HR/9): This will be an average of everybody's numbers from the year before. For players in the minors, I will use a simple adjustment that raises the pitcher's ERA, HR/9 or BB/9 by 1 per level down they were in the Majors. I will penalize their K/9 by 1 per level as well. These will be pooled together and a total score will be calculated. In other words, if one pitcher throws one scoreless inning and another puts up a 4.50 ERA in 50 innings, the combined ERA will be 4.41 and not 2.25.
%Innings pitched by the starters: The effect of the starting pitching needs to be measured. The easiest way to measure this will be to show the percentage of the team's innings that the starters were able to pitch. More innings from the bullpen should mean that it is stretched thin.
Defense: Because of availability of statistics, I will have to settle for defensive efficiency instead of zone rating. I strongly suspect that an improved defense will help the bullpen.
Health/Effectiveness of Intended Relievers: This will be calculated by the actual innings pitched by the opening day and opening day disabled relievers.
Transaction Data: How many were signed as MLB free agents? How many Minor League free agents? How many acquired in trades? What was the cost of those trades? How many waiver claims? How many Rule 5 picks?
Deployment: This will use less simple math than the rest. What I will do is calculate the correlation between the pitcher's leverage score and Fair Run Average. This will be for every pitcher who pitched over 20 innings in relief.

What I plan to do is take each of these scores for each of the top bullpens. We'll start tomorrow with the 2005 Indians, and try to work through these bullpens and maybe some of the less successful ones.

No comments: