Case Study Instructions:
- You have to provide your answer in slides. We’d like to point out that how you present the answer matters to us.
- Answer the questions and problems below to the best of your ability.
- You may use any outside resources necessary to help in your answer.
- You may use Python, Data Studio, or any other tool you wish. Please provide the codes used to produce the answers
- Please use the following tools to answer the questions: 1) Google slide for the presentation; 2) Google sheet for any spreadsheet and calculation instead of microsoft excel.
Gaming industry is an interesting field to explore, it would be fun knowing who is the most popular publishers and developers and which games are the most popular.
Content data:
- Name: Name of the game
- Sales: Sales of the game in Millions
- Series: Series of the game
- Release: Release date of the game
- Genre: Genre of the game
- Developer: Developer of the game
- Publisher: Publisher of the game
Find the most popular publishers and developers and which games are the most popular.
The dataset was provided by RevoU in their google sheet. To obtain the dataset, I copied the spreadsheet into my Google drive folder. First thing I did was preparing all Python libraries required for the analysis by importing them in this notebook. I imported Pandas for processing .csv file into a dataframe and plotting it into a basic visualization and Matplotlib for advanced data visualization.
import pandas as pd
import matplotlib.pyplot as plt
After all libraries imported, then I loaded the data into a dataframe using pandas command. Then, I checked the first 5 row of data using head() function in pandas to make sure the data was successfully loaded.
files = "/content/drive/MyDrive/Data Analyst Project/revou-cs3/games_sales.csv"
df = pd.read_csv(files)
df.head(5)
| Name | Sales | Series | Release | Genre | Developer | Publisher | |
|---|---|---|---|---|---|---|---|
| 0 | PlayerUnknown's Battlegrounds | 42.0 | NaN | 12/1/2017 | Battle royale | PUBG Studios | Krafton |
| 1 | Minecraft | 33.0 | Minecraft | 11/1/2011 | Sandbox, survival | Mojang Studios | Mojang Studios |
| 2 | Diablo III | 20.0 | Diablo | 5/1/2012 | Action role-playing | Blizzard Entertainment | Blizzard Entertainment |
| 3 | Garry's Mod | 20.0 | NaN | 11/1/2006 | Sandbox | Facepunch Studios | Valve |
| 4 | Terraria | 17.2 | NaN | 5/1/2011 | Action-adventure | Re-Logic | Re-Logic |
Last step, I checked the dataframe and the data types in there using info() function in pandas. There were 7 columns with 177 rows, only sales column had the float64 data type, the rest of columns had object data type. Release column contains release date of the game, and the data type should be in datetime format. Also look at the series column, it only has 141 non-null values out of 177 total rows. It means that there are 36 null values in the series column. I will handle this in the data process step in the next section.
df.info()
<class 'pandas.core.frame.DataFrame'> RangeIndex: 177 entries, 0 to 176 Data columns (total 7 columns): # Column Non-Null Count Dtype --- ------ -------------- ----- 0 Name 177 non-null object 1 Sales 177 non-null float64 2 Series 141 non-null object 3 Release 177 non-null object 4 Genre 177 non-null object 5 Developer 177 non-null object 6 Publisher 177 non-null object dtypes: float64(1), object(6) memory usage: 9.8+ KB
As mentioned previously, there were 36 null values in the 'series' column. Series column contains of name of the game series, which also contain the game name itself. For example, Diablo was the name of game series Diablo I, II, and III; or The Sims which was the name of game series The Sims 1, 2, and 3. Therefore, I cleaned those null values by replacing them with corresponding game name.
Using pandas, I splitted strings in the 'name' column and defined it as splits array. In s_cl, it only took the first string (which was the first name of the game) from strings that had been splitted in splits array.
splits = df['Name'].str.split()
s_cl = splits.str[0]
df['Series'].fillna(s_cl, inplace = True)
df.head()
| Name | Sales | Series | Release | Genre | Developer | Publisher | |
|---|---|---|---|---|---|---|---|
| 0 | PlayerUnknown's Battlegrounds | 42.0 | PlayerUnknown's | 12/1/2017 | Battle royale | PUBG Studios | Krafton |
| 1 | Minecraft | 33.0 | Minecraft | 11/1/2011 | Sandbox, survival | Mojang Studios | Mojang Studios |
| 2 | Diablo III | 20.0 | Diablo | 5/1/2012 | Action role-playing | Blizzard Entertainment | Blizzard Entertainment |
| 3 | Garry's Mod | 20.0 | Garry's | 11/1/2006 | Sandbox | Facepunch Studios | Valve |
| 4 | Terraria | 17.2 | Terraria | 5/1/2011 | Action-adventure | Re-Logic | Re-Logic |
Next, 'release' column had object/string data type. To format it into the proper datetime data type, I used the to_datetime().
df['Release'] = pd.to_datetime(df['Release'])
After replacing those nulls and changed the 'release' column data type, I then re-checked the dataframe. The result showing that no null values remained, and the data type was changed successfully. By seeing this information, it also showed that there were not any duplication in the dataframe. Now, the data is ready to analyze.
df.info()
<class 'pandas.core.frame.DataFrame'> RangeIndex: 177 entries, 0 to 176 Data columns (total 7 columns): # Column Non-Null Count Dtype --- ------ -------------- ----- 0 Name 177 non-null object 1 Sales 177 non-null float64 2 Series 177 non-null object 3 Release 177 non-null datetime64[ns] 4 Genre 177 non-null object 5 Developer 177 non-null object 6 Publisher 177 non-null object dtypes: datetime64[ns](1), float64(1), object(5) memory usage: 9.8+ KB
To get the oldest games released, I sorted the release date in ascending order. Thus, the oldest release date would be on top of the dataframe. Then, I set the range of index to show the first ten data in the dataframe.
