Welcome to my 2022 Projections for Pitchers 5x5

Projections using Hypertuned model through XGboost

All data is from FanGraphs. I have no affiliation with FanGraphs, but please consider contributing to their website if you found this project informative.

1 Project Scope

1.1 Objective

This project is designed to showcase how Using a Percentile Based Worth System values Fantasy Baseball Players through a Inning Pitched (IP) weighted projection

The Categories used for prediction valuation are year-end rankings for the following metrics:

• Wins
• Saves
• Strike Outs
• ERA ( 9 * Earned Runs per Inning Pitched)
• WHIP (Walks and Hits per Inning Pitched)

---

---

2 Processing the Data

2.1 Getting Data Into R

2.1.1 Load Libraries

First we need to load the packages that R needs to run the analysis

library(sqldf) #SQL in R
library(skimr) #Summaries and useful for removing low % data
library(ggplot2) #Plotting Functions
library(plyr) #slightly deprecated data cleaning
library(dplyr) #slightly updated data cleaning
library(tidyverse) #tidyverse data cleaning universe
library(caret) #wrapper for creating, tuning and validating models
library(xgboost) #package for creating regression tree model
library(vtreat) # useful package for treating data before modeling 
library(Matrix) #creating matricies for xgboost
library(mgcv)
library(moments) #for measuring skewness
library(data.table) #alternative to dplyr we use to create lags
library(pdp) #partial dependence graphs
library(vip) #variable importance 
library(grid) #put multiple plots on one grid
library(gridExtra) #additional grid functionality
library(janitor) #one function used to clean transposed data set
library(ggpubr) #for qq plot 
library(owmr) #Removing Prefixes
library(kableExtra) # formatting HTML Tables
library(formattable) # formatting HTML Tables

The # comments generally explain what additional functionality each library adds to R

2.1.2 Load in Data

All data is downloaded from Fan Graphs from this location. The data is also available on my Github here. There are player level and team data sets

#data read-in
pitcher_data <- read_csv("FanGraphs Leaderboard_Pitching20IP.csv")

#Team datasets
FDG_Team = read_csv("FanGraphs Leaderboard_Team.csv")

#Create a prefix for all team stats that starts with T_
FDG_Team2 <- FDG_Team %>% 
  rename_with( ~ paste0("T_", .x))

2.1.3 Checking Team Data

str give information about an object, while skim provides a customizable summary

#Output not shown for space
#str(FDG_Team2)

skim(FDG_Team2) %>%  
  tibble::as_tibble() #Remove this option for a normal HTML table

---

2.2 Understanding the Dataset

2.2.1 Exploring the dataset

skim let’s us see how the data was imported into R. Documentation can be found here

#Full Dataset dimensions

skimr::skim(pitcher_data) %>% 
  tibble::as_tibble() %>%  #Remove this option for a normal HTML table
  select(skim_type,skim_variable,complete_rate) %>% 
  filter(complete_rate >0.30) #250 Variables

#skim_type - character or numeric
#skim_variable - name of variable
#complete_rate - % of data that is not missing
#filter - only keep variables that have 30% of data populated

---

Additionally let’s look at how variables vary by year to see if there are any discrepancies there

#It looks like one year, there were fewer games played, and there is a clear drop off in home runs
pitcher_data_dist =
pitcher_data %>% 
 group_by(Season) %>% 
  summarize (Max_Games = max(G),
             Avg_W= mean(W)
             )

pitcher_data_dist

#Plot Win Data by Year
ggplot(pitcher_data_dist, aes(Season, Avg_W)) +
  geom_col()+
  ggtitle("Average Wins by Year")+
  theme(plot.title = element_text(hjust = 0.5,size = 22,color ="steel blue"))

---

2.3 Cleaning and Creating Initial Dataset for Model

What are some issues with the data?

1. Many of Variables, such as K%, are being read in as characters

   • Only Team and Player Name should be characters

2. There is spotty data coverage in some of the variables (~Variables have less than 30% Coverage)

3. 2020 Data only includes 60 games worth of data

   • This was a season shortened due to Covid-19

4. Team Data needs to be appended to pitcher Data by Team Name

---

2.3.1 Cleanly Changing all Variables that are characters to numeric.

There are several ways to do this, we will identify the variables we want to change that are mis-identified. parse_number can be used to pull numbers from these variables. Additional ways to tackle this can be found here.

#Select Column names that are characters but not Team or Name, These should be percentages
pitcher_data_chars_to_convert <- pitcher_data %>% 
  select_if(is.character)%>% select(-Team,-Name) %>% 
  mutate_all (function(x) as.numeric(readr::parse_number(x))/100)
#Note : There are additional ways to do this, this is just one solution


#We can exclude the variables we converted and reintroduce them
pitcher_data_num <- pitcher_data %>% select(-colnames(pitcher_data_chars_to_convert))

pitcher_data2 = cbind(pitcher_data_num,pitcher_data_chars_to_convert) %>% 
  select (colnames(pitcher_data)) %>%  #preserve original order 
  dplyr::rename(flyball_perc = `FB%...50`,fastball_perc = `FB%...74`) #rename two ambiguous columns
  
skim(pitcher_data2) %>% 
  as_tibble() %>% 
  group_by(skim_type) %>% 
  count()

#Logical variables are R's best guess, in our case they are all NA's and will be removed at a later step

The same can be done for the Team Data that is loaded

#Select Column names that are characters but not Team or Name, These should be percentages
FDG_Team2_chars_to_convert <- FDG_Team2 %>% 
  select_if(is.character)%>% select(-T_Team) %>% 
  mutate_all (function(x) as.numeric(readr::parse_number(x))/100)
#Keep in mind, parse number may make actual characters into numerical variables so carefully check your data before using

#We can exclude the variables we converted and reintroduce them
FDG_Team2_num <- FDG_Team2 %>% select(-colnames(FDG_Team2_chars_to_convert))

FDG_Team3 = cbind(FDG_Team2_num,FDG_Team2_chars_to_convert) %>% 
  select (colnames(FDG_Team2)) %>%  #preserve original order
dplyr::rename(T_flyball_perc = `T_FB%...45`,T_fastball_perc = `T_FB%...72`)  #rename two ambiguous columns

skim(FDG_Team3) %>% 
  as_tibble() %>% 
  group_by(skim_type) %>% 
  count()

---

2.3.2 Filtering Data with Low Coverage

I choose 30% coverage of data necessary but this can be adjusted up or down. This will also get rid of columns that are all NA.

# Keep variables with enough values (Need 30% data coverage rate here)
Player_cols_to_keep =
skim(pitcher_data2) %>% 
  dplyr::select(skim_type, skim_variable, complete_rate) %>% 
  filter (complete_rate > 0.30)

#Transpose Rows to get column names as skim melts the data
Player_cols_to_keep_transpose = t(Player_cols_to_keep) 

#extract the colnames we would like to keep
Player_cols_to_keep = colnames(janitor::row_to_names(Player_cols_to_keep_transpose,row_number = 2))

#Only keep the columns designated to have over 30% of their data populated or greater
pitcher_data3 = pitcher_data2 %>% 
  select(one_of(Player_cols_to_keep)) 

Repeat the process for Team Variables

Team_cols_to_keep =
skim(FDG_Team3) %>% 
  dplyr::select(skim_type, skim_variable, complete_rate) %>% 
  filter (complete_rate > 0.30)


#Transpose Rows to get column names as skim melts the data
Team_cols_to_keep_transpose = t(Team_cols_to_keep) 

#extract the colnames we would like to keep
Team_cols_to_keep = colnames(janitor::row_to_names(Team_cols_to_keep_transpose,row_number = 2))

#Only keep the columns designated to have over 30% of their data populated or greater
FDG_Team4 = FDG_Team3 %>% 
  select(one_of(Team_cols_to_keep)) 

---

2.3.3 Creating Variables Normalized by Year

Some Variables will need to be normalized by Innings_Pitched (IP) if they aren’t a percentage already. Remaining Variables are percentages or indices so will not need to be transformed. The full data dictionary for these variables can be found on FanGraph’s website here. for pitching variables and here. for hitting variables.

pitcher_data4 = pitcher_data3 %>% 
  mutate( #create new variables based on existing variables
    W_IP = W/IP,
    L_IP =  L/IP, 
    ShO_IP = ShO/IP,
    SV_IP = SV/IP,
    BS_IP = BS/IP,
    TBF_IP = TBF/IP,
    H_IP = H/IP,
    R_IP = R/IP,
    ER_IP = ER/IP,
    HR_IP=HR/IP,
    BB_IP=BB/IP,
    IBB_IP=IBB/IP,
    HBP_IP=HBP/IP,
    WP_IP= WP/IP,
    BK_IP=BK/IP,
    SO_IP=SO/IP,
    GB_IP = GB/IP,   #Groundballs
    FB_IP =  FB/IP,  #FlyBalls
    LD_IP = LD/IP,   #LineDrives
    IFFB_IP = IFFB/IP,  #Infield Fly balls
    Balls_IP= Balls/IP,
    Strikes_IP= Strikes/IP,
    Pitches_IP= Pitches/IP,
    RS_IP= RS/IP,
    IFH_IP= IFH/IP,
    BU_IP= BU/IP,
    BUH_IP= BUH/IP,
    Pulls_IP= Pulls/IP,
    HLD_IP= HLD/IP,   
    SD_IP= SD/IP,    
    MD_IP= MD/IP,    
    Barrels_IP= Barrels/IP,
    HardHits_IP= HardHit/IP
  ) %>% select(-L,-G,-IP,-ShO,-BS,-(TBF:BK),-(GB:BUH),-Pulls,-(SD:MD),-Barrels,-HardHit,-Events)
               
#will be removed after data is lagged -FIP,-(RAR:WPA),,-(wFB:wCH),-(`ERA-`:`xFIP-`),-SIERA,-(`RA9-WAR`:`Age Rng`),-kwERA,-`wCH (pi)`:`wSL (pi)`,`K/9+`:`HR/FB%+`) 

#skim(pitcher_data4) %>% as_tibble()

Repeat the process for Team Variables

FDG_Team5 = FDG_Team4 %>% 
  mutate( #create new variables based on existing variables
    T_H_T_PA = T_H/T_PA,
    T_x1B_T_PA = T_1B/T_PA, #note: R can't have variables start with a number
    T_x2b_T_PA = T_2B/T_PA,
    T_x3b_T_PA = T_3B/T_PA,
    T_HR_T_PA = T_HR/T_PA,
    T_R_T_PA = T_R/T_PA,
    T_RBI_T_PA = T_RBI/T_PA,
    T_BB_T_PA = T_BB/T_PA,
    T_IBB_T_PA = T_IBB/T_PA,
    T_SO_T_PA=T_SO/T_PA,
    T_HBP_T_PA=T_HBP/T_PA,
    T_SF_T_PA=T_SF/T_PA,
    T_SH_T_PA=T_SH/T_PA,
    T_GDP_T_PA= T_GDP/T_PA,#ground into double play
    T_SB_T_PA=T_SB/T_PA,
    T_CS_T_PA=T_CS/T_PA,
    T_GB_T_PA = T_GB/T_PA,   #Groundballs
    T_FB_T_PA =  T_FB/T_PA,  #FlyBalls
    T_LD_T_PA = T_LD/T_PA,   #LineDrives
    T_IFFB_T_PA = T_IFFB/T_PA,  #Infield Fly balls
    T_Pitches_T_PA= T_Pitches/T_PA,
    T_Balls_T_PA= T_Balls/T_PA,
    T_Strikes_T_PA= T_Strikes/T_PA,
    T_IFH_T_PA= T_IFH/T_PA,
    T_BU_T_PA= T_BU/T_PA,
    T_BUH_T_PA= T_BUH/T_PA,
    T_PH_T_PA= T_PH/T_PA,
    T_Barrels_T_PA= T_Barrels/T_PA,
    T_HardHits_T_PA= T_HardHit/T_PA
  ) %>% select(-(T_H:T_CS),-(T_GB:T_BUH),-T_PH,-T_Barrels,-T_HardHit,-T_Events) #Drop the old variables


#skim(FDG_Team5) %>% as_tibble()

---

2.3.4 Creating Lagged Variables

There are several ways to lag a dataset BY GROUP.
* Dplyr way is here..
* The data.table (the method used below) is here.

