Editing and Removing Variables:

Both datasets have quite a few variables so I removed and narrowed down the variables to the ones I found relevant to my project purpose. This was done by creating a new dataframe and selecting only variables that I needed.

library(tidyverse)
library(dplyr)
LoL_Champions_RawData <- read.csv("LoL_Champions.csv")
LoL_Champions <- LoL_Champions_RawData %>% select(id, title, 
    tags, stats.hp, stats.armor, stats.spellblock, stats.attackrange, 
    stats.attackdamage, stats.attackspeed)

TFT_Champions_RawData <- read.csv("TFT_Champions.csv")
TFT_Champions <- TFT_Champions_RawData %>% select(name, health, 
    defense, attack, attack_range, speed_of_attack, origin, class)

For the rest of the project, variables that are game specific (i.e base Hp) have the game abbreviation in the front(i.e LoL.Hp for LoL base health and TFT.Hp for TFT). Shared traits (i.e Champion name) remain as is.

LoL_Champions <- LoL_Champions %>% select(Champion = id, Title = title, 
    LoL.Class = tags, LoL.Hp = stats.hp, LoL.Armor = stats.armor, 
    LoL.MagicArmor = stats.spellblock, LoL.Range = stats.attackrange, 
    LoL.Damage = stats.attackdamage, LoL.AttackSpeed = stats.attackspeed)

TFT_Champions <- TFT_Champions %>% select(Champion = name, TFT.Hp = health, 
    TFT.TotalArmor = defense, TFT.Damage = attack, TFT.Range = attack_range, 
    TFT.AttackSpeed = speed_of_attack, TFT.Origin = origin, TFT.Class = class)

This was done using select() to create a new data frame with the selected variables being set to the new names

Seperating Classes:

In both datasets, the Class columns have cases where there are two Classes. Using tidyr functions, I tidied the datasets so that there was a new column titled second Class.

LoL_Champions <- LoL_Champions %>% separate(LoL.Class, c("LoL.Class1", 
    "LoL.Class2"), ",")

TFT_Champions <- TFT_Champions %>% separate(TFT.Class, c("TFT.Class1", 
    "TFT.Class2"), ",")

In order to create a second column for the second class, I separated the class column at the comma using separate() and created a second one. I did this for both datasets.

Pivoting :

My data was already tidy for the purpose of my project. However, to practice pivoting I decided to look into two variables within the TFT dataset. In this case, I wanted to see the if, at a glance, there was any association between Hp and Origin. Origin’s purpose in TFT is that when you have more than one champion on your team from said Origin, those champions are given a buff. With that in mind, perhaps the Origin of each champ was determined using the base stats to make the buff more effective. At a glance, there does not appear to be any association between Hp and Origin. Perhaps, Origin is just that and is determined based on the lore/background of each champion. Or perhaps Origin may be based off of other stats.

TFT_Origin <- TFT_Champions %>% select(Champion, TFT.Origin, 
    TFT.Hp)
TFT_Origin <- TFT_Origin %>% pivot_wider(names_from = TFT.Origin, 
    values_from = TFT.Hp)
head(TFT_Origin)

## # A tibble: 6 x 11
##   Champion `Space Pirate` `Star Guardian` Celestial `Dark Star` `Mech-Pilot`
##   <fct>             <int>           <int>     <int>       <int>        <int>
## 1 Gangpla…           1000              NA        NA          NA           NA
## 2 Graves              650              NA        NA          NA           NA
## 3 Neeko                NA             800        NA          NA           NA
## 4 Darius              750              NA        NA          NA           NA
## 5 Rakan                NA              NA       600          NA           NA
## 6 Lux                  NA              NA        NA         600           NA
## # … with 5 more variables: Cybernetic <int>, Rebel <int>, Valkyrie <int>,
## #   Void <int>, Chrono <int>

After selecting for the variables I want to look into, I pivoted the set wider with the names of each column being taken from the Origin column and the values under each being Hp.

