PUBG.Rmd

---
title: "PUBG"
author: "Alex Pearson"
date: "11/12/2018"
output: pdf_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
library(ggplot2)
library(reshape2)
library(tidyr)
library(dplyr)
library(scales)
setwd("/Users/mariette/Desktop/PUBG_Project/")
PUBG_full <- read.csv("data/train_V2.csv", header=TRUE, sep=",")
PUBG_test <- read.csv("data/test_V2.csv", header=TRUE, sep=",")
```

## Process the Data

We will get a lot of different

```{r}
# Full Data
full_response <- PUBG_full$winPlacePerc
full_data_matrix <- PUBG_full[ , !(names(PUBG_full) %in% c("winPlacePerc"))]

duo_fpp_data_with_response <- PUBG_full[which(PUBG_full$matchType=='duo-fpp'),] 
sorted_duo_fpp_with_response <- duo_fpp_data_with_response[order(duo_fpp_data_with_response$matchId,duo_fpp_data_with_response$groupId), ]
match_1_with_response <- sorted_duo_fpp_with_response[which(sorted_duo_fpp_with_response$matchId=="0006eb8c17708d" ),] 
match_1_groups_with_response <- unique(match_1_with_response$groupId)
head(match_1_with_response)
length(match_1_groups_with_response)
```


```{r}
subset_and_order_full_data_by_game_type <- function(game_type, full_data = PUBG_full) {
  subset_data <- full_data[which(full_data$matchType==game_type),] 
  ordered_data <- subset_data[order(subset_data$matchId,subset_data$groupId), ]
  return(ordered_data)
}
length(unique(PUBG_full$matchId))
length(unique(PUBG_full$matchId)) + length(unique(PUBG_test$matchId))
sorted_solo_full_data <- subset_and_order_full_data_by_game_type("solo")

PUBG_full_ordered <- PUBG_full[order(PUBG_full$matchId,PUBG_full$groupId), ]
head(PUBG_full_ordered)

```


```{r}
# Solo Data
solo_data_matrix <- full_data_matrix[which(full_data_matrix$matchType=='solo'),] 
solo_entries <- PUBG_full[which(PUBG_full$matchType=='solo') , ]
solo_response <- solo_entries$winPlacePerc

# Solo FPP Data
solo_fpp_data_matrix <- full_data_matrix[which(full_data_matrix$matchType=='solo-fpp'),] 
solo_fpp_entries <- PUBG_full[which(PUBG_full$matchType=='solo-fpp') , ]
solo_fpp_response <- solo_fpp_entries$winPlacePerc

# Duo Data
duo_data_matrix <- full_data_matrix[which(full_data_matrix$matchType=='duo'),] 
duo_entries <- PUBG_full[which(PUBG_full$matchType=='duo') , ]
duo_response <- duo_entries$winPlacePerc

# Duo FPP Data
duo_fpp_data_matrix <- full_data_matrix[which(full_data_matrix$matchType=='duo-fpp'),] 
duo_fpp_entries <- PUBG_full[which(PUBG_full$matchType=='duo-fpp') , ]
duo_fpp_response <- duo_fpp_entries$winPlacePerc

# squad Data
squad_data_matrix <- full_data_matrix[which(full_data_matrix$matchType=='squad'),] 
squad_entries <- PUBG_full[which(PUBG_full$matchType=='squad') , ]
squad_response <- squad_entries$winPlacePerc

# squad FPP Data
squad_fpp_data_matrix <- full_data_matrix[which(full_data_matrix$matchType=='squad-fpp'),] 
squad_fpp_entries <- PUBG_full[which(PUBG_full$matchType=='squad-fpp') , ]
squad_fpp_response <- squad_fpp_entries$winPlacePerc
```


```{r}
sorted_duo_fpp <- duo_fpp_data_matrix[order(duo_fpp_data_matrix$matchId,duo_fpp_data_matrix$groupId), ]
match_1 <- sorted_duo_fpp[which(sorted_duo_fpp$matchId=="0006eb8c17708d" ),] 
match_1_groups <- unique(match_1$groupId)
head(match_1_groups)
head(match_1)
```


```{r}
#pivoted_data <- summarise(group_by(PUBG_full,groupId), Number_Players=mean(numGroups))
#head(pivoted_data)
```