#Note we will only be lagging the player level data, as the previous year's team performance shouldn't impact current performance


#Order the dataset by lag columns
pitcher_data5 =  arrange(pitcher_data4, playerid,Season) #playerid is the Fangraph id assigned to each player

# Convert dataframe to data.table format
DT_pitcher = data.table(pitcher_data5)

#designate columns to lag - which is all of them
cols1 = colnames(pitcher_data5)
anscols = paste("lag", cols1, sep="_")
DT_pitcher[, (anscols) := data.table::shift(.SD, 1, NA, "lag"),by ='playerid', .SDcols=cols1] #Create 1 period lags by year

pitcher_data6 = as.data.frame(DT_pitcher) %>% select(-lag_playerid, -lag_Team, -lag_Season, -lag_Age,-lag_Name)

ncol(pitcher_data5) #251 - no lags

[1] 251

ncol(pitcher_data6) #497 - lagged data ~ (251 * 2)-5

[1] 497

---

2.3.5 Merging Team and Player Data

We can use either the merge function or the SQL functionality provided by the sqldf package to join the lagged player level data to the Team level data

df_pitching_init = sqldf(
  "
  select a.*, b.*
  from pitcher_data6 a
  left join FDG_Team5 b
  on a.Team = b.T_Team and a.Season = b.T_Season
  
  "
)  %>% select(-T_Team,-T_Season,-T_Age,-T_G,-T_AB)# Unncessary Team Variables


nrow(df_pitching_init) - nrow(pitcher_data6) #check if any rows are duplicated

[1] 0

---

3 Creating Rankings for Players Based On Percentiles

We can use Percentile based ranking to get rankings for players from the 2021 season.

3.1 Worth of each stat

3.1.1 Calculating past performance

Each player goes from a 0% to 100% on each percentile stat that is used for creating a scoring opportunity. Data is not normalized by IP as certain stats such as Wins will be worth more when we do.

#Categories I include are:
#Wins, Saves, WHIP, ERA, SOs, Holds

df_pitching_init2 =  df_pitching_init %>%
#  arrange(player_id,year) %>% 
  group_by(Season) %>% 
  mutate(
    Wins_share = order(order(rank(W_IP,ties.method = 'average'),decreasing = FALSE))/n(),
     SO_share = order(order(rank(SO_IP,ties.method = 'average'),decreasing = FALSE))/n(),
     SV_share = order(order(rank(SV_IP,ties.method = 'average'),decreasing = FALSE))/n(),
     WHIP_share = order(order(rank(WHIP,ties.method = 'average'),decreasing = FALSE))/n(),
     ERA_share = order(order(rank(ERA,ties.method = 'average'),decreasing = FALSE))/n(),
    HLD_share = 0,
    Worth = Wins_share+SO_share+SV_share+WHIP_share+ERA_share+HLD_share
    ) %>% 
  ungroup() 

Chart of the Distribution of initial percentiles
As the chart below shows, the data is roughly normal.

skewness((df_pitching_init2$Worth))

[1] 0.086

ggplot2::qplot(df_pitching_init2$Worth, main="Total Pitching Worth Dataset") + geom_histogram(colour="black", fill="steelblue")

`stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
`stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

min(df_pitching_init2$Worth)

[1] 0.32

max(df_pitching_init2$Worth)

[1] 4.6

ggpubr::ggqqplot(df_pitching_init2$Worth)

shapiro.test(df_pitching_init2$Worth)

    Shapiro-Wilk normality test

data:  df_pitching_init2$Worth
W = 1, p-value = 0.002

---

3.2 2021 Player Rankings - Per IP performance

3.2.1 2021 Player Rankings - Top Worth Players with Holds

Total Rankings for the players (Using 5x5 Scoring) can be found here. While it looks like many of the top players have low worth scores, it is because we haven’t applied a modifier for IP yet. Wins are harder to come by relative to any other stat and require more innings pitched.

df_pitching_init2_raw =  df_pitching_init %>%
#  arrange(player_id,year) %>% 
  group_by(Season) %>% 
  mutate(
    Wins_share_raw = order(order(rank(W,ties.method = 'average'),decreasing = FALSE))/n(),
     SO_share_raw = order(order(rank(SO,ties.method = 'average'),decreasing = FALSE))/n(),
     SV_share_raw = order(order(rank(SV,ties.method = 'average'),decreasing = FALSE))/n(),
     WHIP_share = order(order(rank(WHIP,ties.method = 'average'),decreasing = FALSE))/n(),
     ERA_share = order(order(rank(ERA,ties.method = 'average'),decreasing = FALSE))/n(),
    HLD_share_raw = 0,
    Worth = Wins_share_raw+SO_share_raw+SV_share_raw+WHIP_share+ERA_share+HLD_share_raw
    ) %>% 
  ungroup() %>% 
select(-W,-SO,-SV,-WHIP,-ERA,-HLD)



options(digits=2)

df_pitching_init2021_raw =
df_pitching_init2_raw %>% 
  group_by(Name) %>% 
  filter(Season == 2021) %>% 
  arrange(desc(Worth)) %>% 
  select(Name,Wins_share_raw,SO_share_raw,SV_share_raw,WHIP_share,ERA_share,Worth)


df_pitching_init2021_raw %>%
  filter (Worth>3.5) %>% 
  kbl() %>% 
 kable_material(c("striped", "hover","condensed","responsive"),full_width = F,fixed_thead = T)

Name	Wins_share_raw	SO_share_raw	SV_share_raw	WHIP_share	ERA_share	Worth
Daniel Bard	0.79	0.69	0.97	0.87	0.75	4.0
Garrett Richards	0.79	0.84	0.86	0.87	0.68	4.0
Jesus Luzardo	0.77	0.78	0.55	0.89	0.92	3.9
Brady Singer	0.72	0.89	0.66	0.82	0.69	3.8
Brad Keller	0.86	0.85	0.31	0.92	0.77	3.7
Jose Alvarado	0.82	0.60	0.90	0.87	0.51	3.7
Mitch Keller	0.69	0.76	0.40	0.96	0.87	3.7
Justus Sheffield	0.81	0.56	0.37	0.98	0.95	3.7
Nick Pivetta	0.89	0.95	0.72	0.51	0.60	3.7
Rafael Montero	0.65	0.34	0.91	0.81	0.91	3.6
Josh Fleming	0.93	0.58	0.78	0.59	0.73	3.6
Alec Mills	0.74	0.73	0.71	0.69	0.73	3.6
Joe Jimenez	0.75	0.49	0.73	0.80	0.85	3.6
Paul Fry	0.58	0.52	0.83	0.80	0.87	3.6
Wil Crowe	0.61	0.82	0.53	0.84	0.78	3.6
Erick Fedde	0.82	0.87	0.38	0.70	0.78	3.5
Adam Ottavino	0.77	0.62	0.93	0.70	0.51	3.5
Bryan Garcia	0.50	0.22	0.85	0.98	0.98	3.5
Alex Reyes	0.92	0.77	0.98	0.58	0.25	3.5

---

3.3 2021 Player Rankings - Actual Performance

3.3.1 2021 Player Rankings - Top Worth Players with Holds

While it looks like many of the top players have low worth scores, it is because we haven’t applied a modifier for IP yet.

options(digits=2)

df_pitching_init2021 =
df_pitching_init2 %>% 
  group_by(Name) %>% 
  filter(Season == 2021) %>% 
  arrange(desc(Worth)) %>% 
  select(Name,Wins_share,SO_share,SV_share,WHIP_share,ERA_share,HLD_share,Worth)


df_pitching_init2021 %>%
  filter (Worth>2.9) %>% 
  kbl() %>% 
 kable_material(c("striped", "hover","condensed","responsive"),full_width = F,fixed_thead = T)

