I recently got my hands on a great data set related to video games sales worldwide. Not only does it includes sales data, but also genre, Metacritic score and user score, among others. All this juicy data allowed me to create a nice little dashboard.

The raw data was prepared by Rush Kirubi, which you can find over at **Kaggle.com**.

As always, I'll go through my discoveries, and then review some of the techniques I used in the dashboard; including the **radar chart** custom visual, **buttons **and bookmarks.

**Video Game Sales Analysis**

I started out with two main questions...

1. Do scores (critic and user scores) affect sales similarly?

2. Do all regions buy the same genres equally?

The answer, I found, was "no" to both questions. I found it most interesting that when I plotted out average sales on the Y axis, and score on the X axis separately for user and critic scores, the shapes were clearly different. **Critic scores** yielded what seemed to be *exponential* sales, whereas **user scores **yielded a sales in a *linear* function.

This was surprising to me. "Do user scores matter at all?" To find out, I needed to do a bit more digging. What would it look like if I divided it by genre? Let's look at just video games under the category of "Action", the most popular genre.

**Action Genre**

Hmmm... Critic scores didn't see much change in the shape. User scores look slightly different with a couple of points *hinting *at an exponential function. Let's take a look at another genre; Spots - the second most popular genre.

**Sports Genre**

It seems this genre is a bit flakey with its scores and sales. Hmmm maybe something to do with EA pushing out so many sports games and their advertising power... Even still, we see an exponential trend with critic scores and linear with user scores.

It turns out, this pattern appears in all genres. The closest that a user score comes to looking like an exponential function is with the "Action" genre, and maybe the "Puzzle" genre.

**So How Much Do User Scores Matter?**

If critic scores show more promising results in terms of sales, how much, if any, do user scores matter? To answer this, I jumped over to everyone's favorite friend, Excel, to do a bit of linear regression. **Disclaimer: I'm not a statistician!**

I ran three different regressions on Global Sales (Y input); one for each variable of user score and critic score, and one using both critic and user score.

**Regression for Critic Score:**

Multiple R (Correlation): 0.389

R Squared: 0.151

Standard Error: 2.665

**Regression for User Score:**

Multiple R (Correlation): 0.320

R Squared: 0.103

Standard Error: 2.740

**Regression for Critic & User Score:**

Multiple R (Correlation): 0.432

R Squared: 0.187

Standard Error: 2.608

It seems to me that the **user score plays a supporting role** in the generation of sales, as opposed to the other way around. The correlation is slightly higher for critic score (0.389) than it is for user score (0.320), and both are lower than the correlation for critic and user score (0.432). This suggests that games with both higher critic scores and user scores typically generate higher sales. But let's be honest here, I didn't need to do a regression analysis to figure that one out.

### How Many Sales Can I Predict Given a User and Critic Score?

The coefficients for critic and user scores were as follows:

Intercept: -0.074

User Score: -0.270

Critic Score: 0.039

With that, we can create the following formula:

Sales (in millions) = -0.074 + UserScore(-0.270) + CriticScore(0.039)

For Example: I expect my game to have a *critic score* of 75 out of 100 and a *user score* of 8 out of 10. The estimated sales would be:

Sales (in millions) = -0.074 + 8*(-0.270) + 75*(0.039)

Sales (in millions) = 0.691 million unit sales

**Alien Says:** This is obviously a *very *rough estimate. There are obviously many more factors that play into how many units a game title will sell, but it may be helpful if you are planning to launch your own game and need a quick sales estimate.

**Japan Likes Their Role Playing Games**

**Total Sales**

Looking at sales by genre across regions, it becomes apparent that Japan has a clear, favorite genre of video games. By far, role playing games (RPG's) out sold every other genre in Japan. Where it takes the 6th place in number of units sold in other regions, RPG's are 1st in Japan, followed by Action games which take the 1st place in other regions.

**Average Sales**

Results are slightly different if we look at *average *sales per title, as opposed to *total *sales, which can be skewed by a couple of record selling games. RPG's are still the number one genre for Japan, but interestingly enough puzzle games come in second. Shooter games are towards the bottom, which I think is a direct reflection of Japan's attitude toward guns.

This chart represents an aggregation over a period from 1988 to 2016. Has Japan always liked RPG's? I investigated further by using line charts broken up by region and genre... The highlited blue line represents the RPG genre:

The bottom left-hand chart represents sales in Japan. This definitively shows that Japan has always enjoyed their RPG's. Cool stuff! Personally, I do enjoy a good JRPG myself **cough* Final Fantasy 10 *cough**.

**Are there any trends with Critic and User Scores?**

Until around 2010, critic scores were typically lower than user scores. Since then, they've flip-flopped and now user scores are typically about 6 points lower than critic scores (on a scale of 0-100).

**How Well Do Critic and User Scores Agree?**

We see in the line chart above that critic and user scores are relatively consistent, only differing by about 6 points at any given time (except for the very beginning years). Are there any genre's in particular that are responsible for the difference? 2016 has a difference of about 6 points. Let's take a look at the genre's.

We can see that Platform, Shooter, Sports and Strategy have the largest gaps in user and critic scores - the largest being Platform and Sports (both have a gap of 17.7 points). Since the sports genre is bigger in terms of sales, I can conclude that:

**Sports games** have the biggest disconnect between user and critic scores.

**Do Ratings (E, T, M) Have an Effect on Scores?**

To break it down even further, I used radar charts to map out genres into different rating groups for E, T, and M rated games.

For E rated games, sports and platform games have the biggest gap. For T (and E+10) rated games, strategy games are thrown into the mix, but it is Platform games that seem to have the largest difference. M rated games don't seem to have much of a difference between user and critic scores.

**Bonus Questions!**

**Question: Are there games that sell with a ***good ***critic score (above 50) but a ***bad ***user score (50 or less)?**

Yes, at 4.4% of the data set, this represents 240 million units sold and an average of 0.8 million units sold per title for a period between 1988 to 2016.

**Question: Are there games that sell with a ***bad ***critic score (50 or less) but a ***good ***user score (above 50)?**

Yes, at a slightly higher number of 5.3% of the data set, this represents 100 million units sold and only an average of 0.3 million units sold per title for a period between 1988 to 2016.

**Power BI Dashboard Development**

** - Radar Chart -**

I decided to use a radar chart to display the differences among ratings and genres. From **Wikipedia**:

*"Typically, radar charts are generated in a multi-plot format with many stars on each page and each star representing one observation."*

I had never used a radar chart before, so I thought this would be a good opportunity to try them out. The one I'm using here can be imported from the Marketplace on Power BI (just called "Radar Chart"). It's quite easy to use. Just pass in the category and values and you're done.

** - Assigning Actions to Buttons -**

It's not obvious, but you can assign actions to buttons on Power BI. The way you accomplish this is by creating ** bookmarks**. I think they should have called them "views" or "snapshots" instead, as these terms actually describe what they are.

For example, I have several buttons on the second page of my dashboard. Each button is assigned to a specific bookmark that I previously made.

Clicking on the "Back 5 Years" button will show the graph for a time period 5 years back. What's actually happening is that it's taking me to a different "view" of the page, i.e. bookmark. To make a bookmark, unable the bookmarks pane in the View tab. From there, you can add a bookmark by pressing "Add". Doing this will create a "snapshot" view of the current page and whatever filters are applied. For my button that takes you 5 years back, I adjusted the filter to show the data 5 years back, then pressed the "Add" button in the bookmarks pane.

Let me know in the comments what you thought! I hope you learned something interesting!

Want to improve on your Power BI skills even further? I highly recommend Supercharge Power BI by Matt Allington.