TFT_Origin <- TFT_Origin %>% pivot_longer(2:11, names_to = "TFT.Origin", 
    values_to = "TFT.Hp", values_drop_na = T)
head(TFT_Origin)

## # A tibble: 6 x 3
##   Champion  TFT.Origin    TFT.Hp
##   <fct>     <chr>          <int>
## 1 Gangplank Space Pirate    1000
## 2 Graves    Space Pirate     650
## 3 Neeko     Star Guardian    800
## 4 Darius    Space Pirate     750
## 5 Rakan     Celestial        600
## 6 Lux       Dark Star        600

With the dataset wide like this, there is many NA values due to each champion only having one Origin. So I pivoted the dataset longer in order to return it to its original state. Also, I used values_drop_na, so that the NA values would be dropped.

Mutate: Total Armor and DPS

In LoL, champions have two armor types, magic and regular. Instead of having two separate variables, I wanted just one column for Total Armor. Also, using the data for Attack Speed and Damage, I wanted to calculate the DPS or Damage Per Second each champion has.

Joined_Champions <- Joined_Champions %>% mutate(LoL.TotalArmor = LoL.Armor + 
    LoL.MagicArmor)

Joined_Champions <- Joined_Champions %>% mutate(LoL.DamagePerSec = LoL.Damage * 
    LoL.AttackSpeed)
Joined_Champions <- Joined_Champions %>% mutate(TFT.DamagePerSec = TFT.Damage * 
    TFT.AttackSpeed)
Joined_Champions <- Joined_Champions %>% select(Champion, Title, 
    LoL.Class1, LoL.Class2, LoL.Hp, LoL.TotalArmor, LoL.Range, 
    LoL.Damage, LoL.AttackSpeed, LoL.DamagePerSec, TFT.Origin, 
    TFT.Class1, TFT.Class2, TFT.Hp, TFT.TotalArmor, TFT.Range, 
    TFT.Damage, TFT.AttackSpeed, TFT.DamagePerSec)

In order to create these new columns. I used mutate() which took the values from preexisting columns and applied a function to them in order to create the new ones. For LoL.TotalArmor, the values of LoL.MagicArmor and LoL.Armor were added together. For the DamagePerSec for both LoL and TFT, the values in these columns were the products of Attack Speed and Damage. Finally, I used select() to edit the dataset to get rid of LoL.MagicArmor and LoL.Armor, which were no longer needed.

Summary: Hp and TotalArmor

In LoL, the primary class of a champion is important when determining whether it suits the needs of the player. For a player who likes to be defensive and be a shield, high Hp and Total Armor is important. The champion with the highest Hp was Kayle, a fighter. The Champion with the lowest Hp was Caitlyn, a marksman. The highest Total Armor was Leona, a Tank. The lowest Total Armor was Lux, a mage.

library(kableExtra)
Joined_Champions %>% summarise(max(LoL.Hp), min(LoL.Hp), max(LoL.TotalArmor), 
    min(LoL.TotalArmor))

##   max(LoL.Hp) min(LoL.Hp) max(LoL.TotalArmor) min(LoL.TotalArmor)
## 1         600         481                79.1               48.72

Joined_Champions %>% filter(LoL.Hp == max(LoL.Hp)) %>% select(Champion)

##   Champion
## 1    Kayle

Joined_Champions %>% filter(LoL.Hp == min(LoL.Hp)) %>% select(Champion)

##   Champion
## 1  Caitlyn

Joined_Champions %>% filter(LoL.TotalArmor == max(LoL.TotalArmor)) %>% 
    select(Champion)

##   Champion
## 1    Leona

Joined_Champions %>% filter(LoL.TotalArmor == min(LoL.TotalArmor)) %>% 
    select(Champion)