Name	Wins_share	SO_share	SV_share	WHIP_share	ERA_share	HLD_share	Worth
Daniel Bard	0.93	0.85	0.97	0.87	0.75	0	4.4
Joe Jimenez	0.98	0.86	0.76	0.80	0.85	0	4.2
Paul Fry	0.83	0.87	0.82	0.80	0.87	0	4.2
Rafael Dolis	0.63	0.84	0.89	0.95	0.81	0	4.1
Ben Bowden	0.83	0.82	0.57	0.97	0.92	0	4.1
Jose Alvarado	0.96	0.85	0.89	0.87	0.51	0	4.1
Tanner Rainey	0.22	0.93	0.89	0.93	0.97	0	3.9
Sean Newcomb	0.62	0.93	0.79	0.93	0.65	0	3.9
Ryan Hendrix	1.00	0.73	0.51	0.82	0.85	0	3.9
Adam Ottavino	0.95	0.76	0.94	0.70	0.51	0	3.9
Rafael Montero	0.91	0.26	0.91	0.81	0.91	0	3.8
Enyel De Los Santos	0.55	0.94	0.44	0.95	0.90	0	3.8
Bryan Garcia	0.76	0.21	0.85	0.98	0.98	0	3.8
Tanner Scott	0.87	0.90	0.40	0.84	0.74	0	3.8
Aroldis Chapman	0.93	1.00	0.99	0.52	0.27	0	3.7
Sean Poppen	0.39	0.79	0.83	0.96	0.74	0	3.7
Alex Reyes	0.98	0.90	0.98	0.58	0.25	0	3.7
James Karinchak	0.97	0.96	0.94	0.35	0.47	0	3.7
Bryan Abreu	0.83	0.52	0.78	0.73	0.82	0	3.7
Sean Reid-Foley	0.90	0.89	0.36	0.77	0.75	0	3.7
Jeurys Familia	0.99	0.84	0.72	0.66	0.45	0	3.7
Rex Brothers	0.55	0.96	0.73	0.68	0.75	0	3.7
Paul Campbell	0.76	0.49	0.63	0.86	0.91	0	3.6
Trevor Megill	0.35	0.90	0.40	0.98	0.99	0	3.6
Lucas Sims	0.92	0.99	0.92	0.20	0.56	0	3.6
Pete Fairbanks	0.71	0.92	0.90	0.69	0.35	0	3.6
Jesus Luzardo	0.63	0.57	0.55	0.89	0.92	0	3.6
Gregory Soto	0.89	0.83	0.96	0.59	0.28	0	3.6
David Hess	0.91	0.32	0.32	1.00	1.00	0	3.5
Amir Garrett	0.04	0.89	0.92	0.83	0.86	0	3.5
Phil Maton	0.87	0.88	0.42	0.72	0.65	0	3.5
Jacob Barnes	0.26	0.78	0.87	0.73	0.88	0	3.5
Matt Barnes	0.94	0.98	0.98	0.21	0.41	0	3.5
Ryan Helsley	0.97	0.52	0.75	0.67	0.61	0	3.5
Cesar Valdez	0.36	0.47	0.93	0.91	0.84	0	3.5
Trevor May	0.94	0.92	0.86	0.43	0.35	0	3.5
Jacob Webb	0.98	0.45	0.79	0.79	0.50	0	3.5
Brad Brach	0.24	0.69	0.81	0.87	0.89	0	3.5
Adam Morgan	0.80	0.72	0.88	0.56	0.53	0	3.5
Luis Oviedo	0.24	0.62	0.63	0.99	0.99	0	3.5
Glenn Otto	0.10	0.83	0.61	0.93	0.99	0	3.5
Sam Howard	0.66	0.92	0.33	0.73	0.80	0	3.4
Reid Detmers	0.44	0.39	0.67	0.97	0.97	0	3.4
Michael Feliz	0.02	0.70	0.85	0.91	0.96	0	3.4
Aaron Ashby	0.90	0.86	0.80	0.28	0.61	0	3.4
Griffin Canning	0.80	0.51	0.54	0.75	0.80	0	3.4
Brett de Geus	0.59	0.21	0.65	0.95	0.98	0	3.4
J.B. Wendelken	0.87	0.33	0.83	0.78	0.54	0	3.4
Matt Foster	0.49	0.55	0.77	0.70	0.85	0	3.4
Aaron Bummer	0.86	0.92	0.81	0.44	0.33	0	3.4
Tyler Chatwood	0.22	0.80	0.80	0.72	0.81	0	3.4
Kyle Finnegan	0.75	0.56	0.93	0.74	0.34	0	3.3
Sean Doolittle	0.60	0.66	0.74	0.72	0.60	0	3.3
Justus Sheffield	0.85	0.17	0.37	0.98	0.95	0	3.3
Anthony Kay	0.20	0.79	0.59	0.92	0.81	0	3.3
Chris Stratton	0.86	0.67	0.90	0.50	0.37	0	3.3
Josh Sborz	0.67	0.78	0.72	0.67	0.45	0	3.3
Matt Andriese	0.33	0.59	0.74	0.88	0.75	0	3.3
Jake Diekman	0.45	0.95	0.90	0.56	0.43	0	3.3
Justin Garza	0.71	0.56	0.53	0.84	0.65	0	3.3
Cionel Perez	0.34	0.59	0.51	0.93	0.90	0	3.3
Shane Greene	0.01	0.57	0.83	0.91	0.96	0	3.3
Edwin Uceta	0.10	0.86	0.59	0.80	0.93	0	3.3
Michael Rucker	0.08	0.63	0.81	0.80	0.95	0	3.3
J.D. Hammer	0.47	0.70	0.51	0.88	0.70	0	3.3
Jeffrey Springs	0.95	0.96	0.83	0.20	0.30	0	3.2
Jake Brentz	0.78	0.80	0.80	0.47	0.38	0	3.2
Hirokazu Sawamura	0.89	0.77	0.68	0.71	0.19	0	3.2
Spencer Howard	0.11	0.62	0.65	0.88	0.97	0	3.2
Heath Hembree	0.25	0.96	0.92	0.30	0.79	0	3.2
Devin Williams	0.99	0.99	0.86	0.31	0.09	0	3.2
Seth Lugo	0.84	0.82	0.75	0.49	0.32	0	3.2
Zac Lowther	0.24	0.56	0.57	0.91	0.93	0	3.2
Hansel Robles	0.36	0.70	0.95	0.62	0.58	0	3.2
Paul Sewald	0.99	0.99	0.93	0.10	0.19	0	3.2
Mychal Givens	0.79	0.62	0.92	0.61	0.27	0	3.2
Edwin Diaz	0.80	0.96	0.99	0.13	0.31	0	3.2
Max Kranick	0.50	0.23	0.63	0.94	0.89	0	3.2
Mike Mayers	0.66	0.82	0.77	0.49	0.43	0	3.2
Trevor Stephan	0.42	0.81	0.71	0.66	0.57	0	3.2
Carlos Estevez	0.44	0.48	0.94	0.76	0.56	0	3.2
Anthony Castro	0.32	0.91	0.82	0.44	0.66	0	3.2
Jackson Kowar	0.10	0.44	0.62	1.00	1.00	0	3.2
Brad Hand	0.88	0.41	0.97	0.45	0.44	0	3.1
Daniel Norris	0.26	0.54	0.72	0.76	0.87	0	3.1
Camilo Doval	1.00	0.94	0.90	0.13	0.17	0	3.1
Kyle Zimmer	0.74	0.26	0.81	0.65	0.68	0	3.1
Nick Mears	0.36	0.50	0.66	0.90	0.71	0	3.1
Conner Greene	0.30	0.58	0.35	0.95	0.95	0	3.1
Greg Holland	0.52	0.43	0.91	0.57	0.68	0	3.1
Daniel Lynch	0.58	0.20	0.62	0.90	0.81	0	3.1
Demarcus Evans	0.10	0.87	0.60	0.81	0.74	0	3.1
Edward Cabrera	0.10	0.64	0.62	0.90	0.83	0	3.1
Eduardo Rodriguez	0.82	0.79	0.19	0.64	0.66	0	3.1
Kyle Keller	0.21	0.67	0.47	0.83	0.92	0	3.1
Nick Pivetta	0.57	0.74	0.68	0.51	0.60	0	3.1
Joely Rodriguez	0.36	0.54	0.75	0.81	0.63	0	3.1
Joakim Soria	0.17	0.66	0.92	0.61	0.72	0	3.1
Austin Voth	0.70	0.57	0.29	0.74	0.76	0	3.1
Aaron Slegers	0.64	0.20	0.30	0.98	0.95	0	3.1
Andrew Heaney	0.62	0.77	0.30	0.53	0.84	0	3.1
Victor Gonzalez	0.84	0.39	0.78	0.70	0.35	0	3.0
Kevin Ginkel	0.09	0.70	0.54	0.82	0.89	0	3.0
Dylan Cease	0.79	0.94	0.45	0.42	0.44	0	3.0
Brooks Raley	0.32	0.92	0.82	0.32	0.67	0	3.0
Mitch Keller	0.46	0.35	0.40	0.96	0.87	0	3.0
Stefan Crichton	0.05	0.11	0.93	0.99	0.96	0	3.0
Garrett Richards	0.49	0.24	0.75	0.87	0.68	0	3.0
Anthony Misiewicz	0.87	0.47	0.43	0.64	0.63	0	3.0
Archie Bradley	0.98	0.17	0.82	0.68	0.39	0	3.0
Alex Claudio	0.22	0.38	0.79	0.86	0.79	0	3.0
Yimi Garcia	0.69	0.60	0.96	0.25	0.51	0	3.0
Liam Hendriks	0.95	0.99	0.99	0.01	0.09	0	3.0
Jordan Romano	0.94	0.93	0.97	0.13	0.04	0	3.0
Cody Poteet	0.66	0.62	0.47	0.57	0.71	0	3.0
Alex Colome	0.61	0.31	0.96	0.64	0.49	0	3.0
Brady Singer	0.29	0.54	0.66	0.82	0.69	0	3.0
Raisel Iglesias	0.91	0.98	0.99	0.04	0.10	0	3.0
Sam Coonrod	0.42	0.75	0.84	0.53	0.47	0	3.0
Bruce Zimmermann	0.62	0.29	0.58	0.79	0.72	0	3.0
Darwinzon Hernandez	0.47	0.93	0.53	0.77	0.28	0	3.0
Junior Guerra	0.76	0.38	0.05	0.94	0.86	0	3.0
Codi Heuer	0.92	0.22	0.79	0.53	0.54	0	3.0
Andrew Wantz	0.28	0.95	0.66	0.40	0.70	0	3.0
Hector Neris	0.51	0.91	0.93	0.26	0.37	0	3.0
Sam Hentges	0.12	0.51	0.45	0.96	0.93	0	3.0
Miguel Sanchez	0.77	0.31	0.56	0.85	0.49	0	3.0
Rowan Wick	0.04	0.86	0.95	0.58	0.54	0	3.0
Humberto Mejia	0.10	0.33	0.60	0.98	0.96	0	3.0
Jose Cisnero	0.64	0.53	0.87	0.54	0.38	0	3.0
Richard Lovelady	0.91	0.74	0.85	0.15	0.32	0	3.0
Brad Boxberger	0.78	0.90	0.86	0.15	0.26	0	3.0
Shane McClanahan	0.81	0.76	0.62	0.46	0.31	0	3.0
Phillips Valdez	0.47	0.29	0.76	0.58	0.84	0	2.9
Josh Fleming	0.89	0.04	0.69	0.59	0.73	0	2.9
Brusdar Graterol	0.87	0.20	0.58	0.66	0.62	0	2.9
Jose Quintana	0.02	0.93	0.13	0.94	0.91	0	2.9
Kyle Funkhouser	0.92	0.37	0.71	0.65	0.30	0	2.9
Daniel Ponce de Leon	0.21	0.10	0.86	0.89	0.87	0	2.9
JC Mejia	0.13	0.33	0.59	0.88	0.99	0	2.9
Wandy Peralta	0.90	0.24	0.89	0.62	0.27	0	2.9
Michael Fulmer	0.72	0.61	0.94	0.47	0.17	0	2.9
Alex Lange	0.19	0.71	0.78	0.77	0.47	0	2.9
Brad Keller	0.59	0.32	0.31	0.92	0.77	0	2.9
Shawn Armstrong	0.18	0.84	0.17	0.77	0.94	0	2.9
Braxton Garrett	0.20	0.40	0.63	0.97	0.71	0	2.9
Diego Castillo	0.84	0.89	0.96	0.08	0.13	0	2.9
Chris Rodriguez	0.68	0.50	0.64	0.71	0.38	0	2.9
Josiah Gray	0.19	0.67	0.66	0.60	0.78	0	2.9
Brandon Bielak	0.59	0.36	0.74	0.63	0.59	0	2.9

---

4 Creating Model File

4.1 Additional Data Prep

4.1.1 Remove Variables which are based off current hitting numbers

Not all variables can be used for predictive modeling. Variables that go into the percentile ranking or are non-normalized metrics created after the fact (such as WAR - Wins above Replacement or RS - Raw Run Support) should be removed. However, metrics that are normalized by a per pitch basis (such as wFB/C) can remain as we expect pitchers to have similar performance in these metrics one year out.

#Be careful about RS - Run Support and RS/9

#Creating a new dataset to keep original intact
df_pitching_init3 = df_pitching_init2 %>% 
  select (-Name)

Lagged Percentile (_share) Variables can be used for predictive modeling. However since these variables were created for the Worth metric they must also be removed for modeling purposes.

#Order the dataset by lag columns
df_pitching_init4 =  arrange(df_pitching_init3, playerid,Season) #playerid is the Fangraph id assigned to each player

# Convert dataframe to data.table format
DT_pitcher2 = data.table(df_pitching_init4)

#designate columns to lag - just the new shares
cols1 = (c('Wins_share','SO_share','SV_share', 'ERA_share','WHIP_share','HLD_share','Worth'))
anscols = paste("lag", cols1, sep="_") 
DT_pitcher2[, (anscols) := data.table::shift(.SD, 1, NA, "lag"),by ='playerid', .SDcols=cols1] #Create 1 period lags by year

df_pitching_final = as.data.frame(DT_pitcher2) %>% 
  select(-c(Wins_share,SO_share,SV_share, ERA_share,WHIP_share,HLD_share))%>%
select(-FIP,-(RAR:WPA),-(wFB:wCH),-(`ERA-`:`xFIP-`),
       -SIERA,-(`RA9-WAR`:`Age Rng`),-kwERA,-(`wCH (pi)`:`wSL (pi)`),-(`K/9+`:`HR/FB%+`)) %>% select(-W,-SO,-SV,-HLD,-W_IP,-SO_IP,-SV_IP,-WHIP,-ERA,-HLD_IP)

4.1.2 Creating Training/Test Split

We split the data into Training Data (which is used to create the model) and test data (which is used to validate the model)

set.seed(15674)  # For reproducibility
# Create index for testing and training data
inTrain <- createDataPartition(y = df_pitching_final$Worth, p = 0.80, list = FALSE)
# subset pitching data for training
tr_2021 <- df_pitching_final[inTrain,]
# subset the rest to test and validate trained model
te_2021 <- df_pitching_final[-inTrain,]

nrow(tr_2021)/nrow(df_pitching_final) #check if split is 0.8

[1] 0.8

4.1.3 Treat Missing Data by Imputing Mean Value

Vtreat Package in R is excellent for treating data before using for modeling. Additional documentation can be found here.

treat_plan_2021 <- vtreat::designTreatmentsZ(
  dframe = tr_2021, # training data
  varlist = colnames(tr_2021) %>% .[. != "hitting_score1"], # input variables = all training data columns, except random
  codeRestriction = c("clean", "isBAD", "lev"), # derived variables types (drop cat_P)
  verbose = FALSE) # suppress messages

#clean stands for cleaned numerical variable, isBAD indicates that a value replacement has occurred (which indicates a missing value in this case), and lev is a binary indicator whether a particular value of that categorical variable was present.  

