PitchingPost

0th

Percentile

PitchingPost table

Post season pitching statistics

Keywords
datasets
Usage
data(PitchingPost)
Format

A data frame with 5798 observations on the following 30 variables.

playerID

Player ID code

yearID

Year

round

Level of playoffs

teamID

Team; a factor

lgID

League; a factor with levels AA AL NL

W

Wins

L

Losses

G

Games

GS

Games Started

CG

Complete Games

SHO

Shutouts

SV

Saves

IPouts

Outs Pitched (innings pitched x 3)

H

Hits

ER

Earned Runs

HR

Homeruns

BB

Walks

SO

Strikeouts

BAOpp

Opponents' batting average

ERA

Earned Run Average

IBB

Intentional Walks

WP

Wild Pitches

HBP

Batters Hit By Pitch

BK

Balks

BFP

Batters faced by Pitcher

GF

Games Finished

R

Runs Allowed

SH

Sacrifice Hits allowed

SF

Sacrifice Flies allowed

GIDP

Grounded into Double Plays

Aliases
  • PitchingPost
Examples
# NOT RUN {
library("dplyr")
library(ggplot2)

# Restrict data to World Series in modern era
ws <- PitchingPost %>%
         filter(yearID >= 1903 & round == "WS")
# Pitchers with ERA 0.00 in WS play (> 10 IP)
ws %>%
  filter(IPouts > 30 & ERA == 0.00) %>%
  arrange(desc(IPouts)) %>%
  select(playerID, yearID, teamID, lgID, IPouts, W, L, G, 
         CG, SHO, H, R, SO, BFP) 

# Pitchers with the most IP in a series 
# 1903 Series went eight games - for details, see
# https://en.wikipedia.org/wiki/1903_World_Series
ws %>%
  arrange(desc(IPouts)) %>%
  select(playerID, yearID, teamID, lgID, IPouts, W, L, G, 
         CG, SHO, H, SO, BFP, ERA) %>%
  head(., 10)

# Pitchers with highest strikeout rate in WS
# (minimum 20 IP)
ws %>%
  filter(IPouts >= 60) %>%
  mutate(K_rate = 27 * SO/IPouts) %>%
  arrange(desc(K_rate)) %>%
  select(playerID, yearID, teamID, lgID, IPouts, 
         H, SO, K_rate) %>%
  head(., 10)
  
# Pitchers with the most IP in WS history
ws %>%
  group_by(playerID) %>%
  summarise_at(vars(IPouts, H, ER, CG, BB, SO, W, L), 
               sum, na.rm = TRUE) %>%
  mutate(ERA = round(27 * ER/IPouts, 2),
         Kper9 = round(27 * SO/IPouts, 3),
         WHIP = round(3 * (H + BB)/IPouts, 3)) %>%
  arrange(desc(IPouts)) %>%
  select(-H, -ER) %>%
  head(., 10)

# Plot of K/9 by year
ws %>%
  group_by(yearID) %>%
  summarise(Kper9 = 27 * sum(SO)/sum(IPouts)) %>%
  ggplot(., aes(x = yearID, y = Kper9)) +
    geom_point() +
    geom_smooth() +
    labs(x = "Year", y = "K per 9 innings")

# }
Documentation reproduced from package Lahman, version 8.0-0, License: GPL

Community examples

Looks like there are no examples yet.