Over the past 20 years, so many video games have been released, and some of them have enjoyed a lot of success. As in any artistic medium, critics spend a great deal of time reviewing many of the releases each year. But the question is, are critics in tune with game popularity?
Take for example the popular series Call of Duty. Many of the most popular titles released in the franchise are highly reviewed by critics, but not always. Here’s what a scatter plot looks like if we try to get an idea of how well sales went:
It appears that, for this popular series, the correlation between sales and reviews is there, but very weak. But that’s just one series, what would happen if we take a look at many more games, would we see anything?
It seems like there might be a connection between scores, but we can’t be too sure just by taking a look at the image. So we need to get a number that can represent how closely the two are connected. That’s what a correlation coefficient can tell us. If we calculate that, we find that they are connected, but not too strongly. The correlation is .18 (the maximum correlation something can have is 1), meaning that higher scores are predictive of a successful game, but that there are other factors that seem at play too.
Note that I only included games that made more than $1 million. Some are so low that it makes it very difficult to visualize. If we take the correlation of sales and scores for all the games, the number actually goes up to .25. Just like before, it tells us that scores play a role in how well a game does, but they aren’t everything.
What are other ways we can predict if a game is going to do well?
Let’s go back to the example of the Call of Duty series. Since 2000, at least one new game has been added to the series every year. It must mean that the popularity of the series leads the publisher to ask for more each year. And if we look at a graph, it looks like sales have indeed gone up almost every year.
When we look at the correlation we get .86, which means that sales are strongly related to what year the game was released. It appears that building a brand can really payoff in the gaming industry! But if that’s the case, we should look and see if there are other games with sequels that sell so well. Since programming a way to determine whether a game has sequels could be time consuming, I’m going to leave that question for now. I might come back to it later when I have time.
Does it matter what console a game is sold on? What console makes the most money? Graphing is a simple way to look at that question. If we graph the sums that each console has made, this is what we get:
But don’t assume right away that it’s a bad idea to make a game for the PS4! These are the total sales, we don’t know what a game on each system would normally make. If we instead graph the averages, we get a different picture:
If you take a look at the Playstation consoles, they almost always get the highest average sales of all the systems in their generation (excluding handheld devices, Nintendo has a corner in that market), with the exception of the generation the Nintendo Wii was released. So if you are going to release a game for just one console, shooting for a Playstation console is not a bad idea.
Bad news for PC gamers like me: it looks like PC game sales are pretty low compared to consoles. PC games usually have a longer shelf-life than games for other consoles, since older games can often still be played on newer computers. But even with that extra time, the PC performs much lower than other systems. If you’re wondering why PC ports these days seem like an afterthought, then, you can see why they’re a bad investment for game publishers. On the other hand, getting a game on PC distribution services like Steam is a lot easier than for Playstation, so that opens up the playing field for indie developers. Just don’t expect (if you are one) to make AAA game profits.
Thanks for reading! Let me know if you have any input! If you are interested in taking a look at how I analyzed the data for these visualizations, you can check out my code on Github. It also includes some extra exploratory analysis that I didn’t include here. If you’re interested more on this project, check out the main page.
Note: This blog post was inspired by an activity in the Data Learning Club. If you’re interested in becoming a data scientist, I would recommend joining in with the club and listening to the companion podcast Becoming a Data Scientist.