#### Checking Scoreframe

score_frame <- treat_plan_2021$scoreFrame %>% 
  select(varName, origName, code)

head(score_frame)

tr_treated_2021 <- vtreat::prepare(treat_plan_2021, tr_2021)
te_treated_2021 <- vtreat::prepare(treat_plan_2021, te_2021)


treat_plan_2021 <- vtreat::designTreatmentsZ(
  dframe = DT_pitcher2, # training data
  varlist = colnames(DT_pitcher2) %>% .[. != "hitting_score1"], # input variables = all training data columns, except random
  codeRestriction = c("clean", "isBAD", "lev"), # derived variables types (drop cat_P)
  verbose = FALSE) # suppress messages


total_treated_2021_pitching <- vtreat::prepare(treat_plan_2021, DT_pitcher2)

#tr_treated = tr
#te_treated = te

dim(tr_treated_2021) #note there are dummies for each player and team

[1] 3196 1415

---

4.1.4 Check Distribution of Training Population

The population used for Training should be indicative of Total Population

ggplot2::qplot(tr_treated_2021$Worth, main="Training Set") + geom_histogram(colour="black", fill="steelblue") + theme_bw()

`stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
`stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

#The skewness is actually a bit better than the overall data set
skewness(tr_treated_2021$Worth) 

[1] 0.077

---

5 Running XGboost Model

To keep things simple with modeling, we’ll turn the training data into simple input variables for caret::train, dropping the response variable and converting the data frame to a matrix. Documentation for this approach to XGboost can be found here.

5.1 Tuning the Model

5.1.1 Initial Non-Tuned Model

Break the data set into x and y inputs with x being a matrix. "_isBAD" is a category created by the Vtreat package in case you want to identify rows

input_x <- as.matrix(((tr_treated_2021))%>%
   select(-Worth) %>%                      
   select(!ends_with ("_isBAD")))

input_y <- tr_treated_2021$Worth

XGBoost with Default Hyperparameters:
The Variable Importance (caret::varImp(xgb_base_2021, scale = F) from the caret package shows the contribution of each variable to the initial model. Since this is untuned, we can expect the percentage imporantance to change as the models iterate through potential hyperparameters.
XGBoost documentation can be found for more general models here.

#Defaults for xgboost model
grid_default <- expand.grid(
  nrounds = 100,
  max_depth = 6,
  eta = 0.3,
  gamma = 0,
  colsample_bytree = 1,
  min_child_weight = 1,
  subsample = 1
)

#This is a blank train_control set, this will be updated after
train_control <- caret::trainControl(
  method = "none",
  verboseIter = FALSE, # no training log
  allowParallel = TRUE # FALSE for reproducible results 
)

xgb_base_2021 <- caret::train(
  x = input_x,
  y = input_y,
  trControl = train_control,
  tuneGrid = grid_default,
  method = "xgbTree",
  verbose = TRUE
)

caret::varImp(xgb_base_2021, scale = F  )

xgbTree variable importance

  only 20 most important variables shown (out of 773)

---

5.2 Further Variable Selection

5.2.1 Remove redundant and highly correlated variables

Selection Removal Step 1: Check for high correlations
Normally, this step is done early, but those steps were reserved for preparing the data

dep_cor1 <- t(as.data.frame(cor(tr_treated_2021[ , colnames(tr_treated_2021) != "Worth"],
                tr_treated_2021$Worth)))
dep_cor1 <-
as.data.frame(t(as.data.frame(dep_cor1)%>% 
  select(!starts_with("lag")) %>% #remove lag variables
  select(!contains("_isBAD")))) 

dep_cor1 <- tibble::rownames_to_column(dep_cor1,"VARIABLES")%>% #remove indicators for missing data
  filter(V1 > 0.40|V1 < -0.3)

dep_cor1

dep_cor2 <- colnames(row_to_names(t(dep_cor1),row_number = 1))

Let’s Remove variables with high correlation to worth metric, and metrics that are calculated after a player’s performance (such as WPA/RE24) or redundant (RS_IP)

input_x <- as.matrix(((tr_treated_2021))%>%
   select(-Worth) %>% #Remove some variables variables
     select (-RS_IP,-ER_IP,-R_IP,-REW,-RE24,-Clutch,-WPA_slash_LI,-Season #Remove redundant variables or non/weighted variables
) %>%      
select(!ends_with ("_isBAD"))) #indicator variable for missing data

input_y <- tr_treated_2021$Worth

Run the model on the new dataset to make sure the variable importances look fine

#Note Training parameters were set in initial model set up
xgb_base_2021 <- caret::train(
  x = input_x,
  y = input_y,
  trControl = train_control,
  tuneGrid = grid_default,
  method = "xgbTree",
  verbose = TRUE
)

caret::varImp(xgb_base_2021, scale = F  )

xgbTree variable importance

  only 20 most important variables shown (out of 765)

---

5.3 Model with new data

5.3.1 Tuning All Hyperparameters

A tune grid allows us to test a large amount of hyper-parameters and find the model with the lowest RMSE for predictions.
However, The more values you want to test and the greater the amount of Cross-Fold Validations (method = "cv"), the greater the computational time it will take. More information on the specific parameters can be found here.

# maximum number of trees
nrounds <- 1000

# note to start nrounds from 200, as smaller learning rates result in errors so
# big with lower starting points that they'll mess the scales
tune_grid <- expand.grid(
  nrounds = seq(from = 100, to = nrounds, by = 50),
  eta = c(0.01, 0.025, 0.05, 0.075, 0.1),
  max_depth = c(2, 4, 6, 8, 10),
  gamma = 0,
  colsample_bytree = 1,
  min_child_weight = 1,
  subsample = 1
)

tune_control <- caret::trainControl(
  method = "cv", # cross-validation
  number = 5, # with n folds 
  ## Note this was # out in the original code
  #index = createFolds(tr_treated$Id_clean), # fix the folds
  verboseIter = FALSE, # no training log
  allowParallel = FALSE # FALSE for reproducible results 
)

Running the initial tuning model

#Note I will be timing these runs to give an estimate on how long this model takes to run
start_time <- Sys.time()

xgb_tune_2021 <- caret::train(
  x = input_x,
  y = input_y,
  trControl = tune_control,
  tuneGrid = tune_grid,
  method = "xgbTree",
  verbose = FALSE
  ,verbosity = 0
)

end_time <- Sys.time()

end_time - start_time

Time difference of 21 mins

Tuning Plot and Variable Importance

varImp(xgb_tune_2021, scale = F  ) 

xgbTree variable importance

  only 20 most important variables shown (out of 765)

# helper function for the plots
tuneplot <- function(x, probs = .90) {
  ggplot(x) +
    coord_cartesian(ylim = c(quantile(x$results$RMSE, probs = probs), min(x$results$RMSE))) +
    theme_bw()
}

tuneplot(xgb_tune_2021)

---

5.3.2 Fine Tuning Model

5.3.2.1 Second Tuning: Maximum Depth and Minimum Child Weight

After fixing the learning rate to the best tune from the previous iteration and we’ll also set maximum depth to 3 +-1 (or +2 if max_depth == 2) to experiment a bit around the suggested best tune in previous step. Then, well fix maximum depth and minimum child weight.

tune_grid2 <- expand.grid(
  nrounds = seq(from = 50, to = nrounds, by = 50),
  eta = xgb_tune_2021$bestTune$eta,
  max_depth = ifelse(xgb_tune_2021$bestTune$max_depth == 2,
    c(xgb_tune_2021$bestTune$max_depth:4),
    xgb_tune_2021$bestTune$max_depth - 1:xgb_tune_2021$bestTune$max_depth + 1),
  gamma = 0,
  colsample_bytree = 1,
  min_child_weight = c(1, 2, 3),
  subsample = 1
)

xgb_tune2_2021 <- caret::train(
  x = input_x,
  y = input_y,
  trControl = tune_control,
  tuneGrid = tune_grid2,
  method = "xgbTree",
  verbose = TRUE
)

tuneplot(xgb_tune2_2021)

xgb_tune2_2021$bestTune

varImp(xgb_tune2_2021, scale = F  ) 

xgbTree variable importance

  only 20 most important variables shown (out of 765)

---

5.3.2.2 Third Tuning: Column and Row Sampling

tune_grid3 <- expand.grid(
  nrounds = seq(from = 50, to = nrounds, by = 50),
  eta = xgb_tune_2021$bestTune$eta,
  max_depth = xgb_tune2_2021$bestTune$max_depth,
  gamma = 0,
  colsample_bytree = c(0.4, 0.6, 0.8, 1.0),
  min_child_weight = xgb_tune2_2021$bestTune$min_child_weight,
  subsample = c(0.5, 0.75, 1.0)
)

xgb_tune3_2021 <- caret::train(
  x = input_x,
  y = input_y,
  trControl = tune_control,
  tuneGrid = tune_grid3,
  method = "xgbTree",
  verbose = TRUE
)

tuneplot(xgb_tune3_2021, probs = .95)

xgb_tune3_2021$bestTune

varImp(xgb_tune3_2021, scale = F  ) 

xgbTree variable importance

  only 20 most important variables shown (out of 765)

---

5.3.2.3 Fourth Tuning: Gamma

Next, we again pick the best values from previous step, and now will see whether changing the gamma has any effect on the model fit:

tune_grid4 <- expand.grid(
  nrounds = seq(from = 50, to = nrounds, by = 50),
  eta = xgb_tune_2021$bestTune$eta,
  max_depth = xgb_tune2_2021$bestTune$max_depth,
  gamma = c(0, 0.05,0.1, 0.2,0.4, 0.5, 0.7, 0.9, 1.0),
  colsample_bytree = xgb_tune3_2021$bestTune$colsample_bytree,
  min_child_weight = xgb_tune2_2021$bestTune$min_child_weight,
  subsample = xgb_tune3_2021$bestTune$subsample
)

xgb_tune4_2021 <- caret::train(
  x = input_x,
  y = input_y,
  trControl = tune_control,
  tuneGrid = tune_grid4,
  method = "xgbTree",
  verbose = TRUE
)

tuneplot(xgb_tune4_2021)

Warning: The shape palette can deal with a maximum of 6 discrete values because more than 6 becomes difficult
to discriminate; you have 9. Consider specifying shapes manually if you must have them.
Warning: Removed 60 rows containing missing values (geom_point).

xgb_tune4_2021$bestTune

varImp(xgb_tune4_2021, scale = F  ) 

xgbTree variable importance

  only 20 most important variables shown (out of 765)

---

5.3.2.4 Fifth Tuning: Reducing the Learning Rate

Now, we have tuned the hyperparameters and can start reducing the learning rate to get to the final model:

start_time <- Sys.time()

tune_grid5 <- expand.grid(
  nrounds = seq(from = 100, to = 10000, by = 75),
   eta = c(0.01, 0.015, 0.025,0.035, 0.05,0.75, 0.1),
  max_depth = xgb_tune2_2021$bestTune$max_depth,
  gamma = xgb_tune4_2021$bestTune$gamma,
  colsample_bytree = xgb_tune3_2021$bestTune$colsample_bytree,
  min_child_weight = xgb_tune2_2021$bestTune$min_child_weight,
  subsample = xgb_tune3_2021$bestTune$subsample
)



xgb_tune5_2021 <- caret::train(
  x = input_x,
  y = input_y,
  trControl = tune_control,
  tuneGrid = tune_grid5,
  method = "xgbTree",
  verbose = TRUE
)

#tuneplot(xgb_tune5_2021)

end_time <- Sys.time()

end_time - start_time

Time difference of 28 mins

xgb_tune5_2021$bestTune

varImp(xgb_tune5_2021, scale = F  ) 

xgbTree variable importance

  only 20 most important variables shown (out of 765)

---

5.3.2.5 Fitting Final Model

(final_grid_2021 <- expand.grid(
  nrounds = xgb_tune5_2021$bestTune$nrounds,
  eta = xgb_tune5_2021$bestTune$eta,
  max_depth = xgb_tune5_2021$bestTune$max_depth,
  gamma = xgb_tune5_2021$bestTune$gamma,
  colsample_bytree = xgb_tune5_2021$bestTune$colsample_bytree,
  min_child_weight = xgb_tune5_2021$bestTune$min_child_weight,
  subsample = xgb_tune5_2021$bestTune$subsample
))

