Pokemon Catch em All

Loading, Preview and Cleaning Data
Pokemon Starter Pack
Where is Pikachu?
- Pika Pika Pika
- Electric Types
Pokemon BMI
- Body Masses
- Correlations
Legendary Pokemon

library(ggplot2)
library(corrplot)

## corrplot 0.84 loaded

library(RColorBrewer)

Greetings trainers,
In this page(s) i will try to inform you things about Pokemon

Pokemon is …

Nah, i don’t have any obligation to explain you what Pokemon is, just Google it

What insight we will try to get :
1. What is 6 best Pokemon to start your journey with
2. Where is Pikachu?
3. Pokemon BMI, is it any correlation with attack/defense and versality?
4. What is the strongest legendary Pokemon? Each type

Loading, Preview and Cleaning Data

First we have to load the data

pokemon <- read.csv("pokemonfull.csv")
head(pokemon)

The capture_rate variable has factor type and we want to convert it to numeric

pokemon$capture_rate <- as.numeric(pokemon$capture_rate)

We’re going to separate the common and legendary Pokemons

common.p <- pokemon[pokemon$is_legendary == 0,]
legend.p <- pokemon[pokemon$is_legendary == 1,]

Pokemon Starter Pack

There are many categories those defines a Pokemon categorized as starter pack
1. Capture Rate
2. Most Common Type
3. Versality based of damage taken
4. Attack/Defense ratio

Capture Rate

Analytics

The minimum value of capture rate is 1 and the maximum is 34. Based on those values we make the function

# capture rate function
ccr <- function(x){
  if(x <= 8){
    x <- "Hard"
  }else if(x > 8 & x <= 26){
    x <- "Moderate"
  }else{
    x <- "Easy"
  }
}

common.p$capture_prob <- as.factor(sapply(common.p$capture_rate,ccr))

# put probs in objects
common.p.easy <- common.p[common.p$capture_prob == "Easy",]

Graphics

pie(table(common.p$capture_prob))

As we see from the pie above, we can see that most of the Pokemon not that easy to catch, but not that hard to catch
just like girlfriends ;)

common.p.30 <- head(common.p,30)

ggplot(common.p.30, aes(x = reorder(name, capture_rate), y = capture_rate))+
  geom_col(aes(fill = capture_rate))+
  coord_flip()+
  scale_fill_continuous(high = "#811e09", low = "#fad61d")+
  theme(plot.title = element_text(hjust = 0.5,size = 16),
        panel.background = element_rect(fill = "white"),
        plot.background = element_rect(fill = "white"),
        legend.position = "none")+
  labs(x = NULL, y = NULL,title = "Capture Rate")+
  scale_y_continuous(limits = c(0,40),
                     breaks = seq(0,40,5))

The easiest Pokemon to catch in our list is Sandslash

Most Common Type

To help our new beginning journey we have to know which type of Pokemon are most common
It will help us to identify the solution while facing the enemy

common.type <- as.data.frame(table(common.p.easy$type1))
common.type <- common.type[order(common.type$Freq,decreasing = T),]
head(common.type,6)

From above we know what are the most common types :
* water
* normal
* bug
* grass
* ground
* poison

We only use these 6 kinds of type

common.p.easy <- common.p.easy[common.p.easy$type1 %in% c("water","normal","bug","grass","ground","poison"),]

Versality

Versality means how Pokemon can defend theirselves from many type of attacks
We divide it by 3 variables : Strong, Moderate, Weak

# damage taken by all elements
common.p.easy$vers <- rowSums(common.p.easy[,c(2:19)])/18

# versality function
vr <- function(y){
  if(y <= 0.3472223){
    x <- "Strong"
  }else if(y > 0.3472223 & y <= 1.041667){
    x <- "Moderate"
  }else{
    x <- "Weak"
  }
}

common.p.easy$vr_prob <- as.factor(sapply(common.p.easy$vers,vr))

There is no Pokemon in “Strong” term, so we use “Moderate”

common.p.easy <- common.p.easy[common.p.easy$vr_prob == "Moderate",]

Attack and Defense Ratio

Attack and Defense are important, so we’ll calculate the balance of Attack and Defense

The formula is
((Attack/Defense)+(Defense/Attack))/2

Because Attack is as important as Defense

# attack/defense
common.p.easy$ad <- as.numeric(((common.p.easy$attack/common.p.easy$defense)+(common.p.easy$defense/common.p.easy$attack))/2)

agg.ad <- aggregate(formula = ad ~ name + type1 + height_m + weight_kg, data = common.p.easy, FUN = max)
agg.ad <- agg.ad[order(agg.ad$ad,decreasing = T),]
head(agg.ad,6)

Here it is !!! The best Pokemon for your new beginning journey !!!

