Video games come in a whole range of genres and ideas. Some of those ideas really captivate a large amount of people. Is this there some way one can determine if a video game is gonna sell well?
(Skip this part if you really need to, this was my inital thought process.)
Let’s take a step back, and think about this problem, and what a predictive model should give us. Some genres of video games are going to be more popular in certain years. With shooting games dominating the 2000’s and 2010’s, when adventuring and platforming games were more popular in the late 1980’s and 1990’s. So, we can make an assumption that genre and year are woven together. What else can determine good sales? Having games released by a well-established publisher can certianly boost sales. There can be a bit of influence on the name of the game itself, too. I’m willing to bet if a game had “Super Mario” or “Grand Theft Auto” in the title, that that game would definitely sell more, because people are familiar with those series.
Searching through Kaggle for this type of dataset, I stumbled upon this one. It seemed to have everything I really needed for this project, albeit a bit lacking in some columns. To fix this issue, I saw it had a Github repo with a web scraper. I took a look at that code and made my own scraper to get data, which I then cleaned, so there would be no missing values.
Making the model was simple enough. Since were dealing with numeric data, using a regression model will work, and since almost all the other variables are not, I used a Random Forest algorithm. Surprisingly, the model accuracy is above 99%, which might suggest that it might be overfitting.
Now, we can graph a multitude of parameters to see which ones are correlated with each other.
One that makes sense is Score and Sales.
Better reviewed are obviously gonna get more sales.
Another is Platform and Sales. Unsurprisingly, the Wii has the most sold games. The Wii was accessable to almost everyone, not just your average gamer.
So, it seems like to have a popular game, you need to either: make it accessible for a wide audience, be a really good game, already have brand recognition, or release it on a popular console. I was hoping to find something out of the blue with this project, but instead, I just solidified my prior hypotheses, which is still a plus.
Thanks for reading!