(xgb_model_2021 <- caret::train(
  x = input_x,
  y = input_y,
  trControl = train_control,
  tuneGrid = final_grid_2021,
  method = "xgbTree",
  verbose = TRUE
))

eXtreme Gradient Boosting 

3196 samples
 765 predictor

No pre-processing
Resampling: None 

varImp(xgb_model_2021, scale = F  ) 

xgbTree variable importance

  only 20 most important variables shown (out of 765)

---

5.4 Model Performance

5.4.1 Checking Model on Test Split Data

We don’t need to look too closely at are training data as Xgboost will heavily overfit the model based on that data. The more important part is how the model performs on in predicting our Test Sample that was not included.

y_pred_test <- predict(xgb_model_2021, data.matrix(te_treated_2021))

test_stats= cbind((te_treated_2021$Worth),y_pred_test)

test_statsR2 = cor(test_stats[,1],test_stats[,2])^2

print(test_statsR2)

[1] 0.71

y_pred_train <- predict(xgb_model_2021, data.matrix(tr_treated_2021))

train_stats = cbind((tr_treated_2021$Worth),y_pred_train)

train_statsR2 = cor(train_stats[,1],train_stats[,2])^2

print(train_statsR2)

[1] 0.98

#test dataset
x <- select(te_treated_2021, -Worth)
y <- (te_treated_2021$Worth)

(xgb_model_rmse <- ModelMetrics::rmse(y, predict(xgb_model_2021, newdata = x)))

[1] 0.34

holdout_x <- select(tr_treated_2021, -Worth)
holdout_y <- tr_treated_2021$Worth

(xgb_model_rmse <- ModelMetrics::rmse(holdout_y, predict(xgb_model_2021, newdata = holdout_x)))

[1] 0.081

5.4.1.1 Graphical Representation of Model

ggplot2::ggplot() +
  aes(x = test_stats[,1], y = test_stats[,2]) +
  geom_jitter() +
  xlab("Predicted Values") +
  ylab("Actual Values") +
  ggtitle("Results of Pitching Model on Test Data")+
  theme(plot.title = element_text(hjust = 0.5,size = 22,color ="steel blue"))+
  geom_smooth(method = "lm")

`geom_smooth()` using formula 'y ~ x'

---

6 Creating 2022 Projections from Model

6.1 Re-fit model for Important Variables

Now that we have an acceptable model, we can use it to create projections for how well we think players should do in 2022 based on their hitting statistics in 2021. First let’s reduce

Step 1: Only keep variables with high enough importance in model

vip(xgb_model_2021, num_features = 30)  # 10 is the default, 30 gives a visual on the top 30 most important features of the model

unscalevi = vi(xgb_model_2021, method="model") #shows the numbers behind the plot

unscalevi$Importance_perc = with(unscalevi,Importance/sum(Importance)) #adds percentages 

unscalevi # importance by variables

variables_to_keep_2021 = subset(unscalevi, Importance_perc > 0.0010) %>% select(Variable) #Keep Variables that explain at least a small amount [0.1%] of the model. This is a low threshold for inclusion ,but you can adjust this

variables_to_keep_2021b = t(variables_to_keep_2021)

variables_to_keep_2022 = colnames(row_to_names(variables_to_keep_2021b,row_number = 1))

tr_treated_2022 = tr_treated_2021 %>%  select(Worth,one_of(variables_to_keep_2022),starts_with("Team_lev_x_")) #keep modeled important variables along with team indicator variables

te_treated_2022 = te_treated_2021 %>%  select(Worth,one_of(variables_to_keep_2022),starts_with("Team_lev_x_"))

input_x_2022 = as.matrix(select(tr_treated_2022, -Worth))

input_y_2022 = tr_treated_2022$Worth

---

Step 2: Re-fit model with reduced variable scope
Note from the best tune below the nrounds - is the max I set above and eta is at the lowest possible value. This could cause potential overfitting issues, but from our Actual vs. Predicted Graph we know this not to be the case.

(final_grid_2021 <- expand.grid(
  nrounds = xgb_tune5_2021$bestTune$nrounds,
  eta = xgb_tune5_2021$bestTune$eta,
  max_depth = xgb_tune5_2021$bestTune$max_depth,
  gamma = xgb_tune5_2021$bestTune$gamma,
  colsample_bytree = xgb_tune5_2021$bestTune$colsample_bytree,
  min_child_weight = xgb_tune5_2021$bestTune$min_child_weight,
  subsample = xgb_tune5_2021$bestTune$subsample
))

(xgb_model_2022 <- caret::train(
  x = input_x_2022,
  y = input_y_2022,
  trControl = train_control,
  tuneGrid = final_grid_2021,
  method = "xgbTree",
  verbose = TRUE
))

eXtreme Gradient Boosting 

3196 samples
 125 predictor

No pre-processing
Resampling: None 

vip(xgb_model_2022, num_features = 30)

unscalevi24 = vi(xgb_model_2022, method="model")

unscalevi24$Importance_perc = with(unscalevi24,Importance/sum(Importance)) 

unscalevi24

# Save work for later prediction

save(xgb_model_2022,file = '2022_Pitching5x5_Model.Rdata')

pitching5x5 = xgb_model_2022

pitchinginput = input_x_2022

---

6.2 Get 2022 list of players

6.2.1 Arrange the Data so the Columns are in the exact order as the model

First let’s prepare a file for predicting based on our model object

variableslag5x= row_to_names(as.data.frame(t(variables_to_keep_2022)),row_number = 1)  %>% select (starts_with("lag"))

variables_nolag5x = (owmr::remove_prefix(variableslag5x,"lag" , sep = "_"))

Data_Predict_2022a5x = total_treated_2021_pitching %>% select (one_of(colnames(variables_nolag5x)),Season,playerid)

colnames(Data_Predict_2022a5x) <- paste0("lag_", colnames(Data_Predict_2022a5x))

Data_Predict_2022b5x = total_treated_2021_pitching %>% select (one_of(colnames(variables_nolag5x)))
colnames(Data_Predict_2022b5x) = colnames(variableslag5x)

variables_to_keep_2022_nolag5x = total_treated_2021_pitching %>% select(one_of(variables_to_keep_2022),Season,playerid,starts_with("Team_lev_x_"))%>% select(-one_of(colnames(Data_Predict_2022b5x)))


Data_predict_20225x = sqldf(
  "
  select a.*,b.* from
  Data_Predict_2022a5x a,
  variables_to_keep_2022_nolag5x b
  on b.playerid = a.lag_playerid
  and b.Season = a.lag_Season
  "
) %>% select(-lag_playerid,lag_Season) %>%
  filter(Season == 2021) %>% 
  select(one_of(variables_to_keep_2022),starts_with("Team_lev_x_"))

---

6.3 Create Predictions for Model

6.3.1 Run Projections on Players who Played in 2021

This is the raw prediction score per IP for each pitcher

pitching_predictions5x = as.data.frame(predict(xgb_model_2022,Data_predict_20225x))

names(pitching_predictions5x) = c("Predict_Score")

Data_predict_2022_w_Pitching_Predictions5x = cbind(Data_predict_2022,pitching_predictions5x) %>% select(playerid,Predict_Score)

head(Data_predict_2022_w_Pitching_Predictions5x)

NA

---

6.3.2 Load in Latest 2022 Projections for Innings Pitched

Downloaded from FanGraphs here.

Latest_2022_pitchingdata_FP = read_csv("FanGraph_Fantasy_Baseball_Pitching.csv")

Rows: 817 Columns: 27
-- Column specification -------------------------------------------------------------------------------------------
Delimiter: ","
chr  (3): Name, Team, playerid
dbl (24): GS, G, IP, W, L, QS, SV, HLD, H, ER, HR, SO, BB, WHIP, K/9, BB/9, ERA, FIP, WAR, RA9-WAR, ADP, InterS...

i Use `spec()` to retrieve the full column specification for this data.
i Specify the column types or set `show_col_types = FALSE` to quiet this message.

Latest_2022_pitchingdata_FP

NA

---

As you can see from the chart below there aren’t many elite pitchers in the 87+ Predict score range.

Pitching_Data_NonAdj_Projections5x = sqldf(
  "
  select a.*,b.Predict_Score
  from Latest_2022_pitchingdata_FP a 
  left join 
  Data_predict_2022_w_Pitching_Predictions5x b
  on a.playerid = b.playerid
  "
) %>% filter(ADP<370 | is.na(Predict_Score)==F)


Pitching_Data_Adj_Projections5x =
Pitching_Data_NonAdj_Projections5x %>% 
  mutate(
    Avg_IP = 60,
    AdjPredict_Score_raw = ifelse(is.na(Predict_Score),NA,Predict_Score*(IP/Avg_IP)),
    max_predscore= max(AdjPredict_Score_raw,na.rm = T),
    AdjPredict_Score = ifelse (is.na(AdjPredict_Score_raw),NA,AdjPredict_Score_raw *100/max_predscore),
    WAR_rank = order(order(rank(WAR,ties.method = 'average'),decreasing = TRUE)),
    AdjPredict_Score_Rank = order(order(rank(AdjPredict_Score,ties.method = 'average'),decreasing = TRUE))-sum(is.na(AdjPredict_Score)),
        Ranks_Above_ADP = ADP - AdjPredict_Score_Rank
  ) %>% select (Name,ADP,WAR, WAR_rank,AdjPredict_Score ,AdjPredict_Score_Rank,Ranks_Above_ADP)


  

ggplot2::qplot(Pitching_Data_Adj_Projections5x$AdjPredict_Score, main="Predictions") + geom_histogram(colour="black", fill="grey") + theme_bw()

`stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
`stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

---

7 2022 Projections Full

7.1 Table of Pitching Projections (Players who Didn’t Play in 2021 - Recieve an NA)

AdjPredict_Score are normalized to 100

tableexport =
Pitching_Data_Adj_Projections5x %>%
  arrange (ADP,WAR) %>% 
  kbl() %>% 
 kable_material(c("striped", "hover","condensed","responsive"),full_width = F,fixed_thead = T)

save_kable(tableexport,file = "Pitching5x5.html")

#tableexport

This is a better formatted Table

ft_dt <- Pitching_Data_Adj_Projections5x[1:nrow(Pitching_Data_Adj_Projections5x), 1:ncol(Pitching_Data_Adj_Projections5x)] %>% 
  filter(AdjPredict_Score_Rank>0)%>%  arrange((AdjPredict_Score_Rank))

ft_dt$ADP <- color_tile("white", "red")(ft_dt$ADP)

ft_dt$WAR <- color_bar("lightblue")(ft_dt$WAR)

ft_dt$AdjPredict_Score<- color_bar("lightblue")(ft_dt$AdjPredict_Score)

ft_dt$WAR_Rank <- color_tile("green","orange")(ft_dt$WAR_rank)

ft_dt$Predict_Rank <- color_tile("green","orange")(ft_dt$AdjPredict_Score_Rank) 


ft_dt$Ranks_Above_ADP <- 
  ifelse(
  ft_dt$Ranks_Above_ADP < 0,
  cell_spec(round(ft_dt$Ranks_Above_ADP,2), color = "red", italic = T),
  cell_spec(round(ft_dt$Ranks_Above_ADP,2), color = "green", italic = T)
)


ft_dt2 <- ft_dt[c("Name", "ADP", "WAR", "AdjPredict_Score", "WAR_Rank","Predict_Rank","Ranks_Above_ADP")]



table_export = 
kbl(ft_dt2, escape = F) %>% 
 kable_material(c("striped", "hover","condensed","responsive"),full_width = F,fixed_thead = T) %>%   column_spec(6, width = "3cm") %>%
  add_header_above(c(" ", "Scores" = 3, "Ranks" = 2," "))
save_kable(table_export,file = "Pitching5x5_updated.html")
  
