Analyzing data trends in the video Gaming Industry
The skills the author demoed here can be learned through taking Data Science with Machine Learning bootcamp with NYC Data Science Academy.
Motivation:
Within the last decade, data shows there have been major shifts in the way games are dispensed to consumers. While companies like Game Stop are on the decline, [1] online gaming sites like steam have dominated the mainstream gaming culture. This change in the way games are sold has had an impact on what kind of games are sold. Retailers are taking a sizable risk by selling hard copies.
If the game title doesn’t sell then that impacts the retailers who invested space, and money into that game title. Selling digital copies on a website like steam however, takes virtually no risk. This means steam doesn’t need to be discriminatory as to what titles it takes. This has lead to explosion of Indie titles as the bar for entry has been lowered.
This project is targeted at providing useful insights for small indie teams that don’t necessarily have the same marketing resources that a large production company might provide. This project looks through two websites to see determine what kind of aspects of a game correlate with popularity in this new gaming market.
Method:
The first website is http://store.steampowered.com/. Here we attempt to web scrape tags and game prices that appear on the “top sellers” section as well as the “new releases” section, to get a sense of the most popular and least popular tags in this online gaming store currently. Unfortunately steam limits the number of pages that can be scrapped at once by redirecting URL request. This wasn’t a problem that could be solved by my code. This meant that I had to request each URL that I wanted to scrap individually.
The next site to be web scraped was https://www.gamespot.com/reviews/. Here thousands of game reviews were scraped along side there rating. The goal was to see what kind of words were most associated with commonly played games. We also accumulated meaningful words that were least associated.
Data Results:
Here in this plot we see some words that match our expectation of most common tags on steam. Tags such as indie, multiplayer, action, and adventure.
Looking at the least occurring tags we get the graph below. It should be noted that there are a multitude tags that appeared 0 times. These are tags that may have been used in the past but have not been used in recent games.
Moving on to the next website we determine the most meaningful word in the reviews. Using the Natural Language Processing Tool Kit we removed stop words (words such as the, as, should…) and proceeded to isolate meaningful words. We get the following graph below.
Below is a sample of some of the least commonly mentioned words in the reviews that were still meaningful in terms of Gaming language.
Comparing the results of both methods we notice some similarities. Adventure, Role Playing Gaming and Action Games represent the most widely sold content. Puzzles, Mystery, Detective and Racing games appears to be on the lower end.
Limitations of this study
There are a couple thing that must be noted in this study. First this data collected is biased towards more recent trends simply because as time passes games that were in the new release section would have been removed. Items that were on the top sellers categories may lose there position to new releases as well. Secondly this words that were isolated relied picked partially based on human interpretation. The last thing to note relates to the implication that may be drawn from these trends. Just because a genre is larger doesn’t necessarily mean it’s a good idea to enter that market space This could mean that the market space is saturated in
Articles
[1]:https://www.theverge.com/2017/3/25/15059380/gamestop-store-closings-2017-digital-sales-collectibles-business
[3] steam.powered.com
[4] gamespot.com
About Author