And here it is, if you want to bring each type

common.water <- head(agg.ad[agg.ad$type1 == "water",],1)
common.normal <- head(agg.ad[agg.ad$type1 == "normal",],1)
common.bug <- head(agg.ad[agg.ad$type1 == "bug",],1)
common.grass <- head(agg.ad[agg.ad$type1 == "grass",],1)
common.ground <- head(agg.ad[agg.ad$type1 == "ground",],1)
common.poison <- head(agg.ad[agg.ad$type1 == "poison",],1)

common.element <- rbind(common.water,common.normal,common.bug,common.grass,common.ground,common.poison)
common.element

Where is Pikachu?

Pika Pika Pika

We’ll try to find Pikachu and define his position among others

pikachu <- common.p[common.p$name == "Pikachu",]
pikachu[,c("name","capture_prob","attack","defense","type1")]

# Pikachu's versality
pikachu$vers <- rowSums(pikachu[,c(2:19)])/18
pikachu$vr_prob <- as.factor(sapply(pikachu$vers,vr))

# Pikachu's att/def
pikachu$ad <- as.numeric(((pikachu$attack/pikachu$defense)+(pikachu$defense/pikachu$attack))/2)

#print Pikachu
pikachu[,c("name","capture_prob","ad","vr_prob","type1")]

Electric Types

Analytics

Because of Pikachu is an electric, so we will compare him with his comrades

# Electric Pokemon with Moderate capture_prob
electric <- common.p[common.p$capture_prob == "Moderate" & common.p$type1 == "electric",]

# Electric Pokemon's versality
electric$vers <- rowSums(electric[,c(2:19)])/18
electric$vr_prob <- as.factor(sapply(electric$vers,vr))

# Electric Pokemon's att/def
electric$ad <- as.numeric(((electric$attack/electric$defense)+(electric$defense/electric$attack))/2)

# Print electric
electric[,c("name","capture_prob","ad","vr_prob","type1")]

#Aggregate Electric
agg.electric <- aggregate(formula = ad ~ name + type1 + height_m + weight_kg + vr_prob, data = electric, FUN = max)
agg.electric <- agg.electric[order(agg.electric$ad,decreasing = T),]
agg.electric

Graphics

ggplot(electric, aes(x = reorder(name, ad), y = ad))+
  geom_col(aes(fill = ad))+
  coord_flip()+
  scale_fill_continuous(high = "#fad61d", low = "#fad61d")+
  geom_col(data = pikachu, aes(fill = ad), fill = "#f62d14")+
  theme(plot.title = element_text(hjust = 0.5,size = 16),
        panel.background = element_rect(fill = "white"),
        plot.background = element_rect(fill = "white"),
        legend.position = "none")+
  labs(x = NULL, y = NULL,title = "Electric Power")+
  scale_y_continuous(limits = c(0,2),
                     breaks = seq(0,2,0.1))

statpack.plot <- head(agg.ad,6)
statpack.plot <- statpack.plot[,c("name","ad")]
pikachu.plot <- pikachu[,c("name","ad")]
spp.plot <- rbind(statpack.plot,pikachu.plot)

ggplot(spp.plot, aes(x = reorder(name, ad), y = ad))+
  geom_col(aes(fill = ad))+
  coord_flip()+
  scale_fill_continuous(high = "#fad61d", low = "#fad61d")+
  geom_col(data = pikachu.plot, aes(fill = ad), fill = "#f62d14")+
  theme(plot.title = element_text(hjust = 0.5,size = 16),
        panel.background = element_rect(fill = "white"),
        plot.background = element_rect(fill = "white"),
                legend.position = "none")+
  labs(x = NULL, y = NULL,title = "Where Is Pikachu",subtitle = "Among Top 6 Start Pack")+
  scale_y_continuous(limits = c(0,2),
                     breaks = seq(0,2,0.1))

common.element.plot <- common.element[,c("name","ad")]
pikachu.plot <- pikachu[,c("name","ad")]
cep.plot <- rbind(common.element.plot,pikachu.plot)

ggplot(cep.plot, aes(x = reorder(name, ad), y = ad))+
  geom_col(aes(fill = ad))+
  coord_flip()+
  scale_fill_continuous(high = "#fad61d", low = "#fad61d")+
  geom_col(data = pikachu.plot, aes(fill = ad), fill = "#f62d14")+
  theme(plot.title = element_text(hjust = 0.5,size = 16),
        panel.background = element_rect(fill = "white"),
        plot.background = element_rect(fill = "white"),
                legend.position = "none")+
  labs(x = NULL, y = NULL,title = "Where Is Pikachu",subtitle = "Among Top 6 based on type")+
  scale_y_continuous(limits = c(0,2),
                     breaks = seq(0,2,0.1))