# Game release sorted ascending (from oldest)
gr_old = df.sort_values(by='Release', ascending=True)
# Print 10 oldest games
print(gr_old[['Name', 'Release','Series','Publisher']][:10])
Name Release Series \
133 Hydlide 1984-12-01 Hydlide
34 Where in the World Is Carmen Sandiego? 1985-06-01 Carmen Sandiego
88 International Karate 1985-11-01 International Karate
162 Tetris 1988-01-01 Tetris
22 Last Ninja 2 1988-08-01 The Last Ninja
151 RoboCop 1988-12-01 RoboCop
30 Populous 1989-06-01 Populous
150 Return to Zork 1993-08-01 Zork
51 Myst 1993-09-01 Myst
156 Star Wars: Rebel Assault 1993-11-01 Star Wars
Publisher
133 Technology and Entertainment Software
34 Broderbund
88 Epyx
162 Spectrum HoloByte
22 Activision
151 Data East, Ocean Software
30 Electronic Arts
150 Activision
51 Brøderbund
156 LucasArts
As mentioned previously, gr_old variable sorted the dataframe by release column in ascending order. Which the oldest games were on top, and the newest games were on the last order in the dataframe. To get the newest games released, I reversed the range of index using negative stop value to extract last 10th data in the dataframe and negative step value to decrement one step.
# Print 10 newest games
print(gr_old[['Name', 'Release','Series','Publisher']][:-10:-1])
Name Release Series \
20 Valheim 2021-02-01 Valheim
27 Cyberpunk 2077 2020-12-01 Cyberpunk
110 Crusader Kings III 2020-09-01 Crusader Kings
10 Fall Guys 2020-08-01 Fall
139 Mordhau 2019-04-01 Mordhau
92 Satisfactory 2019-03-01 Satisfactory
26 The Forest 2018-04-01 The
13 Rust 2018-02-01 Rust
96 Kingdom Come: Deliverance 2018-02-01 Kingdom
Publisher
20 Coffee Stain Publishing
27 CD Projekt
110 Paradox Interactive
10 Devolver Digital
139 Triternion
92 Coffee Stain Publishing
26 Endnight Games
13 Facepunch Studios
96 Warhorse Studios
I counted rows of records in the developer and publisher column and grouped it into distinct string in their own column. After that, I extracted the top 5 developers and publishers using .nlargest() function. Then, plot them into a horizontal bar chart.
# Count how many records are in developers column, group by itself
dev = df.groupby(['Developer'])['Developer'].count()
# Top N using .nlargest() in Pandas
top5_dev = dev.nlargest(5)
# Plot the top N into the horizontal bar chart in Pandas
dev_g = top5_dev.plot(kind='barh', figsize=(10, 6))
plt.title("Top 5 Productive Developers")
plt.xlabel("Total Games")
plt.ylabel("Publisher")
plt.show()
# Count how many records are in publisher column, group by itself
pub = df.groupby(['Publisher'])['Publisher'].count()
# Top N using .nlargest() in Pandas
top5_pub = pub.nlargest(5)
# Plot the top N into a horizontal bar chart in Pandas
pub_g = top5_pub.plot(kind='barh', figsize=(10, 6))
# Show the value label using .bar_label() in Matplotlib
plt.title("Top 5 Productive Publishers")
plt.xlabel("Total Games")
plt.ylabel("Developer")
pub_g.bar_label(pub_g.containers[0])
plt.show()
I did the same step as previously to perform top 5 series with the most games. I counted rows of records in series column, then grouped it into distinct string in its column. I took the top 5 series using .nlargest() function, then plot it into a horizontal bar chart.
# Make a raw dataframe to exclude null or NaN values in series column aggregation
raw = pd.read_csv(files)
# Count how many records are in series column, group by itself
sg = raw.groupby(['Series'])['Series'].count()
# Top N using .nlargest() in Pandas
top5_sg = sg.nlargest(5)
# Plot the top N into a horizontal bar in Pandas
sg_bar = top5_sg.plot(kind='barh', figsize=(10, 6))
plt.title("Top 5 Series with The Most Games")
plt.xlabel("Total Games")
plt.ylabel("Series")
plt.show()
Sales analysis also used the same steps as previously, but I summed the sales value and grouped it into distinct series.
# Calculate total sales group by series
s_sales = raw.groupby(['Series'])['Sales'].sum()
# Sort the top N in series sales
top5_sales = s_sales.nlargest(5)
# Plot the top N sales into a bar chart
sales_g = top5_sales.plot(kind='barh', figsize=(10, 6))
plt.title("Top 5 Series with Highest Sales")
plt.xlabel("Total Sales (in million USD)")
plt.ylabel("Series")
sales_g.bar_label(sales_g.containers[0])
plt.show()
All findings in this analysis will be presented in a slide or powerpoint file as requested by the stakeholder.
Hydlide was the oldest game (published in December 1st , 1984) and Valeheim was the newest game (published in February 1st , 2021).
Top 5 publishers with the most games published were Electronic Arts (19 games), Blizzard Entertainment (10 games), Paradox Interactive (7 games), Activision (6 games), and GT Interactive (5 games).
Top 5 developers with the most games developed were Blizzard Entertainment (10 games), Maxis (6 games), Paradox Development Studio (5 games), CD Projekt Red (4 games), and Capcom (4 games).
Top 5 Series with the highest sales were Minecraft (33 million USD), Diablo (26 million USD), The Sims (24 million USD), Half-Life (21 million USD), and StarCraft (21 million USD).
Top 5 series with the most games were StarCraft (6 games), Command & Conquer (5 games), Age of Empires (4 games), Civilization (4 games), and Dark Souls (3 games).