##   Champion
## 1      Lux

Joined_Champions %>% group_by(LoL.Class1) %>% summarise(max.LoL.Hp = max(LoL.Hp)) %>% 
    arrange(desc(max.LoL.Hp)) %>% kbl() %>% kable_styling(bootstrap_options = "striped", 
    full_width = F, position = "center")

Joined_Champions %>% group_by(LoL.Class1) %>% summarize(min.LoL.Hp = min(LoL.Hp)) %>% 
    arrange(min.LoL.Hp) %>% kbl() %>% kable_styling(bootstrap_options = "striped", 
    full_width = F, position = "center")

Joined_Champions %>% group_by(LoL.Class1) %>% summarize(max.LoL.TotalArmor = max(LoL.TotalArmor)) %>% 
    arrange(max.LoL.TotalArmor) %>% kbl() %>% kable_styling(bootstrap_options = "striped", 
    full_width = F, position = "center")

Joined_Champions %>% group_by(LoL.Class1) %>% summarise(min.LoL.TotalArmor = min(LoL.TotalArmor)) %>% 
    arrange(min.LoL.TotalArmor) %>% kbl() %>% kable_styling(bootstrap_options = "striped", 
    full_width = F, position = "center")

In TFT, the same rules apply as those in LoL in that Class1 is important in finding Champions that suit your teams needs. The Champion with the highest Hp is Aurelian Sol, a Starship. The Champion with the lowest Hp is Fiora, a Blademaster. When it comes to Armor in TFT, there are many Champions of various Classes with the same values resulting in no one Champion of Class having the max or min.

Joined_Champions %>% summarise(max(TFT.Hp), min(TFT.Hp), max(TFT.TotalArmor), 
    min(TFT.TotalArmor))

##   max(TFT.Hp) min(TFT.Hp) max(TFT.TotalArmor) min(TFT.TotalArmor)
## 1        1100         450                  40                  20

Joined_Champions %>% filter(TFT.Hp == max(TFT.Hp)) %>% select(Champion)

##      Champion
## 1 AurelionSol

Joined_Champions %>% filter(TFT.Hp == min(TFT.Hp)) %>% select(Champion)

##   Champion
## 1    Fiora

Joined_Champions %>% group_by(TFT.Class1) %>% summarise(max.TFT.HP = max(TFT.Hp)) %>% 
    arrange(desc(max.TFT.HP)) %>% kbl() %>% kable_styling(bootstrap_options = "striped", 
    full_width = F, position = "center")

Joined_Champions %>% group_by(TFT.Class1) %>% summarize(min.TFT.Hp = min(TFT.Hp)) %>% 
    arrange(min.TFT.Hp) %>% kbl() %>% kable_styling(bootstrap_options = "striped", 
    full_width = F, position = "center")

Joined_Champions %>% group_by(TFT.Class1) %>% summarize(max.TFT.TotalArmor = max(TFT.TotalArmor)) %>% 
    arrange(desc(max.TFT.TotalArmor)) %>% kbl() %>% kable_styling(bootstrap_options = "striped", 
    full_width = F, position = "center")

Joined_Champions %>% group_by(TFT.Class1) %>% summarise(min.TFT.TotalArmor = min(TFT.TotalArmor)) %>% 
    arrange(min.TFT.TotalArmor) %>% kbl() %>% kable_styling(bootstrap_options = "striped", 
    full_width = F, position = "center")

In order to find these values for both Lol and TFT, I used summarize max() and min() to find the values. To find the specific Champion I used filter to get only the observations with the specific value and selected for Champion to see who it belongs to. Finally to see the max() mins() for each class, I grouped the summary by Class1.

Summary: Unique Ranges

The map in LoL is quite large in comparison to TFT as TFT’s map is similar to a chess board with hexes instead of squares. Range in LoL is measured in an arbitrary quantity called “units”. While Range in TFT is measured by the number of hexes between Champions. For both games, there are set Ranges for the Champions in the game (there are only a select number of ranges they may have). As a result of these different map qualities, TFT is limited to only have 5 different ranges while LoL can have 13. Then, after filtering by Marksman, which is a Class I know is associated with Range, I wanted to see how the Range stats were distributed.