Based on the graphic, there is no reason that Pikachu is the strongest, the easiest to catch, the fittest or wsoe
Then, why he so famous?
BECAUSE HE IS CUTE AS HELL!!!
and Ryan Reynolds casts him too :3

Pokemon BMI

Body Masses

Analytics

Not only human have the BMI (Body Mass Index), Pokemon have to
The formula is Weight (Kg) / Height^2 (m)
So let’s begin

cp.bmi <- common.p
cp.bmi$vers <- rowSums(cp.bmi[,c(2:19)])/18
cp.bmi$vr_prob <- as.factor(sapply(cp.bmi$vers,vr))
cp.bmi$ad <- as.numeric(((cp.bmi$attack/cp.bmi$defense)+(cp.bmi$defense/cp.bmi$attack))/2)
cp.bmi$bmi <- as.numeric((cp.bmi$weight_kg)/(cp.bmi$height_m^2))
cp.bmi[,c("name","vr_prob","ad","bmi")]

Oopsie! There are many NA in our data, it means the data is not that good
What we have supposed to do? stop here?
Heck no!
We’ll try to remove those NAs and start over

cp.bmi <- na.omit(cp.bmi)
cp.bmi[,c("name","vr_prob","ad","bmi")]

Here are the categories according to BMI value
Underweight : < 18.5
Normal weight : 18.5–24.9
Overweight : 25–29.9
Obesity : BMI of 30 or greater

# BMI function

bmi <- function(b){
  if(b < 18.5){
    b <- "Underweight"
  }else if(b >= 18.5 & b <= 24.9){
    b <- "Normal Weight"
  }else if(b >= 25 & b <= 29.9){
    b <- "Overweight"
  }else{
    b <- "Obesity"
  }
}

cp.bmi$bmi_cat <- as.factor(sapply(cp.bmi$bmi,bmi))
cp.bmi[,c("name","vr_prob","ad","bmi","bmi_cat")]

# Put each category in objects
underweight <- cp.bmi[cp.bmi$bmi_cat == "Underweight",]
normalweight <- cp.bmi[cp.bmi$bmi_cat == "Normal Weight",]
overweight <- cp.bmi[cp.bmi$bmi_cat == "Overweight",]
obesity <- cp.bmi[cp.bmi$bmi_cat == "Obesity",]

cp.bmi[order(cp.bmi$bmi,decreasing = T), c("name","bmi","bmi_cat","type1")]

Seems not normal right? How come a creature has 375 BMI?
It means you’re looking at a 1x1 m creature with 375 Kg weight!!
But this is Pokemon world right? We are as data scientist must put the "normality’ aside LOL

Graphics

pie(table(cp.bmi$bmi_cat))

Correlations

Analytics

We want to see the correlation between the variables
* BMI
* vers
* ad
* speed
* height
* weight
* attack
* defense
* capture_rate

bmicor <- cp.bmi[,c("bmi","vers","ad","hp","speed","height_m","weight_kg","attack","defense","capture_rate")]
bmicor <-cor(bmicor)
bmicor

##                      bmi         vers           ad          hp       speed
## bmi           1.00000000  0.022457688  0.037711265 -0.01555890 -0.22415104
## vers          0.02245769  1.000000000  0.009666441  0.01488890 -0.06326199
## ad            0.03771126  0.009666441  1.000000000 -0.10566317 -0.09301751
## hp           -0.01555890  0.014888900 -0.105663170  1.00000000  0.09498438
## speed        -0.22415104 -0.063261993 -0.093017508  0.09498438  1.00000000
## height_m     -0.13810360  0.049758668  0.003418695  0.40097680  0.16051648
## weight_kg     0.26360237  0.117750362 -0.001992684  0.41010271 -0.02282505
## attack        0.02775554  0.024497092 -0.075502625  0.36697855  0.30158556
## defense       0.14730570  0.054982747  0.196888458  0.19199004 -0.06169946
## capture_rate -0.04394980  0.024202017 -0.035682234  0.24987589  0.24000998
##                  height_m    weight_kg      attack     defense
## bmi          -0.138103601  0.263602372  0.02775554  0.14730570
## vers          0.049758668  0.117750362  0.02449709  0.05498275
## ad            0.003418695 -0.001992684 -0.07550262  0.19688846
## hp            0.400976803  0.410102710  0.36697855  0.19199004
## speed         0.160516484 -0.022825047  0.30158556 -0.06169946
## height_m      1.000000000  0.600558255  0.37242702  0.35986364
## weight_kg     0.600558255  1.000000000  0.41920700  0.45555563
## attack        0.372427017  0.419206995  1.00000000  0.45167441
## defense       0.359863641  0.455555629  0.45167441  1.00000000
## capture_rate  0.258003585  0.209173878  0.34412752  0.31038775
##              capture_rate
## bmi           -0.04394980
## vers           0.02420202
## ad            -0.03568223
## hp             0.24987589
## speed          0.24000998
## height_m       0.25800359
## weight_kg      0.20917388
## attack         0.34412752
## defense        0.31038775
## capture_rate   1.00000000