table_export  

	Scores			Ranks
Name	ADP	WAR	AdjPredict_Score	WAR_Rank	Predict_Rank	Ranks_Above_ADP
Dylan Cease	81.2	3.3	100.0	20	1	80.2
Eduardo Rodriguez	150.1	3.5	99.8	15	2	148.1
Shane Bieber	31.2	4.4	93.2	4	3	28.2
Aaron Nola	41.0	4.2	89.8	7	4	37
Nick Pivetta	350.8	1.7	88.7	86	5	345.8
Tarik Skubal	197.6	2.0	88.0	61	6	191.6
Germán Márquez	261.7	2.9	88.0	29	7	254.7
Tyler Mahle	132.2	2.6	87.8	39	8	124.2
Brady Singer	492.4	2.2	87.8	53	9	483.4
Jon Gray	236.9	2.3	87.5	51	10	226.9
Shane McClanahan	109.6	2.6	86.3	40	11	98.6
Brad Keller	582.1	1.6	85.3	91	12	570.1
Luis Castillo	103.2	3.3	84.9	21	13	90.2
Zac Gallen	154.3	2.3	84.8	50	14	140.3
Josiah Gray	290.7	1.6	84.3	89	15	275.7
José Quintana	575.5	1.0	84.2	143	16	559.5
Patrick Corbin	456.0	1.6	82.6	95	17	439
Alek Manoah	93.0	2.8	82.0	33	18	75
Gerrit Cole	7.1	5.0	81.9	3	19	-11.9
Yusei Kikuchi	301.3	1.7	81.1	83	20	281.3
Sean Manaea	141.4	2.8	80.6	31	21	120.4
Logan Gilbert	161.8	2.2	80.0	52	22	139.8
Lucas Giolito	44.2	3.7	79.5	10	23	21.2
Logan Webb	69.9	3.7	79.1	11	24	45.9
Andrew Heaney	299.4	1.4	78.9	102	25	274.4
Blake Snell	113.4	2.4	78.6	49	26	87.4
Frankie Montas	92.8	3.3	78.6	19	27	65.8
Corbin Burnes	10.8	5.3	77.9	1	28	-17.2
Freddy Peralta	54.0	3.6	77.9	12	29	25
Kyle Hendricks	279.8	1.7	77.6	79	30	249.8
Jesús Luzardo	289.5	0.8	76.2	158	31	258.5
Walker Buehler	18.6	3.8	75.6	8	32	-13.4
Framber Valdez	149.6	3.1	75.0	26	33	116.6
Aaron Civale	269.5	1.5	74.8	96	34	235.5
José Berríos	77.5	3.2	74.6	23	35	42.5
Steven Matz	249.2	1.9	74.5	67	36	213.2
Luis Garcia	144.7	2.0	74.4	64	37	107.7
Robbie Ray	49.9	3.0	74.2	27	38	11.9
Kevin Gausman	71.4	3.2	73.8	24	39	32.4
Ian Anderson	151.1	2.1	73.5	56	40	111.1
Sonny Gray	168.2	2.7	73.3	38	41	127.2
Yu Darvish	100.3	2.7	72.2	34	42	58.3
Kyle Gibson	393.9	1.9	72.1	71	43	350.9
Cole Irvin	532.2	1.5	71.8	97	44	488.2
Nathan Eovaldi	135.1	3.6	71.6	13	45	90.1
Trevor Rogers	92.9	3.1	71.2	25	46	46.9
Mitch Keller	531.5	1.2	71.0	123	47	484.5
Triston McKenzie	229.0	1.6	70.7	93	48	181
Julio Urías	34.7	3.5	70.5	14	49	-14.3
Jordan Montgomery	208.7	2.7	70.4	35	50	158.7
Patrick Sandoval	196.0	2.4	69.9	46	51	145
Zac Lowther	999.0	0.8	69.2	159	52	947
Antonio Senzatela	585.0	2.0	68.9	62	53	532
JT Brubaker	537.7	1.4	68.3	105	54	483.7
Kris Bubic	536.0	1.1	68.1	125	55	481
Bruce Zimmermann	600.8	1.3	67.9	114	56	544.8
Shohei Ohtani	9.2	2.8	67.9	32	57	-47.8
Dylan Bundy	498.5	1.3	67.7	109	58	440.5
Luis Patiño	290.7	1.5	67.5	99	59	231.7
Mike Minor	532.2	2.0	67.5	63	60	472.2
Kyle Freeland	574.4	1.5	67.4	98	61	513.4
Max Fried	69.8	3.5	67.3	16	62	7.8
Zach Plesac	346.6	1.5	66.8	100	63	283.6
Dane Dunning	484.6	1.9	66.7	72	64	420.6
Hyun Jin Ryu	207.3	2.4	66.6	45	65	142.3
Alex Cobb	242.9	1.9	66.4	70	66	176.9
Brandon Woodruff	21.2	4.2	66.3	6	67	-45.8
Tanner Houck	206.4	2.4	66.2	47	68	138.4
Joe Ryan	221.5	2.2	66.1	54	69	152.5
Jordan Lyles	578.9	0.6	65.8	196	70	508.9
Joe Musgrove	70.6	3.2	65.4	22	71	-0.4
Max Scherzer	19.7	3.8	65.1	9	72	-52.3
Dallas Keuchel	567.3	1.2	64.9	119	73	494.3
Alex Wood	219.9	1.7	64.8	78	74	145.9
Carlos Hernández	412.5	1.2	64.4	121	75	337.5
Bailey Ober	261.3	2.2	64.4	55	76	185.3
Chris Flexen	413.6	1.9	64.0	73	77	336.6
Michael Kopech	161.1	2.5	63.5	43	78	83.1
Matt Manning	543.4	1.1	63.1	124	79	464.4
Sandy Alcantara	43.4	3.3	63.0	18	80	-36.6
Huascar Ynoa	245.8	1.8	63.0	76	81	164.8
Zack Wheeler	31.0	4.4	62.9	5	82	-51
Cristian Javier	284.0	1.2	62.8	118	83	201
Charlie Morton	94.3	2.9	62.5	30	84	10.3
Austin Gomber	551.9	1.2	62.4	116	85	466.9
Reid Detmers	427.4	1.1	62.4	134	86	341.4
Lance Lynn	70.0	3.4	62.0	17	87	-17
Keegan Akin	600.0	1.0	61.8	139	88	512
Chris Bassitt	128.5	2.4	61.7	48	89	39.5
Ranger Suárez	179.3	2.4	61.2	44	90	89.3
Carlos Rodón	110.2	2.9	61.2	28	91	19.2
Merrill Kelly	526.0	1.7	60.8	81	92	434
Spencer Howard	575.8	0.9	60.6	150	93	482.8
Vladimir Gutierrez	585.4	0.2	59.4	322	94	491.4
Corey Kluber	324.4	1.1	58.9	130	95	229.4
Cal Quantrill	267.3	1.7	58.9	82	96	171.3
Carlos Carrasco	284.6	1.3	58.6	110	97	187.6
Casey Mize	279.4	1.7	58.5	84	98	181.4
Zach Davies	595.9	0.2	58.4	343	99	496.9
James Kaprielian	448.3	1.1	58.4	126	100	348.3
Aaron Ashby	264.1	1.5	58.2	101	101	163.1
Ryan Yarbrough	541.5	1.4	58.0	106	102	439.5
Marco Gonzales	319.6	1.7	57.9	80	103	216.6
Erick Fedde	589.7	0.6	57.8	188	104	485.7
Pablo López	140.6	2.7	57.6	36	105	35.6
Tony Gonsolin	300.6	0.9	57.4	148	106	194.6
J.A. Happ	597.0	0.6	57.4	199	107	490
Elieser Hernandez	339.6	1.0	57.3	145	108	231.6
Eric Lauer	306.3	1.4	57.3	103	109	197.3
José Suarez	535.1	1.3	57.2	113	110	425.1
Taylor Hearn	591.2	0.7	56.8	183	111	480.2
Wade Miley	502.0	1.2	56.5	120	112	390
Marcus Stroman	187.5	2.6	56.5	41	113	74.5
Alec Mills	594.1	0.6	55.6	211	114	480.1
Chris Sale	85.9	1.9	55.4	66	115	-29.1
Anthony DeSclafani	212.6	1.8	54.9	77	116	96.6
Jack Flaherty	115.4	1.6	54.8	87	117	-1.6
Jameson Taillon	299.6	2.1	54.7	59	118	181.6
Clayton Kershaw	149.4	2.5	54.2	42	119	30.4
José Urquidy	212.1	2.0	54.1	65	120	92.1
Taijuan Walker	414.5	0.6	53.7	185	121	293.5
Madison Bumgarner	507.1	1.1	53.7	131	122	385.1
Michael Lorenzen	543.3	0.6	53.4	203	123	420.3
Zach Eflin	485.0	1.9	53.3	69	124	361
Jacob deGrom	19.7	5.1	53.0	2	125	-105.3
Tylor Megill	334.2	1.1	52.7	129	126	208.2
Bryse Wilson	600.5	0.5	52.3	223	127	473.5
Chris Paddack	436.8	1.2	52.2	117	128	308.8
Wil Crowe	600.8	0.1	51.9	379	129	471.8
Drew Rasmussen	282.4	1.6	51.5	88	130	152.4
Stephen Strasburg	294.0	1.3	51.5	111	131	163
Drew Smyly	577.4	0.6	51.3	208	132	445.4
Daniel Bard	579.3	0.4	51.1	266	133	446.3
Adrian Houser	465.0	1.4	50.5	104	134	331
Luke Weaver	514.4	1.1	50.3	128	135	379.4
Martín Pérez	598.8	0.8	49.4	168	136	462.8
John Means	217.1	2.7	49.2	37	137	80.1
Daniel Lynch	559.5	0.7	48.9	174	138	421.5
Michael Wacha	549.7	0.9	48.9	149	139	410.7
A.J. Alexy	596.2	0.6	48.5	186	140	456.2
Edward Cabrera	523.6	0.6	47.9	202	141	382.6
Glenn Otto	582.9	0.9	47.2	152	142	440.9
José Alvarado	587.6	0.5	47.2	237	143	444.6
Adam Wainwright	194.7	2.1	46.9	58	144	50.7
Zack Greinke	312.8	1.6	46.7	90	145	167.8
Adam Ottavino	598.1	0.2	46.3	328	146	452.1
James Karinchak	468.6	0.7	45.3	171	147	321.6
Nestor Cortes	333.1	1.4	45.3	107	148	185.1
Randy Dobnak	600.4	0.8	45.3	162	149	451.4
Aroldis Chapman	83.3	0.9	45.2	151	150	-66.7
Rich Hill	477.4	0.7	44.9	173	151	326.4
Garrett Whitlock	239.2	1.2	44.7	122	152	87.2
Johnny Cueto	585.8	0.3	44.7	283	153	432.8
Justus Sheffield	599.5	0.4	44.0	269	154	445.5
Michael Pineda	465.2	1.4	43.9	108	155	310.2
Josh Fleming	600.8	0.5	43.8	222	156	444.8
Trevor May	565.3	0.4	43.6	239	157	408.3
Kyle Finnegan	351.5	0.2	43.4	336	158	193.5
Matt Barnes	245.2	0.8	43.2	164	159	86.2
Tyler Anderson	564.5	0.7	43.0	172	160	404.5
Paul Sewald	289.0	0.5	43.0	234	161	128
Jake Diekman	585.9	0.5	42.9	232	162	423.9
Héctor Neris	493.9	0.4	42.9	242	163	330.9
Tanner Rainey	371.9	0.2	42.8	325	164	207.9
Gregory Soto	194.3	0.5	42.6	228	165	29.3
Miles Mikolas	512.8	1.6	42.5	94	166	346.8
Amir Garrett	586.6	0.2	42.5	333	167	419.6
Phil Maton	600.4	0.4	42.2	248	168	432.4
Dillon Peters	600.7	0.0	42.1	421	169	431.7
Scott Barlow	159.3	0.9	41.8	155	170	-10.7
Carlos Estévez	528.6	0.2	41.7	337	171	357.6
David Peterson	595.2	0.7	41.6	175	172	423.2
Lance McCullers Jr.	260.0	1.7	41.4	85	173	87
Pete Fairbanks	517.3	0.6	40.9	194	174	343.3
Lucas Sims	243.6	0.7	40.9	180	175	68.6
Justin Steele	598.7	0.4	40.8	245	176	422.7
