What Makes Video Games Sell?
By Colin Bowers
Filed under Historical Data
The global video game market is expected to be a $180 billion behemoth by 2021. With all that money up for grabs, companies spend lots of time, effort, and cash trying to identify what makes a video game successful. Thanks to a Kaggle dataset with data and sales records for over 16,500 games, we can get in on the fun.
The dataset contains information on platform and publisher, game genre, critic and user scores, and sales by region as well as globally. A quick look at the top ten of the global sales list shows that Nintendo, especially Wii, dominates the global market. Wii Sports has sold twice as many units globally as any other game! That will become a trend throughout this analysis.
The rest of this analysis tries to predict global sales using other data on video games. This approach should allow us to identify what features make a game successful. However, the dataset also contains information on sales by region. Looking at a correlation matrix of this information, it is clear that North American, European and Global sales are all correlated. However, Japan’s video games habits appear to be different. Specifically, Wii games are much less popular in Japan than North America or Europe (Wii Sports is the 25th highest selling game in Japan).
When trying to build a model of video game sales, our first assumption was that highly reviewed games would sell well and poorly reviewed games would sell poorly. A regression model using Critics Score and User Score to predict Global Sales showed that this data can explain 30.6% of the variation in sales data. This means critic and user ratings can provide some information on what game will sell but they are not overly predictive. A look at the regression coefficients for these features shows that higher reviewed games by critics tend to sell better (since the coefficient is positive). Counterintuitively, the coefficient for user reviews is slightly negative. This value is worthy of further exploration but may be because the most highly reviewed games by users tend to be niche games without wide viewership or a high number of reviews.
Since ratings are only moderately predictive, our next theory was that game genre would predict global sales. Unlike our ratings hypothesis, which was moderately correct, this theory was just wrong. Game genre only predicts 8.9% of the variation in global sales. In hindsight, this finding should have been obvious from a look at the top ten list since no genre is highly represented on that list. A look at the regression coefficients for game genre shows that Platform games are slightly more popular than other games, likely driven by the popularity of the Mario franchise.
Unlike genre, a look at the publisher feature on the top ten list shows a clear trend. Namely, Nintendo owns the global video game market. A regression using game publisher can explain 28.8% of the variation in global sales. A look at the coefficients for this regression confirms Nintendo’s clear dominance and shows that they are followed by Microsoft Game Studios, who produce the highest selling non-Nintendo game in Kinect Adventures.
As final step in this analysis, we combined the publisher and ratings models to build an overall regression to try and understand what features impact global video game sales. This model explained 37.3% of the variation in global sales and confirmed our finding that the most important predictive feature in the global video games market is whether or not a game was produced by Nintendo (followed by how it was received by critics).
A useful next step of this analysis would be to split the video game market into Nintendo and non-Nintendo games and analyze them separately. We could also look at gaming platform or even platform type (console, handheld). Ultimately there are a nearly unlimited number of ways to cut up and look at the global video game market. This look has shown that the next time your wondering how popular a video game will be, the most important question to ask is did Nintendo produce the game?
Message us on Twitter @StatXP or on Facebook with questions and/or comments!