Graphics

bmicor <- cp.bmi[,c("bmi","vers","ad","hp","speed","height_m","weight_kg","attack","defense","capture_rate")]
bmicor <-cor(bmicor)
corrplot(bmicor, type="upper", order="hclust", method = "number",
         col=brewer.pal(n=8, name="RdYlBu"))

The highest positive correlation is height_m and weight_kg = 0.6
The highest negative correlation is bmi and speed = -0.22

ggplot(data = cp.bmi, aes(x = capture_prob, y = bmi_cat))+
  geom_jitter(aes(size = ad, col = vers))+
  facet_grid( ~ as.factor(vr_prob),labeller = label_value)+
  theme(legend.position = "right")+
  labs(x = "Capture Probability", y = "BMI Category")

Legendary Pokemon

Legendary Pokemons are believed as the mighty Pokemon, and some say they have godlike ability

Based on Attack

Analytics

legend.p.att <- head(legend.p[order(legend.p$attack,decreasing = T),],12)
legend.p.att[,c("name","attack")]

Graphics

ggplot(legend.p.att, aes(x = reorder(name, attack), y = attack))+
  geom_col(aes(fill = attack))+
  coord_flip()+
  scale_fill_continuous(high = "#811e09", low = "#fad61d")+
  theme(plot.title = element_text(hjust = 0.5,size = 16),
        panel.background = element_rect(fill = "white"),
        plot.background = element_rect(fill = "white"),
        legend.position = "none")+
  labs(x = NULL, y = NULL,title = "Strongest Pokemon based on Attack")

Based on Defense

Analytics

legend.p.def <- head(legend.p[order(legend.p$defense,decreasing = T),],12)
legend.p.def[,c("name","defense")]

Graphics

ggplot(legend.p.def, aes(x = reorder(name, defense), y = defense))+
  geom_col(aes(fill = defense))+
  coord_flip()+
  scale_fill_continuous(high = "#811e09", low = "#fad61d")+
  theme(plot.title = element_text(hjust = 0.5,size = 16),
        panel.background = element_rect(fill = "white"),
        plot.background = element_rect(fill = "white"),
        legend.position = "none")+
  labs(x = NULL, y = NULL,title = "Strongest Pokemon based on Defense")

Based on Attack Defense Ratio

Analytics

legend.p$ad <- as.numeric(((legend.p$attack/legend.p$defense)+(legend.p$defense/legend.p$attack))/2)
legend.p.ad <- head(legend.p[order(legend.p$ad,decreasing = T),],12)
legend.p.ad[,c("name","ad")]

Graphics

ggplot(legend.p.ad, aes(x = reorder(name, ad), y = ad))+
  geom_col(aes(fill = ad))+
  coord_flip()+
  scale_fill_continuous(high = "#811e09", low = "#fad61d")+
  theme(plot.title = element_text(hjust = 0.5,size = 16),
        panel.background = element_rect(fill = "white"),
        plot.background = element_rect(fill = "white"),
        legend.position = "none")+
  labs(x = NULL, y = NULL,title = "Strongest Pokemon based on Attack Defense Ratio")

Attack Defense Ratio and Versality

Analytics

legend.p.ad$ver <- rowSums(legend.p.ad[,c(2:19)])/18
legend.p.ad$vr_prob <- as.factor(sapply(legend.p.ad$ver,vr))
legend.p.ad[,c("name","ver","vr_prob")]

Graphics

ggplot(data = legend.p.ad, aes(x = attack, y = defense))+
  geom_jitter(aes(size = ad, col = ver))+
  facet_wrap( ~ as.factor(name))+
  theme(legend.position = "right")