Page 268 - Contributed Paper Session (CPS) - Volume 7
P. 268
CPS2094 Yoshimitsu Morinishi et al.
recent years, we construct causal model on whether we can adapt at NPB and
examine whether the "flyball revolution" at NPB is correct or not. I will also
describe the usefulness of causality reasoning in sports and the usefulness of
causal reasoning in observational research in future society.
3. Method
Study the literature of past causal inference and establish a method to
extract causal effects from data. For details about causality inference, refer to
Rubin (1974, 1976), Rosebaum and Rubin (1983), Rosenbaum (2002), Iwasaki
(2015), Hoshino(2009). Formulate and verify a causal model for baseball 'flyball
revolution at NPB'. Construct a causal model and adopt SEM (structural
equation model) as a measure method of causal effect. From the result of the
constructed causal model, we evaluate and verify flyball revolution at NPB and
verify the effectiveness of causal reasoning in sports. Specifically, analysis was
carried out according to the following procedure.
1. Basic calculation of results for each batting: NPB Aggregate "one bat
data" in 2016 and 2017 and calculate the batting result (out or not) or
batting strategy (whether frying), Batter player ID, pitcher player ID, and
so on.
2. Basic calculations of pitcher · batter's grades: results of pitchers necessary
for constructing a causal model (four-ball rate, ball speed etc.) Basic
calculate the results of the batter (batting rate, base rate, base stolen,
etc.) and combine it with the data of 1.
3. Replacement of swing speed with striking number and homerun count:
In this research, the objective is to verify whether "flyball revolution" is
effective for each player's swing speed. However, the data of the two
years of NBL used this time did not include the batter's swing speed.
Therefore, it was necessary to find a substitute variable to replace the
swing speed. In the interview active baseball club members, knowledge
that the number of strikes and the number of homeruns may have a
positive correlation with the swing speed and a causal effect was
obtained, so in this study, Principal component analysis was performed
on two variables, number and homerun count, and we decided to use a
factor common to the two variables as a swing speed substitute variable.
Principal component analysis was carried out as described above and
the main component score of the obtained factor called swing speed
was combined with the data set of 2.
4. Data set creation for 2 left and right batting groups and 10 swing speed
groups, totalling 20 groups: Data sets created in 3. Are divided into right
and left batting seats, then divided into 10 groups with slow swing
speed, fast respectively. Create a total of 20 data datasets.
255 | I S I W S C 2 0 1 9