Joined_Champions %>% summarise(n_distinct(LoL.Range), n_distinct(TFT.Range))

##   n_distinct(LoL.Range) n_distinct(TFT.Range)
## 1                    13                     5

Joined_Champions %>% filter(LoL.Class1 == "Marksman") %>% select(Champion, 
    LoL.Range) %>% arrange(desc(LoL.Range)) %>% kbl() %>% kable_styling(bootstrap_options = "striped", 
    full_width = F, position = "center")

To find these values, I first did summarize(n.distinct()) to give me the number of unique values. Then to see how the Range was distributed in LoL Marksmans, I first filtered to the desired Class1 then selected for Champion and Range.

Summary: Damage

Joined_Champions %>% summarise(quantile(LoL.Damage), quantile(TFT.Damage))

##   quantile(LoL.Damage) quantile(TFT.Damage)
## 1               47.000                   10
## 2               54.104                   45
## 3               60.000                   50
## 4               63.000                   60
## 5               69.000                   90

Joined_Champions %>% filter(TFT.Damage > 60) %>% group_by(TFT.Origin, 
    TFT.Class1) %>% summarize(mean.TFT.Damage = mean(TFT.Damage)) %>% 
    arrange(desc(mean.TFT.Damage)) %>% kbl() %>% kable_styling(bootstrap_options = "striped", 
    full_width = F, position = "center")

For both Damage and DPS, I used quantile() to just see the range/distribution of these stats. For Damage, I grouped by Origin and Class1 because I wanted to see which Origin/Class1 combos were in the 75% in TFT.Damage. This was done by filtering and using the condition of being in the 75 Percentile.

Summary: Damage Per Second

Though the lowest base DPS for TFT was a lot lower than the lowest LoL DPS by about 21, the highest base DPS for TFT was higher than that of LoL by almost 35. This makes sense as TFT is a round-based game and certain Champions(considered to be stronger Champions) are not entered until later rounds.

Joined_Champions %>% summarise(quantile(LoL.DamagePerSec), quantile(TFT.DamagePerSec))

##   quantile(LoL.DamagePerSec) quantile(TFT.DamagePerSec)
## 1                   29.18739                        8.0
## 2                   34.66809                       31.5
## 3                   37.66528                       35.0
## 4                   41.08000                       44.5
## 5                   46.92000                       81.0

Summary: Attack Speed

TFT is a round based game so in the time of one LoL game many rounds take place. This is reflected in the mean Attack Speed for both games. Since TFT is a faster paced game, the average Attack Speed is higher.

Joined_Champions %>% summarise(mean(LoL.AttackSpeed), mean(TFT.AttackSpeed))

##   mean(LoL.AttackSpeed) mean(TFT.AttackSpeed)
## 1             0.6439804             0.7303922

Joined_Champions %>% group_by(LoL.Class1) %>% summarise(mean.LoL.AttackSpeed = mean(LoL.AttackSpeed)) %>% 
    arrange(desc(mean.LoL.AttackSpeed)) %>% kbl() %>% kable_styling(bootstrap_options = "striped", 
    full_width = F, position = "center")

Joined_Champions %>% group_by(TFT.Class1) %>% summarise(mean.TFT.AttackSpeed = mean(TFT.AttackSpeed)) %>% 
    arrange(desc(mean.TFT.AttackSpeed)) %>% kbl() %>% kable_styling(bootstrap_options = "striped", 
    full_width = F, position = "center")

After using mean() to find the average AttackSpeed for each game, I used group_by() to observe the average AttackSpeed for each Class which I thought might be influence each other.

