Streaky measures of hitting from Statcast data.

statcast_streaky.Rmd

---
title: "Statcast Streaky Analysis"
author: "Jim Albert"
date: "10/5/2020"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE,
                      message = FALSE,
                      warning = FALSE)
```

#### Read in Data

Read in statcast data for 2020 season;

```{r}
library(tidyverse)
sc <- read_csv("~/Dropbox/2020 WORK/statcast2020/sc2020final.csv")
```

#### Main Functions

Here is the main function to compute the streaky statistics described in the post.

The inputs to ```streak_data3()``` are the statcast dataset and the player of interest.

The output is a list with four components:

- Swing - data frame with the swing and miss data
- InPlay - data frame with the BIP data
- pv - p-value of the permutation test
- r2 - value of the R^2 statistic

```{r}
streak_data3 <- function(sc, player){
  require(tidyverse)
  require(BayesTestStreak)
  require(lubridate)
  pdata <- filter(sc, player_name == player) 

  swing_situations <- c("bunt_foul_tip",           
  "foul", "foul_bunt", "foul_tip",                          
  "hit_into_play", "hit_into_play_no_out",   
  "hit_into_play_score", "missed_bunt",            
  "swinging_strike", "swinging_strike_blocked")

  pdata %>% 
    filter(description %in% swing_situations) %>% 
    mutate(Miss = ifelse(description %in% 
                    c("swinging_strike", 
              "swinging_strike_blocked"), 1, 0), 
        Contact = 1 - Miss, 
        Count = paste(balls, strikes, sep = "-")) ->
  pdata_swing  

  pdata_swing %>% 
  arrange(game_date,
          game_pk,
           at_bat_number,
           pitch_number) -> pdata_swing
  
   pdata_inplay <- filter(pdata_swing, type == "X")
   pdata_inplay$prob <- 
     as.numeric(pdata_inplay$estimated_ba_using_speedangle)
   
   pdata_inplay$Week <- 
       factor(week(ymd(pdata_inplay$game_date)))
   fit <- lm(prob ~ Week,
               data = pdata_inplay)
   r2 <- summary(fit)$r.squared

   df1 <- data.frame(Swing = 1:dim(pdata_swing)[1],
                     Miss = pdata_swing$Miss)
   df2 <- data.frame(Inplay = 1:dim(pdata_inplay)[1],
                     Week = pdata_inplay$Week,
                     Prob = pdata_inplay$prob)
   pv <- permutation.test(df1$Miss)
   
   list(Player = player,
        Swing = df1,
        InPlay = df2,
        pv = pv,
        r2 = r2)
}
```

The function ```make_plots()``` will create associated graphs from the output of the ```streak_data3()``` function.

```{r}
make_plots <- function(sd){
  require(BayesTestStreak)
  require(gridExtra)
  require(ProbBayes)
  p1 <- plot_streak_data(sd$Swing$Miss) +
    ggtitle(paste(sd$Player, 
                  "Misses on Swings: P-Value =",
                  sd$pv)) + 
    centertitle()
  
  p2 <- ggplot(sd$InPlay, 
               aes(Week, Prob)) +
    xlab("Week") +
    geom_boxplot() +
    ggtitle(paste(sd$Player, 
                  "P(H) on BIP: R2 =",
                  round(sd$r2, 3))) + 
    centertitle() 
  grid.arrange(p1, p2, ncol = 1)
}
```

#### Example

Here is an example of using the functions for Bryce Harper.

```{r}
streak_data3(sc, "Bryce Harper") %>% make_plots()
```

#### Many Players

Want to collect streaky stats for many players.  Write function that will collect the p-value and R squared statistics for a given player

```{r}
collect_pvs <- function(player){
  out <- streak_data3(sc, player)
  c(out$pv, out$r2)
}
```

Collect number of pitches for all players.

```{r}
sc %>% 
  group_by(player_name) %>% 
  count() -> S
```


We focus on hitters who faced at least 600 pitches during the 2020 season.

```{r}
player_list <- filter(S, n >= 600) %>% 
  pull(player_name)
```

Collect stats for many hitters by use of the ```sapply()``` function.


```{r}
out <- sapply(player_list, collect_pvs)

out_df <- data.frame(Player = player_list,
                     Swing_PV = out[1, ],
                     XBA_R2 = out[2, ])
row.names(out_df) <- NULL
```

Construct some graphs of these streaky statistics:

```{r}
ggplot(out_df, aes(Swing_PV, XBA_R2)) +
  geom_point()
ggplot(out_df, aes(Swing_PV)) +
  geom_density()
ggplot(out_df, aes(XBA_R2)) +
  geom_density()
```

Look at several interesting streaky hitters with respect to R2.

Output the players whose R^2 value exceeeds 0.2.

```{r}
filter(out_df, XBA_R2 > .2)
```

Here are the plots for these streaky players.

```{r}
streak_data3(sc, "Luis Robert") %>% 
  make_plots()

streak_data3(sc, "Daniel Vogelbach") %>% 
  make_plots()

streak_data3(sc, "Gleyber Torres") %>% 
  make_plots()
```