Data Analysis of Video Game Industry's Disconnect
The skills the author demonstrated here can be learned through taking Data Science with Machine Learning bootcamp with NYC Data Science Academy.
Data Science Background
According to the market research firm, The NPD Group, COVID helped propel U.S. video game sales to a record $56.9 billion in 2020, a 27% increase over the previous year based on data.
The global gaming market was valued at USD 173.7 billion in 2020, and it is expected to reach a value of USD 314.4 billion by 2026, registering a CAGR of 9.64% over 2021-2026. According to Fortune Business Insights™, the global gaming market is expected to gain momentum by reaching USD 545.98 billion by 2028 while exhibiting a CAGR of 13.20 % between 2021 and 2022. Recently, there has been a growing disparity between what video game players are looking for and what video game companies are developing for the market (i.e., Battlefield 2042, Grand Theft Auto, Cyberpunk 2077).
Research Questions
This project is focused on responding to two key research questions:
- Is there a disconnect between video game users and critics regarding scores?
- Do critic scores or user scores positively or negatively affect video game sales?
Dataset
The datasets used for this analysis were sourced from Kaggle.com and consisted of information on video game industry sales from 1985 to 2016. The datasets included each video game sold, its name, platform (PS2, Xbox, PC, etc.), genre, year of release, global sales, critic score, and user score.
Feature generation was then used to create three more variables that could help answer our two key research questions. Those generated features included the difference in scores between critics and users, grouped critic scores, and grouped user scores. After importing the data and gathering additional features generated, data cleaning was necessary. After merging our datasets on the video game title, I dropped empty values and removed duplicates, leaving us with a dataset of almost 8,000 rows and 12 columns.
Data Analysis
I began my analysis by plotting the global sales of the video game industry by genre. The median was used to detail the difference between one genre and another to remove any potential outliers. The resulting plot demonstrated that the top three genres in terms of sales were Sports, Shooters, and Platform games; the bottom three genres were Strategy, Adventure, and Puzzle games.
After reviewing which genres had the highest sales over our 31-year span, I wanted to observe the relationship between the scores that critics gave to video games compared to users. To do so, I plotted another bar plot illustrating the median critic score given to each genre and sorted it in descending order. Per the plot below, Sports, Strategy, and Shooter games were the highest scored by critics, while Adventure, Action, and Miscellaneous ranked the lowest.
After understanding which genres critics tend to give higher ratings to, the analysis examined the genres users preferred. Using a barplot of median user scores, we could observe the highest-ranked genres for users were Role-Playing, Strategy, and Platform, while Misc, Action, and Puzzle were the lowest-ranked genres.
The feature-generated difference in scores column was used to understand the difference between the users' and critics' points of view. For each video game title, the critic score was subtracted from the user score and grouped to illustrate the difference in scores between one genre and another. After calculating the scores for all titles in each genre, we could observe that users rank Adventure, Platform, and Simulation games higher on average than critics. In contrast, critics tend to have a stronger preference for Sports games.
After completing our analysis of the scores given to video games by their users and critics, it was time to move on to answering our second key research question, do critic scores or user scores positively or negatively affect video game sales? To do so, it was necessary to visualize the relationship between the two variables and plot them against the global sales. As we can see, there seems to be a positive correlation between critic scores and game sales, but there is almost no correlation with user scores. Consequently, which games sell more copies cannot be predicted based on the scores given by users and critics.
To analyze the correlations, a correlation matrix was built. As we can see, there is a positive 0.29 correlation between critic scores and sales and a 0.053 correlation between user scores and global sales. In terms of the difference between critics and users, there is a negative correlation, meaning the higher the difference between the user score and the critic's score, the lower the number of overall global sales of a video game.
Conclusion
The video game industry has gone through many changes over the last decade. With the rise of new technologies and the expanding interest in video games, the market is poised to continue growing. Unfortunately for users, it seems as though the market is not following their interests.
An analysis of the data brings to light a slight disconnect between users and critics. Critics tend to rank certain games the highest, while users prefer other genres of games. And in terms of scores affecting game sales, critic scores seem to have a positive but weak correlation with sales, while user scores do not correlate with sales.
Future Work
As for future work, several actions could be taken to solidify this report. One would be gathering more data to help paint a bigger picture. The dataset was restricted only to 2016, pre-COVID, and it would be interesting to see how much it has changed in the aftermath of the adoption of new gaming consoles that allow players from different consoles to play the same game.
A predictive model could be built using training and test data to see if we could predict the user, critic scores, and global sales based on our existing features or new features that could be generated. These new features could include the month of release instead of the year to determine if seasonality has an effect and segmentation by region to observe differences in different areas of the world.
About Author