Paolo Espino	599.2	0.2	40.7	327	177	422.2
Edwin Díaz	63.5	1.1	40.6	135	178	-114.5
Aaron Bummer	595.4	1.0	40.6	137	179	416.4
Chad Green	403.2	1.0	40.4	147	180	223.2
Seth Lugo	598.1	0.5	40.3	236	181	417.1
Chris Stratton	519.4	0.3	40.2	284	182	337.4
Tanner Scott	599.0	0.7	40.0	182	183	416
David Price	581.5	0.5	39.4	216	184	397.5
Zach Thompson	546.8	1.0	39.3	146	185	361.8
Mike Mayers	594.7	0.5	39.2	221	186	408.7
Griffin Canning	594.2	0.5	39.1	213	187	407.2
Sam Hentges	999.0	0.3	38.9	306	188	811
Devin Williams	287.1	1.1	38.9	133	189	98.1
Heath Hembree	600.8	0.2	38.7	338	190	410.8
Mychal Givens	549.7	0.1	38.5	419	191	358.7
Alex Colomé	378.0	0.1	38.5	387	192	186
Domingo Germán	494.6	1.1	38.4	132	193	301.6
Raisel Iglesias	47.6	1.0	38.1	138	194	-146.4
Brad Boxberger	588.3	0.2	38.1	331	195	393.3
Brent Suter	587.6	0.4	38.1	268	196	391.6
Giovanny Gallegos	113.5	1.0	38.1	140	197	-83.5
Daniel Norris	999.0	0.1	38.0	398	198	801
Hansel Robles	596.0	-0.1	37.9	505	199	397
Austin Voth	600.4	0.0	37.8	420	200	400.4
Jaime Barría	999.0	0.5	37.7	238	201	798
Jordan Romano	88.3	0.8	37.7	160	202	-113.7
Kyle Funkhouser	999.0	0.1	37.6	386	203	796
Will Smith	120.8	0.5	37.4	227	204	-83.2
Taylor Rogers	172.5	1.1	37.3	127	205	-32.5
Yimi García	581.1	0.2	37.3	352	206	375.1
Tyler Wells	454.6	0.9	37.2	153	207	247.6
Camilo Doval	159.8	0.5	37.1	229	208	-48.2
Justin Dunn	577.6	0.6	37.1	198	209	368.6
Joe Jiménez	999.0	0.2	36.9	329	210	789
Garrett Richards	598.7	0.4	36.8	272	211	387.7
Hirokazu Sawamura	589.6	0.0	36.6	443	212	377.6
Jake Odorizzi	533.5	0.8	36.6	163	213	320.5
Ryne Stanek	598.8	0.2	36.6	330	214	384.8
J.C. Mejía	999.0	0.1	36.4	389	215	784
Bryan Shaw	597.8	-0.1	36.3	507	216	381.8
Luis Gil	525.9	0.8	36.2	165	217	308.9
Rowan Wick	302.0	0.4	36.2	249	218	84
Rafael Dolis	999.0	0.0	36.1	434	219	780
Tyler Alexander	586.8	0.8	36.1	167	220	366.8
Liam Hendriks	32.0	1.6	36.1	92	221	-189
J.B. Wendelken	600.1	0.2	36.0	341	222	378.1
Brad Hand	520.0	0.2	35.9	351	223	297
Jeurys Familia	597.5	0.3	35.8	305	224	373.5
Trevor Stephan	999.0	0.1	35.8	368	225	774
Michael Fulmer	347.0	0.6	35.8	192	226	121
Griffin Jax	600.8	0.1	35.7	363	227	373.8
Ryan Pressly	64.3	1.0	35.5	136	228	-163.7
Jackson Kowar	594.6	0.5	35.5	233	229	365.6
José Ureña	999.0	0.0	35.3	467	230	769
Blake Treinen	147.6	0.8	35.3	166	231	-83.4
José Cisnero	999.0	0.3	35.2	281	232	767
Cole Sulser	427.2	0.7	35.1	177	233	194.2
Joely Rodríguez	600.9	0.5	35.1	214	234	366.9
Lou Trivino	237.2	0.2	35.1	316	235	2.2
Jeff Hoffman	999.0	0.0	35.0	426	236	763
Craig Kimbrel	164.4	0.7	34.9	176	237	-72.6
Jake Brentz	600.8	0.3	34.8	301	238	362.8
Diego Castillo	409.2	0.6	34.7	191	239	170.2
Caleb Smith	592.3	0.0	34.6	441	240	352.3
Josh Sborz	600.9	0.3	34.5	290	241	359.9
Chad Kuhl	600.7	0.2	34.4	349	242	358.7
Dean Kremer	999.0	0.6	34.4	210	243	756
Keegan Thompson	999.0	0.0	34.4	429	244	755
Matt Wisler	599.6	0.4	34.3	270	245	354.6
Ryan Helsley	600.2	0.1	34.3	380	246	354.2
Anthony Banda	999.0	-0.1	34.2	514	247	752
Andrew Kittredge	257.6	0.8	34.1	169	248	9.6
Garrett Crochet	549.8	0.9	34.0	154	249	300.8
Tony Santillan	600.3	0.4	34.0	263	250	350.3
Eli Morgan	600.5	0.2	33.9	346	251	349.5
Kenley Jansen	90.0	0.6	33.7	187	252	-162
Andrew Wantz	999.0	0.4	33.7	262	253	746
Spencer Patton	599.0	0.1	33.4	410	254	345
Josh Hader	30.5	1.3	33.3	115	255	-224.5
Ross Stripling	583.7	0.4	33.2	256	256	327.7
Trevor Richards	600.5	0.4	33.1	261	257	343.5
Génesis Cabrera	596.6	0.6	32.9	205	258	338.6
Robert Stephenson	590.5	0.0	32.8	466	259	331.5
Vince Velasquez	599.4	0.3	32.8	308	260	339.4
Matthew Boyd	591.0	1.0	32.4	144	261	330
Josh Staumont	505.8	0.4	32.4	244	262	243.8
Adbert Alzolay	444.0	0.7	32.4	179	263	181
Corey Knebel	151.3	0.6	32.4	204	264	-112.7
Anthony Misiewicz	999.0	0.3	32.3	295	265	734
Dylan Floro	221.3	0.4	32.3	251	266	-44.7
A.J. Minter	596.5	0.5	32.3	215	267	329.5
Dinelson Lamet	400.5	0.9	32.2	157	268	132.5
Brusdar Graterol	567.4	0.7	32.0	178	269	298.4
Sam Howard	999.0	0.3	32.0	310	270	729
Sam Coonrod	600.9	0.4	31.8	271	271	329.9
Jonathan Loáisiga	393.7	1.0	31.8	142	272	121.7
Clay Holmes	600.3	0.6	31.7	201	273	327.3
Mark Melancon	131.8	0.4	31.6	247	274	-142.2
Jakob Junis	600.1	0.2	31.6	332	275	325.1
Archie Bradley	600.7	0.2	31.5	348	276	324.7
Tyler Duffey	577.8	0.4	31.3	260	277	300.8
Brett Anderson	600.9	0.6	31.3	190	278	322.9
Anthony Bender	412.8	0.7	31.1	184	279	133.8
Tim Mayza	599.3	0.6	31.1	197	280	319.3
Wandy Peralta	999.0	0.2	30.9	340	281	718
Ian Kennedy	409.8	-0.1	30.9	512	282	127.8
David Bednar	190.8	0.9	30.8	156	283	-92.2
Caleb Thielbar	999.0	0.6	30.8	195	284	715
Emmanuel Clase	57.5	1.3	30.7	112	285	-227.5
Jorge Alcala	459.0	0.6	30.7	209	286	173
Collin McHugh	545.5	0.6	30.4	189	287	258.5
Trevor Williams	601.0	0.2	30.4	326	288	313
Emilio Pagán	551.7	0.1	30.4	364	289	262.7
Patrick Murphy	999.0	0.4	30.2	254	290	709
J.P. Feyereisen	597.6	0.0	30.1	487	291	306.6
Kendall Graveman	531.0	0.4	30.1	240	292	239
Jorge López	599.9	0.4	30.0	265	293	306.9
Tyler Rogers	534.3	0.5	30.0	230	294	240.3
Kyle Zimmer	999.0	-0.2	29.7	528	295	704
Max Kranick	999.0	0.3	29.6	297	296	703
Joe Ross	599.6	0.7	29.4	170	297	302.6
Sergio Romo	600.5	0.0	29.4	450	298	302.5
Brooks Raley	600.9	0.3	29.3	296	299	301.9
Daniel Hudson	550.2	0.5	29.1	225	300	250.2
Pierce Johnson	433.3	0.6	29.1	200	301	132.3
Ryan Hendrix	999.0	0.0	29.1	468	302	697
Tim Hill	999.0	0.3	29.0	293	303	696
Nick Wittgren	999.0	0.1	29.0	374	304	695
Mike Foltynewicz	600.4	-0.1	28.9	511	305	295.4
Cionel Pérez	999.0	0.5	28.8	235	306	693
Lucas Gilbreath	596.4	0.0	28.7	471	307	289.4
Kwang Hyun Kim	593.0	0.3	28.6	303	308	285
Taylor Widener	600.5	0.0	28.4	424	309	291.5
Alex Vesia	600.0	0.3	28.4	289	310	290
Sean Doolittle	598.5	0.1	28.3	402	311	287.5
Rafael Montero	999.0	0.1	28.0	394	312	687
Connor Brogdon	600.4	0.4	27.9	252	313	287.4
Jake Woodford	599.5	0.0	27.9	460	314	285.5
John King	999.0	0.5	27.7	219	315	684
Josh Taylor	999.0	0.5	27.7	217	316	683
José Quijada	999.0	0.5	27.6	212	317	682
Josh Tomlin	999.0	-0.3	27.4	532	318	681
Sean Newcomb	600.1	0.1	27.3	400	319	281.1
Tyler Kinley	999.0	0.0	27.3	433	320	679
Justin Wilson	999.0	0.1	27.0	377	321	678
Luke Jackson	598.1	0.4	27.0	267	322	276.1
Bailey Falter	599.8	0.5	27.0	218	323	276.8
Jake McGee	251.1	0.4	27.0	241	324	-72.9
Darwinzon Hernandez	999.0	0.2	27.0	324	325	674
Tucker Davidson	584.9	0.6	26.9	193	326	258.9
Duane Underwood Jr.	999.0	0.1	26.8	417	327	672
Ryan Tepera	581.3	0.4	26.7	246	328	253.3
Matt Harvey	999.0	0.3	26.7	286	329	670
Miguel Castro	600.9	0.1	26.6	372	330	270.9
Johan Oviedo	999.0	0.2	26.3	356	331	668
Austin Warren	999.0	0.5	26.2	231	332	667
Joe Smith	999.0	0.1	26.1	361	333	666
Bryan Garcia	999.0	-0.2	26.1	530	334	665
Nick Sandlin	599.8	0.6	26.1	207	335	264.8
Drew Steckenrider	401.5	0.3	26.1	294	336	65.5
Alexander Wells	999.0	0.4	25.9	276	337	662
Alex Lange	999.0	0.1	25.8	366	338	661
Sean Reid-Foley	999.0	0.0	25.7	438	339	660
Jake Cousins	600.1	0.5	25.6	226	340	260.1
Zach Pop	999.0	0.2	25.5	345	341	658
Logan Allen	600.8	0.1	25.3	418	342	258.8
Sammy Long	596.1	0.3	25.2	307	343	253.1
Demarcus Evans	600.6	0.0	25.2	472	344	256.6
Zack Littell	599.3	0.1	25.2	367	345	254.3
JT Chargois	600.3	0.3	25.2	279	346	254.3
Reynaldo López	557.4	0.4	25.2	257	347	210.4
Jarlín García	600.8	0.1	25.1	411	348	252.8
Dillon Tate	600.9	0.4	25.0	264	349	251.9
Matt Shoemaker	999.0	0.0	25.0	457	350	649