```{r}
groups_and_duration <- select(PUBG_full_ordered,matchId,numGroups,matchDuration)
num_Groups_vs_match_duration <- unique(groups_and_duration)
length(num_Groups_vs_match_duration$numGroups) == length(unique(PUBG_full$matchId))
linear_model_of_num_groups_v_match_duration <- lm(num_Groups_vs_match_duration$matchDuration ~ num_Groups_vs_match_duration$numGroups)
plot(linear_model_of_num_groups_v_match_duration)

plot_num_groups_vs_duration <- function(ordered_data = PUBG_full_ordered) {
  groups_and_duration <- select(ordered_data,matchId,numGroups,matchDuration)
  num_Groups_vs_match_duration <- unique(groups_and_duration)
  #plot(num_Groups_vs_match_duration$numGroups,num_Groups_vs_match_duration$matchDuration)
  ggplot(num_Groups_vs_match_duration, aes(x=numGroups, y=matchDuration)) +
  geom_hline(yintercept=1600, linetype="dashed", color = "red") +
  geom_point()
}

subset_and_order_full_data_by_game_type <- function(game_type, full_data = PUBG_full) {
  subset_data <- full_data[which(full_data$matchType==game_type),] 
  ordered_data <- subset_data[order(subset_data$matchId,subset_data$groupId), ]
  return(ordered_data)
}


plot_num_groups_vs_duration(subset_and_order_full_data_by_game_type("solo"))
plot_num_groups_vs_duration(subset_and_order_full_data_by_game_type("solo-fpp"))
plot_num_groups_vs_duration(subset_and_order_full_data_by_game_type("duo"))
plot_num_groups_vs_duration(subset_and_order_full_data_by_game_type("duo-fpp"))
plot_num_groups_vs_duration(subset_and_order_full_data_by_game_type("squad"))
plot_num_groups_vs_duration(subset_and_order_full_data_by_game_type("squad-fpp"))
```

Main Question: Can we use regression and clustering to figure out the best strategies for placing highly in PUBG?  In particular we want to create

## EDA

TODO: write motivation for EDA

First lets try and get a feel for the different types of PUBG game stats.  

```{r}
all_types <-table(PUBG_full$matchType)
# Over 99% of the data falls into 6 main types
main_types <- all_types[c("solo","solo-fpp","duo","duo-fpp","squad","squad-fpp")]

df_types <- data.frame(Rules=c('Original','FPP Only'), Alone =c(main_types[1],main_types[2]), Duo=c(main_types[3],main_types[4]), Squad=c(main_types[5],main_types[6]))
dat.g <- gather(df_types, type, value, -Rules)

# Stupid hack switching Solo to Alone in order to order the bar chart correctly
ggplot(dat.g, aes(x = type, y = value )) + 
  geom_bar(aes(fill = Rules), stat = "identity", position = "dodge") +
  labs(x = "Group Size", y = "Number of PGaPS") +
  scale_y_continuous(label=comma) + 
  ggtitle("Number of PGaPS for Different Types of PUBG Games") +
  theme(plot.title = element_text(hjust = 0.5)) 
```



```{r}
#dat <- data.frame(country=c('USA','Brazil','Ghana','England','Australia'), Stabbing=c(15,10,9,6,7), #Accidents=c(20,25,21,28,15), Suicide=c(3,10,7,8,6))

#dat.g <- gather(dat, type, value, -country)
#ggplot(dat.g, aes(type, value)) + 
 # geom_bar(aes(fill = country), stat = "identity", position = "dodge")

```

Correlation Plot
```{r}
library(corrplot)
test <- PUBG_full[,!(names(PUBG_full) %in% drop_categories)]
M <- cor(test[sapply(test, function(x) !is.factor(x))])
corrplot(M, method="square", type="upper", tl.col="black", tl.srt=45, diag=FALSE, tl.cex = 0.5)
```

Sub Correlation Plot
```{r}
total_distance_traveled <- function(df, method1, method2, method3){
df <- df[method1+"Distance"] + df[method2+"Distance"] + method[method3+"Distance"]
return(df)
}

healsAndBoosts <- function(df, method1, method2, method3){
df[] <- df[method1+"Distance"] + df[method2+"Distance"] + method[method3+"Distance"]
return()
}
```