| |
Serious
Baseball: Introduction to PCA
by Matthew Souders and Frank Bundy III
January 25, 2005
This is the Serious Baseball column for AHP. My partner
(Matthew Souders) and I will be writing baseball articles with an
analytical flair, showing baseball trends, etc. using a database
generated from a system developed by Mr. Souders called PCA (Pythagorean
Comparative Analysis).
Before any articles are written, or databases referenced, it is
only fair and necessary to explain how PCA works. If we didn’t,
you would have no idea what we were talking about in any work that
will appear in this column.
Basically, PCA is a rating system for all aspects of any particular
player: offense, defense, and pitching. In each of those three categories
a player is measured against what a “marginal” (replacement-level)
player would do in exactly the same situation as the original
player. How much a player performed better than the “marginal”
player is the final factor in determining a player’s value.
The primary statistic used in PCA is BPO (Bases per Out). The reason
for this is because outs are a constant factor throughout every
facet of baseball. Getting more chances between outs, or taking
away chances between outs, is the goal of every offensive and defensive
player. The fact that outs are symmetrical allows the system to
be fair and balanced, and allows us to focus in on one statistic
of record that can be compared, and calculated easily.
The fact that PCA systematically eliminates elements of the game
that are out of the control of each player, including place, time
(park and era), luck, and the skills of their teammates, etc, means
that you will frequently see two players who appear to be similar,
based on traditional statistics (AVG/OBP/SLG, HR, RBI etc.), end
up scoring differently—either better or worse, in PCA.
Here is an example from the 2004 season.
Manny Ramirez (2004 Boston Red Sox):
568 AB, .308 AVG/.397 OBP/.613 SLG, 43 HR, 130 RBI, .967 Fielding
%.
(PCA: Marginal Offensive Value – 240, Marginal Defensive value
– 12, Total Marginal Value – 252)
Adam Dunn (2004 Cincinnati Reds):
568 AB, .266 AVG/.388 OBP/.569 SLG, 46 HR, 102 RBI, .970 Fielding
%.
(PCA: Marginal Offensive Value – 288, Marginal Defensive Value
– 5, Total Marginal Value – 293)
As you can clearly see, Ramirez has better “everyday”
statistics than Dunn, but has a smaller value according to PCA.
This is because a marginal player dropped into the same situation
as Ramirez in 2004 would be expected to do better than a marginal
player in Dunn’s situation; thereby lessening the gap between
the marginal player and Ramirez.
Basically, the difference between a marginal left fielder on the
Red Sox in 2004 and Manny Ramirez is smaller than the difference
between a marginal left fielder on the Reds in 2004 and Adam Dunn.
What you are looking at with these statistics is a value given to
each player based on how many more bases he created or saved than
a marginal player in the same number of outs. This concept applies
to all three aspects of baseball, offense (MOV-Marginal Offensive
Value), Defense (MDV-Marginal Defensive Value), and Pitching (MPV-Marginal
Pitching Value)—which are added together to come up with TMV
(Total Marginal Value).
This is just an introduction to PCA. A series of articles will be
published in the near future explaining PCA in more detail.
This article should, in the very least, let you know what you are
looking at when you see PCA statistics throughout our work.
Thank you for reading.
If you have any questions, comments, concerns, or suggestions, please
post them in our forums or contact us directly at frnkbndy@yahoo.com,
or my colleague Matthew Souders at m_souders@yahoo.com.
|
|