Steven Brault	600.9	0.0	24.8	432	351	249.9
Yency Almonte	999.0	-0.2	24.8	524	352	647
Brett Martin	999.0	0.4	24.7	275	353	646
Craig Stammen	599.4	0.4	24.7	259	354	245.4
Anthony Bass	600.9	0.1	24.6	412	355	245.9
Albert Abreu	999.0	-0.1	24.6	496	356	643
Jeffrey Springs	999.0	0.3	24.5	287	357	642
Austin Davis	999.0	0.2	24.5	320	358	641
Phil Bickford	600.9	0.4	24.5	258	359	241.9
Tyler Clippard	600.8	-0.2	24.5	529	360	240.8
Blake Taylor	999.0	0.2	24.4	353	361	638
Greg Holland	600.2	0.0	24.4	473	362	238.2
Dominic Leone	600.8	0.1	24.2	371	363	237.8
Alex Reyes	395.7	0.2	24.0	335	364	31.7
Ryan Weathers	600.8	0.3	24.0	300	365	235.8
Austin Adams	600.0	0.4	24.0	273	366	234
Josh Rogers	600.8	-0.1	23.8	506	367	233.8
Art Warren	482.3	0.8	23.7	161	368	114.3
Yusmeiro Petit	601.0	-0.3	23.6	531	369	232
Sean Poppen	999.0	0.2	23.6	319	370	629
Tyler Gilbert	599.9	0.4	23.5	255	371	228.9
Adam Morgan	999.0	0.1	23.4	391	372	627
Michael King	600.5	0.5	23.4	220	373	227.5
Chris Martin	600.3	0.4	23.4	277	374	226.3
Dennis Santana	999.0	0.0	23.1	442	375	624
Jacob Webb	999.0	0.0	23.1	447	376	623
Joe Kelly	600.4	0.4	23.1	243	377	223.4
Wander Suero	999.0	0.2	23.0	317	378	621
Chi Chi González	999.0	0.1	22.9	396	379	620
Brad Brach	999.0	0.0	22.8	436	380	619
Kolby Allard	601.0	0.2	22.7	312	381	220
Brandon Kintzler	999.0	0.0	22.6	464	382	617
Trevor Bauer	208.4	1.0	22.5	141	383	-174.6
Richard Rodríguez	556.5	0.1	22.5	414	384	172.5
Touki Toussaint	598.6	0.1	22.3	369	385	213.6
Ralph Garza Jr.	999.0	-0.1	22.1	501	386	613
Jhoulys Chacín	999.0	-0.4	22.0	533	387	612
Tyler Matzek	573.2	0.7	22.0	181	388	185.2
Ryan Thompson	999.0	0.3	22.0	278	389	610
Luis Cessa	588.8	0.3	21.8	299	390	198.8
Andrew Miller	999.0	-0.1	21.8	510	391	608
Ross Detwiler	999.0	0.0	21.7	474	392	607
Taylor Clarke	999.0	0.0	21.5	437	393	606
Deolis Guerra	599.9	0.2	21.3	342	394	205.9
Ben Bowden	999.0	0.1	21.3	395	395	604
Joe Barlow	216.1	0.2	21.1	344	396	-179.9
Andrew Chafin	587.9	0.6	20.8	206	397	190.9
Nick Mears	999.0	0.0	20.8	480	398	601
Kyle Muller	590.8	0.3	20.8	302	399	191.8
Packy Naughton	999.0	0.3	20.7	292	400	599
Drew Smith	999.0	0.1	20.7	360	401	598
Michael Rucker	999.0	0.0	20.4	427	402	597
Derek Holland	999.0	-0.1	20.3	498	403	596
Andres Machado	999.0	-0.1	20.2	513	404	595
Trevor Cahill	999.0	0.2	20.2	313	405	594
Cody Ponce	999.0	0.2	20.2	354	406	593
Aaron Loup	594.4	0.4	20.0	274	407	187.4
Carlos Martínez	596.0	0.0	19.9	477	408	188
Jake Arrieta	999.0	0.0	19.8	481	409	590
Junior Guerra	999.0	0.0	19.8	490	410	589
Bryan Abreu	999.0	0.1	19.7	392	411	588
Charlie Barnes	999.0	0.2	19.7	339	412	587
Jay Jackson	999.0	0.2	19.7	347	413	586
Brandon Bielak	999.0	0.1	19.6	362	414	585
Nabil Crismatt	999.0	0.1	19.5	403	415	584
Wily Peralta	600.8	0.1	19.3	388	416	184.8
Tony Watson	999.0	0.0	19.2	479	417	582
Steve Cishek	595.3	0.0	19.1	439	418	177.3
Jesse Chavez	999.0	0.1	18.6	408	419	580
Mitch White	562.7	0.3	18.6	288	420	142.7
Steven Okert	999.0	0.3	18.6	304	421	578
Adam Cimber	600.6	0.3	18.6	282	422	178.6
Alex Claudio	999.0	0.0	18.5	463	423	576
Joe Mantiply	999.0	0.2	18.3	334	424	575
Trent Thornton	999.0	0.1	18.2	401	425	574
Buck Farmer	999.0	-0.2	18.2	519	426	573
Matt Foster	999.0	0.1	18.1	373	427	572
Domingo Tapia	999.0	0.0	18.0	492	428	571
Jharel Cotton	600.8	0.1	18.0	376	429	171.8
Ryan Borucki	999.0	0.2	17.8	355	430	569
Hunter Strickland	599.9	0.0	17.6	423	431	168.9
Paul Blackburn	999.0	0.2	17.4	315	432	567
Erik Swanson	600.1	0.3	17.1	291	433	167.1
Jordan Holloway	999.0	-0.1	17.1	502	434	565
Ervin Santana	999.0	-0.5	17.0	534	435	564
Blake Parker	999.0	0.1	17.0	393	436	563
Jacob Barnes	999.0	0.1	16.8	381	437	562
Brandon Workman	999.0	-0.2	16.7	521	438	561
Mason Thompson	999.0	0.0	16.7	448	439	560
Paul Campbell	999.0	-0.1	16.6	495	440	559
Michael Feliz	999.0	0.0	16.4	425	441	558
Tyler Zuber	999.0	0.0	16.3	462	442	557
Noé Ramirez	600.9	0.1	16.2	390	443	157.9
Justin Garza	999.0	-0.2	16.2	523	444	555
Yohan Ramirez	599.8	0.1	16.1	413	445	154.8
Miguel Sánchez	999.0	0.0	16.0	484	446	553
Kyle McGowin	999.0	0.2	15.9	321	447	552
Héctor Santiago	999.0	-0.2	15.6	525	448	551
Ryan Burr	999.0	0.0	15.5	440	449	550
Humberto Castellanos	999.0	0.1	15.5	370	450	549
JD Hammer	999.0	0.0	15.4	483	451	548
Phillips Valdez	999.0	-0.1	15.1	493	452	547
Kodi Whitley	999.0	0.2	14.4	350	453	546
Junior Fernández	999.0	0.1	14.3	415	454	545
Danny Coulombe	999.0	0.1	14.3	382	455	544
Drew Pomeranz	582.3	0.1	14.2	383	456	126.3
Rex Brothers	999.0	0.0	14.1	461	457	542
Juan Minaya	999.0	0.1	14.0	397	458	541
Anthony Castro	999.0	0.1	13.6	375	459	540
Tommy Nance	999.0	0.1	13.5	385	460	539
T.J. McFarland	999.0	0.1	13.3	378	461	538
Joel Payamps	999.0	0.3	13.2	285	462	537
Richard Bleier	600.7	0.5	13.1	224	463	137.7
Sean Guenther	999.0	0.2	13.0	323	464	535
Nick Neidert	999.0	-0.2	13.0	517	465	534
Ryne Harper	999.0	0.1	13.0	416	466	533
Anthony Kay	999.0	0.0	12.8	422	467	532
Hoby Milner	999.0	0.1	12.8	405	468	531
Matt Moore	999.0	-0.1	12.8	503	469	530
Dan Winkler	999.0	-0.1	12.7	515	470	529
Humberto Mejía	999.0	0.0	12.7	453	471	528
Matt Peacock	999.0	0.0	12.5	488	472	527
Brett de Geus	999.0	0.0	12.4	475	473	526
Aaron Sanchez	600.6	0.0	12.4	451	474	126.6
José Ruiz	999.0	0.2	12.4	357	475	524
Wade LeBlanc	999.0	0.2	12.4	358	476	523
Chris Mazza	999.0	0.3	12.3	309	477	522
Shane Greene	999.0	-0.1	12.1	499	478	521
Keynan Middleton	999.0	-0.1	12.1	500	479	520
Danny Duffy	587.9	0.2	12.0	318	480	107.9
Cody Poteet	999.0	0.1	11.7	407	481	518
José Álvarez	600.2	0.2	11.7	359	482	118.2
Chase Anderson	999.0	0.0	11.6	491	483	516
Louis Head	600.5	0.3	11.6	311	484	116.5
Kohei Arihara	999.0	0.0	11.5	449	485	514
Jordan Sheffield	999.0	-0.2	11.4	520	486	513
Ashton Goudeau	999.0	-0.2	10.6	527	487	512
Victor González	999.0	0.1	10.6	365	488	511
Yennsy Díaz	999.0	-0.1	10.4	516	489	510
Braxton Garrett	999.0	0.0	10.4	428	490	509
Chasen Shreve	999.0	-0.1	10.1	509	491	508
César Valdez	600.3	0.0	10.0	431	492	108.3
Sam Selman	999.0	0.0	9.5	444	493	506
Cam Bedrosian	999.0	0.0	9.1	458	494	505
Dustin May	559.5	0.3	9.1	298	495	64.5
Sam Clay	999.0	0.0	8.7	430	496	503
Caleb Baragar	999.0	-0.2	8.7	522	497	502
Dillon Maples	999.0	0.1	8.6	409	498	501
Daniel Ponce de Leon	600.7	0.0	8.5	456	499	101.7
Erasmo Ramírez	999.0	-0.2	7.9	526	500	499
Edwin Uceta	999.0	0.0	7.9	455	501	498
Miguel Diaz	999.0	-0.1	7.8	494	502	497
Marcos Diplán	999.0	0.0	7.4	459	503	496
Enyel De Los Santos	999.0	0.1	7.3	404	504	495
Tayler Saucedo	999.0	0.1	7.1	399	505	494
Robert Gsellman	999.0	-0.2	7.1	518	506	493
Wes Benjamin	999.0	-0.1	7.0	504	507	492
Thomas Eshelman	999.0	0.0	6.8	476	508	491
Reiss Knehr	999.0	0.0	6.7	446	509	490
Jake Faria	999.0	0.0	6.7	469	510	489
Luke Farrell	999.0	0.0	6.3	489	511	488
Robert Dugger	999.0	0.0	6.2	478	512	487
Casey Sadler	600.5	0.3	5.8	280	513	87.5
Shaun Anderson	999.0	0.0	5.7	445	514	485
Drew Hutchison	999.0	-0.1	5.5	497	515	484
Sean Nolin	999.0	0.0	5.3	485	516	483
Adrian Sampson	999.0	-0.1	4.9	508	517	482
Tyler Glasnow	598.8	0.2	4.9	314	518	80.8
Shawn Armstrong	999.0	0.0	4.6	454	519	480
Kevin Ginkel	999.0	0.0	4.3	470	520	479
Carson Fulmer	999.0	0.0	4.1	482	521	478
Edgar Santana	999.0	0.0	4.1	452	522	477
Kyle Crick	999.0	0.0	4.0	465	523	476
Daniel Castano	999.0	0.0	3.7	486	524	475
Kenta Maeda	600.7	0.1	3.3	384	525	75.7
John Gant	600.8	0.0	3.3	435	526	74.8
Chris Ellis	600.8	0.1	2.5	406	527	73.8

NA
NA
NA
NA
NA
NA

---