Dial Player Index (DPI) Methodology

15 views
Skip to first unread message

Chris Dial

unread,
Oct 26, 1999, 3:00:00 AM10/26/99
to

DPI is a rating system I developed (am developing) for
position players to
incorporate Extrapolated Runs (see J. Furtado's
www.baseballstuff.com or www.backatcha.com ) and Defensive
Runs.

For each player, I calculated his XR, and adjusted every
player's XR to a
season's worth of outs. This turns out to be ~465 outs. I
then normalized
every batter's XR to XR/465, and adjusted for park factor
for runs.

After reading exchanges between Nelson Lu and Keith Woolner,
I settled
on 99 PF averaged with the STATS 3-yr PF of 96-98.

This is the XR a player would contribute if he were Cal
Ripken: played every
inning of every game.

A player's offensive value is XR+ (park-adjusted, normalized
to 465 outs).

Defense is a real bear to quantify. I used to score for
STATS, so I have a
really good grasp of what they do and how well it works (in
terms of
measuring balls in play). I think it is an excellent
scoring system. How
STATS happens to sort the data and assign it isn't the best,
but it is what
we have.

I want to thank Dale Stephenson for his suggestions and DR
raw data he's
posted from time to time.

I normalize the defense as well. To do this, I establish
how many plays
each position will get during the course of a season. And a
defensive
season is defined by the Cal Ripken season: every inning of
every game.
That is 162*8.75 = 1418. For each position, I determine the
average # of
plays per game each defender playing in 125+ games gets.
That average
multiplied by 162 is the number of plays each position is
normalized to.
That makes each position with a different number. Here
are the plays for the 1999 season a Ripken can expect:

Pos AL NL
1B: 277 275
2B: 545 480
3B: 430 430
SS: 541 495
LF: 386 358
CF: 505 519
RF: 382 378

AL: 3066
NL: 2935

Assumption:
For infielders, everything tends to even out. Since I
normalize the chances,
lefty-righty staffs and groundball/flyball staffs don't
affect the data too
much. It is possible that the players used to calc the
seasons chances play
for LHP/RHP or GB/FB, but each was averaged with at least
four starters.
Hard hit balls, tough chances etc. tend to even out. I'd
rather have the raw
data (plays in each zone for every player), but I takes what
I gots.

Nonetheless, the plays I do have are certainly outs. At
worst, ZR
undervalues defense (depending on pop-ups and line drives,
and DPs).

With the season total of plays, I multiply by the fielder's
ZR. That's
how many plays he would have made. I think this converts to
runs easily,
and almost obviously: any play not made results in a
baserunner.

Each out has a value: if it takes away a single, that's
worth 0.47 (hit) +
0.28 (out). Each subsequent base is worth 0.31 runs (per
Linear Wts).

SS/2B: (0.987*plays made*0.75) + (0.013*pm*1.06)
1B: (0.86*pm*0.75)+(0.125*pm*1.06)+(0.015*pm*1.37)
3B: (0.84*pm*0.75)+(0.158*pm*1.06)+(0.002*pm*1.37)
LF: [(0.76*pm*0.75)+(0.213*pm*1.06)+(0.026*pm*1.37)]*pitf +
Assists*0.59
CF:[(0.76*pm*0.75)+(0.183*pm*1.06)+(0.057*pm*1.37)]*pitf +
Assists*0.59
RF:[(0.76*pm*0.76)+(0.18*pm*1.06)+(0.06*pm*1.37)]*pitf +
Assists*0.59

pitching staff factor is calculated by: SLG+ = (total bases
from (fly
balls - hr))/ (fb - hr)

I get the total bases by: fly balls*%singles allowed by
team*0.48 (the value
of a single) + (2B, 3B)
then set as a percentage of league average (less that team's
SLG+) for the
pitf.


Assumption:
OFers that play behind bad pitchers face tougher chances.
This increases the value of plays made by OF who play behind
bad pitchers.

Problem: LF in Fenway cannot be measured with present
available data. Troy O'Leary's defense cannot be measured
without the removal of balls off the wall hits. His
defensive data has been removed from AL LFs. This also
applies to Colorado. Dante Bichette's and Larry Walker's
defensive data cannot be analyzed equitably. Their
defensive data is not included in NL LF/RF.

I compare the players to one another, rather than
replacement value. I am
working on a replacement value, but for defense, replacement
value is
often *higher*, like Belliard or any of the assorted
defensive replacements.

The ratings are runs above average, so the two factors can
be added for a
total player index.

RSV/XRV/TRV
These ratings are new for this season. A critique of my
system pointed out that my system did not actually describe
what a player *was* worth, but rather what he could have
been worth. As such, I worked up how each player performed
compared to what an average player would have done with the
same number of chances (outs or fielding opps). This
yielded a Runs Saved Value, an XR value and a Total Run
Value. This describes what each player actually did for
his team. All values are park adjusted/pitcher adjusted.
This stat makes it easy to select the MVP, as it shows which
player had the most value to his team. Players that miss
time are effectively "hurt" by their absence.


Any and all input is helpful, and appreciated (regardless of
my response tone)

Chris Dial


--
* Keith Woolner, Moderator for rec.sport.baseball.analysis *
* Submissions: rs...@stathead.com *
* Questions/Info/Contact: rsba-r...@stathead.com *
* Charter: http://www.stathead.com/rsba-charter.htm *

Reply all
Reply to author
Forward
0 new messages