Correlation Heatmap

The correlation heatmap helped to answer my purpose of this project, to see if the base stats of LoL influenced the base stats of TFT. It is also interesting to see correlation between the stats themselves. Looking at the heatmap, all base stats of TFT have a moderate to high positive correlation to the corresponding base stats of LoL. The highest correlation between the two games is that of the Ranges. Another interesting correlation is that for both TFT and LoL, Range and Total Armor had a strong negative correlation. It is interesting that in LoL, there is a decently high positve correlation between Damage and Total Armor but in TFT there is almost no effect.

corMatrix <- Joined_Champions %>% select_if(is.numeric) %>% cor(use = "pair")
tidycor <- corMatrix %>% as.data.frame %>% rownames_to_column("var1") %>% 
    pivot_longer(-1, names_to = "var2", values_to = "correlation")
tidycor %>% ggplot(aes(var1, var2, fill = correlation)) + geom_tile() + 
    geom_text(aes(label = round(correlation, 2)), color = "black", 
        size = 3) + theme(axis.text.x = element_text(angle = 90, 
    hjust = 1)) + scale_fill_gradient2(low = "white", mid = "light blue", 
    high = "blue") + ggtitle("Correlation Heatmap")

Range v TotalArmor of Classes

Joined_Champions %>% ggplot(aes(LoL.Range, LoL.TotalArmor, color = LoL.Class1)) + 
    geom_point() + ggtitle("Range v Total Armor of Classes in LoL") + 
    xlab("Champion Range") + ylab("Champion Total Armor") + labs(color = "Class")

Joined_Champions %>% ggplot(aes(TFT.Range, TFT.TotalArmor, color = TFT.Origin)) + 
    geom_point() + ggtitle("Range v Total Armor of Class1 in TFT") + 
    xlab("Champion Range") + ylab("Champion Total Armor") + labs(color = "Origin") + 
    facet_wrap(~TFT.Class1)

Average Damage of Classes

Once again, I wanted to see if I could find any patterns in how Origin may have been determined outside of just lore. This time I looked at the Average Damage for each Class1. This graph shows that some Origins have fewer Classes than others (Void and Valkyrie). In terms of Average Damage, the graph does not seem to show much. However, it is a bit difficult to tell, but it appears that the Dark Star Origin seems to have the Classes with the higher average Damage.

Joined_Champions %>% ggplot(aes(x = TFT.Class1, fill = TFT.Class1)) + 
    geom_bar(aes(y = TFT.Damage), stat = "summary", fun = mean) + 
    scale_y_continuous(breaks = seq(0, 100, 10)) + facet_wrap(~TFT.Origin, 
    ncol = 4) + theme(axis.text.x = element_text(angle = 90, 
    hjust = 1)) + ggtitle("Average Damage across Class1 and Origin in TFT") + 
    xlab("Class1") + ylab("Average Damage") + labs(color = "Class1")

Range and Total Armor

library(cluster)
pam_data <- Joined_Champions %>% select(LoL.Range, LoL.TotalArmor)
sil_width <- vector()
for (i in 2:10) {
    pam_fit <- pam(pam_data, k = i)
    sil_width[i] <- pam_fit$silinfo$avg.width
}
ggplot() + geom_line(aes(x = 1:10, y = sil_width)) + scale_x_continuous(name = "k", 
    breaks = 1:10)

library(cluster)
pam1 <- Joined_Champions %>% select(LoL.Range, LoL.TotalArmor) %>% 
    pam(k = 2)
pamclust = Joined_Champions %>% mutate(cluster = as.factor(pam1$clustering))
pamclust %>% ggplot(aes(LoL.Range, LoL.TotalArmor, color = cluster)) + 
    geom_point()

pamclust %>% group_by(cluster) %>% select(cluster, LoL.Range, 
    LoL.TotalArmor) %>% summarize_if(is.numeric, mean, na.rm = T) %>% 
    kbl() %>% kable_styling(bootstrap_options = "striped", full_width = F, 
    position = "center")

