A new way to estimate steam game sales.

1/20/2023

Currently, the only way to estimate Steam game sales is using reviews. Review based estimations can show a trend, but are not very accurate at estimating individual game sales, as sales/review ratio can vary wildly between games.

A new way to estimate sales:

If we know the concurrent number of players playing a game at any point in time, we can calculate the total number of hours a game has ever been played. By dividing this number by the average playtime (total number of hours played by an average player), we can estimate the total number of players for the game. Since we can get concurrent player count as an exact number via Steam API, accuracy of this method totally depends on getting accurate playtime estimates.

How do we get average playtimes?

We can aggregate average playtime from all public indicators, like public profiles and game reviews.

To test this, I made a script that aggregates playtimes and compared the estimates to some games with known sales figures. To my surprise, this method seams to be consistently accurate with an accuracy of around 70%. Of course, there are outliers, but this method seems to be more consistent than the review-based approach.

Conclusion

Good thing about this method is that it's not biased like review-based approximation. Sales/review ratios can vary greatly depending on genres, price, demographics, etc..., but this method is unbiased for all games.