All Downloads are FREE. Search and download functionalities are using the official Maven repository.

data.3news-bydate.test.rec.sport.baseball.104789 Maven / Gradle / Ivy

There is a newer version: 0.6.3
Show newest version
From: [email protected] (Harold_Brooks)
Subject: RBI, RISP, and SLG
Organization: Bill's Safety Cab and Record Bar, Chickasha, OK
Lines: 49

Off and on over the last several months, threads about RBIs and
related topics have gotten me to thinking about how well we can 
predict a player's RBIs using information about his overall
performance and the number of runners in scoring position (RISP)
that he bats with.  In the Brock2 model, Bill James calculated
predicted RBIs as RBI=.235*(Total Bases) + Home Runs.  This 
completely ignores the context, which was all that Brock2 
could do, since context was unknown to it.  So I thought I'd
take that idea as a starting point and look how good a fit to
the data you get by comparing (RBI-Home Runs) to SLG*RISP.

I've started with team data, using data from the Elias's that
I've picked up over the years when a) I could afford them and
b) I could stomach the thought of increasing Elias's profits.
That gave me the years 1984-1986, 1988, and 1990.  (I don't 
have team RBIs for '87 or I could add that year.)  If you
run a simple least squares fit to the data you get 

(RBI-Home Runs) = 0.81*SLG*RISP.

The correlation between the LHS and the RHS is 0.86, which is
significant at a ridiculously high level.  So, I feel like the
fit is good at the team level.  I've no started to move on to 
the player level and have looked at 4 players (Will Clark,
Ozzie Smith, Joe Carter, and Don Mattingly).  I hope to 
add quite a few more during my copious free time this year.

It doesn't do too badly, except the equation underpredicts the
low HR hitter (Smith), which may be a fault of the model or it
could just be Ozzie.  The results:

                           RBI-HR
         Years        Actual   Predicted
Carter  (84-88,90)     400       402.6
Clark   (87,88,90,92)  269       269.6
Matt'ly (84-88,90)     471       460.8
Smith   (84-88,90)     317       280.6

I think we can make a case (and I hope to make it stronger) that
RBIs can be predicted simply from knowing how a player slugs overall
and how many men are in scoring position when he comes up.

More later,
Harold
-- 
Harold Brooks                    [email protected]
National Severe Storms Laboratory (Norman, OK)
"I used to work for a brewery, too, but I didn't drink on the job."
-P. Bavasi on Dal Maxvill's view that Florida can win the NL East in '93




© 2015 - 2024 Weber Informatics LLC | Privacy Policy