Joined_Champions %>% slice(pam1$id.med) %>% select(Champion, 
    LoL.Range, LoL.TotalArmor) %>% kbl() %>% kable_styling(bootstrap_options = "striped", 
    full_width = F, position = "center")

pam1$silinfo$avg.width %>% kbl() %>% kable_styling(bootstrap_options = "striped", 
    full_width = F, position = "center")

Damage and Attack Speed

library(cluster)
pam_data <- Joined_Champions %>% select(LoL.Damage, LoL.AttackSpeed)
sil_width <- vector()
for (i in 2:10) {
    pam_fit <- pam(pam_data, k = i)
    sil_width[i] <- pam_fit$silinfo$avg.width
}
ggplot() + geom_line(aes(x = 1:10, y = sil_width)) + scale_x_continuous(name = "k", 
    breaks = 1:10)

library(cluster)
pam1 <- Joined_Champions %>% select(LoL.Damage, LoL.AttackSpeed) %>% 
    pam(k = 9)
pamclust = Joined_Champions %>% mutate(cluster = as.factor(pam1$clustering))
pamclust %>% ggplot(aes(LoL.Damage, LoL.AttackSpeed, color = cluster)) + 
    geom_point()

pamclust %>% group_by(cluster) %>% select(cluster, LoL.Damage, 
    LoL.AttackSpeed) %>% summarize_if(is.numeric, mean, na.rm = T) %>% 
    kbl() %>% kable_styling(bootstrap_options = "striped", full_width = F, 
    position = "center")

Joined_Champions %>% slice(pam1$id.med) %>% select(Champion, 
    LoL.Damage, LoL.AttackSpeed) %>% kbl() %>% kable_styling(bootstrap_options = "striped", 
    full_width = F, position = "center")

pam1$silinfo$avg.width %>% kbl() %>% kable_styling(bootstrap_options = "striped", 
    full_width = F, position = "center")

Damage and Total Armor

Damage and Total Armor has 2 clusters. This is a relatively strong structure with an average silhouette width of 0.539. This clustering makes sense in that typically Champions are considered to have one of two roles: Damage Dealers and Damage Takers. When looking at the Champions that best represent these clusters, we have Syndra, a Mage, and JarvinIV, a Tank. This is consistent with the common knowledge in league that Mages are considered to be the best Damage Dealers while Tanks are considered the best defense as they are Damage Takers.

library(cluster)
pam_data <- Joined_Champions %>% select(LoL.Damage, LoL.TotalArmor)
sil_width <- vector()
for (i in 2:10) {
    pam_fit <- pam(pam_data, k = i)
    sil_width[i] <- pam_fit$silinfo$avg.width
}
ggplot() + geom_line(aes(x = 1:10, y = sil_width)) + scale_x_continuous(name = "k", 
    breaks = 1:10)

library(cluster)
pam1 <- Joined_Champions %>% select(LoL.Damage, LoL.TotalArmor) %>% 
    pam(k = 2)
pamclust = Joined_Champions %>% mutate(cluster = as.factor(pam1$clustering))
pamclust %>% ggplot(aes(LoL.Damage, LoL.TotalArmor, color = cluster)) + 
    geom_point()

pamclust %>% group_by(cluster) %>% select(cluster, LoL.Damage, 
    LoL.TotalArmor) %>% summarize_if(is.numeric, mean, na.rm = T) %>% 
    kbl() %>% kable_styling(bootstrap_options = "striped", full_width = F, 
    position = "center")

Joined_Champions %>% slice(pam1$id.med) %>% select(Champion, 
    LoL.Damage, LoL.TotalArmor) %>% kbl() %>% kable_styling(bootstrap_options = "striped", 
    full_width = F, position = "center")

pam1$silinfo$avg.width %>% kbl() %>% kable_styling(bootstrap_options = "striped", 
    full_width = F, position = "center")