How to estimate how many sales a Steam game has made

(click to enlarge)

I was having a discussion with Ryan Clark about using the number of reviews a game has on Steam to estimate its sales. He tells me the idea came from Mike Boxleiter.

Anyway, I recently got shown this article which lists the most popular Steam games of 2014 (Thanks Lars Doucet) and some other useful data like number of owners. So I thought I’d look up the review count of the top 50 games of 2014 and do an analysis to find out what the average amount of owners (sales) are per review using the data in that article.

Units sold and Revenue estimates

This means you can look up any game on Steam and multiply the review count by a magic number (or range) – see below – and estimate how many sales it has made. Then if you multiply by the price, you can work out the maximum gross revenue the game has made, although of course sales and bundles will greatly reduce this.

DISCLAIMER:

I’m not saying this is totally accurate or anything, but it’s interesting and fun if you are into this kind of stuff. Enjoy!

Some observations and the magic numbers

1) There are 22 F2P games in the top 50 (44%)

2) The date format is the stupid American format.

3) Average owners/review for all games is 148. However, it’s 201 for F2P games and 106 for paid games. An important difference.

4) For paid games (I’m not that interested in F2P games. sorry!), Football Manager and Dizzel were outliers with over 400 owners per review, so I removed them and got an average of 81. If I take out two more above 150, the average is more like 72.

5) The lowest average for paid games was 32. If I “eyeball” the graph of values (see image at top of post), I’d said the average looks more like 60-70 as that’s the longest flat area before it hikes up.

6) So to conclude I’d say you can use 30-100 as an approx guide. There won’t be any below 30, and there’s only a few over 100. Another way to look at it (based solely on this data) is to pick 70 as a kind of mid value and use 70/2 as the lower end and 70×1.5 as the upper end (=35 to 105). Someone who is better at stats will probably rip this apart, but you get the idea.

7) Average Players/Owners is 73% for all games (57% for F2P and 85% for paid … interesting) Lowest for paid was 40% and highest was 99%. If I take out two lowest paid outliers of 40% and 45% (next lowest is 69%) this gives an 89% average. I’m not sure if any of these games have been in bundles. If so I imagine that would drop this ratio right down.

Some things I have not taken into account:

1) Date released: Some games have been around longer to get more reviews. However, we know none of these games are more than 14.5 months old and they are all at least 2.5 months old. If I didn’t have to fix the dumb date format by hand in Excel (there’s probably a smart way), then maybe I could do more analysis.

2) I haven’t noted if they are indie games or AAA games.

3) I haven’t noted the genre the games.

4) Hours played (maybe people who play more vote more? Actually I checked this with a trendline and it’s true! So if you know a game is a short game, bank on a HIGH owners/reviews value, and if it’s a long game, go for a LOW owners/reviews value.)

5) I used Owners per review, *not* Players. However, the data clearly shows that the larger % of players to owners, the lower the sales per review, which makes sense because it means that more of the owners played it and decided to leave a review.

6) These games were super-popular. It’s possible that if you check less popular games their review counts change/become more unpredictable.

I hope you found this interesting, please let me know if you think I’ve screwed up anywhere.

Also if anyone wants to do the next 100, go ahead, I need to do some work now.

4 Responses to “How to estimate how many sales a Steam game has made”

  1. S.E. Gordon Says:

    “the stupid American format”

    Seriously?

  2. Grey Alien Games Says:

    Yes the format is stupid, not the Americans. Guess I could have made that clearer. It’s a constant bone of contention for Europeans who use dd/mm/yy which is more logical in my opinion.

  3. S.E. Gordon Says:

    I don’t know a lot of people who say, “Today is the 7th day of March, 2015″ or “Today is the 7th of March, 2015.” March 7th, 2015 is more compact and to the point.

    If the format is indeed derived from oral speech, how is your remark not commentary of the underlying people? Aren’t most of your customers Americans?

    The European format screws me up every now and then when I’m downloading sales reports, so touché.

  4. Grey Alien Games Says:

    In the UK we often say 18th of March or similar, and also March 18th. The point is that dd/mm/yy is a logical sequence as is the reverse (which also has the advantage of being alphabetically sorted), but moving mm out of the middle isn’t. Most of the world doesn’t use it and finds it confusing: http://en.wikipedia.org/wiki/Date_format_by_country

    Critiquing a date format is in no way a reflection on the American people. The two are totally separate. Lots of countries have their own oddities, some of which seem pretty weird to everyone else, but it’s not a reflection on their populations as a whole.