Out of all the statistics about Steam games discussed online, the most prominent is gross revenue - how many units did the game sell and for what price.

Across the web you’ll find reported revenue for games already released and calculators to help estimate the revenue your game might generate, but how do these calculations work? Below, we briefly summarise some of the widely used methods, give some insight into how we estimate reveunue at Game Oracle, and then we’ll offer some parting advice for how to evaluate your game’s performance.

Boxleiter Method - a.k.a Review Multiples

The most common method for estimating revenue is to use the number of public reviews the game has recieved on Steam. It assumes that for each public review, there are a certain number of sales. This ratio is called the “Boxleiter number”.

The basic formula is:

Estimated Sales = Number of Reviews x Boxleiter Number.

The original Boxleiter number most people used was around 50, meaning 50 sales per review. But more recent data has caused that number to vary widely and there are a number of factors that can effect the Boxleiter number that is used:

• Game genre
• Release year
• Price
• Relative size of the developer/ publisher

What is important to remember is that the method provides a rough estimate, not an exact figure and accuracy can vary significantly between games. Additionally, the introduction of Steam’s “Would you like to review this?” prompt in October 2019 has affected review rates, resulting in more reviews for some games and therefore a lower Boxleiter number¹.

While not perfect, the Boxleiter method provides a useful starting point for estimating Steam game performance based on publicly available data.

Polling Public Steam Profiles

Publically shared Steam profiles can be polled, usually using automated scripts that run daily or weekly, to measure total game ownership. This creates a sample of game ownership across the Steam userbase which can be used to estimate what percentage of all Steam users own a particular game. Sites like SteamSpy used to implement this method but in April 2018 Steam made profiles private by default², significantly reducing the sample size of public profiles and therefore impacting the accuracy of this method.

Given Steam’s massive userbase, you need a very large number of Steam profiles for any gross revenue estimate to be statistically significant, so Steam’s change in privacy poilicy had a large impact on this method. The other limitation is that “ownership” doesn’t necessarily equate to “sales” as it includes free copies, bundle purchases, discounts etc.

Concurrent Player Count

Steam provides public API access to real-time concurrent player counts (CPCs) and third-party sites like SteamDB and SteamCharts collect and aggregate this data over time. The data shows how many players are simultaneously playing a game at any given time.

Total playtime estimates can then be made by summing the CPCs for each hour over a period (e.g. daily or monthly). By dividing the total playtime by an estimated average playtime per player this provides an approximation of unique players in that period. If you assume a percentage of unique players represents new sales you can multiply estimated new players by the game’s price to estimate the revenue.

The problem with this method is that CPCs don’t directly correlate to sales, especially for single-player games. It also requires assumptions about average playtime and player retention and may overestimate for games with high replayability or multiplayer focus.

Game Oracle’s Modified Boxleiter Method

So what do we use at Game Oracle? At the time of writing we are using a modified Boxleiter method to estimate revenue, but are actively developing predictive models that take full advantage of our bespoke Steam Map. You can stay up-to-date with our latest product developments by subscribing to our free monthly newsletter.

Our Boxleiter method uses a Boxleiter number that accounts for:

• Release year of the game - older games have a higher Boxleiter number due to changes on Steam that have encouraged more reviews
• Percentage of positive reviews - data has shown that games with more positive reviews recieve more reviews overall
• Initial price of the game - data has shown that games with a higher price tend to have a lower Boxleiter number
• Size of the studio/publisher - games released by big budget studios tend to have higher Boxleiter numbers given the force of their marketing machines

Since we are aware that the Boxleiter number is often inaccurate, we always report our estimates as a range from 0.5 times the Boxleiter number up to 1.5 times the Boxleiter number.

Advice for Game Developers

A famous statistician called George Box once said “All models are wrong but some are useful” and the same applies here. No one knows the true number of sales for a game except the developer and publisher themselves and everything you read on the internet is an estimate.

Even where Boxleiter numbers have been “validated” it is often using a small collection of games from surveyed game developers or using outdated sources like the 2018 Steam revenue leaked dataset. You must therefore always take estimated revenue “with a pinch of salt” because it will always be wrong.

When evaluating the performance of games in a genre or area of interest, it is better to focus on statistics that we know are accurate:

• How many games have been released and how many games are coming soon or in early access?
• How many reviews has the game recieved?
• How many concurrent players does the game have and how has that changed over time?
• What percentage of reviews are positive and more importantly what are the players saying in those reviews?

References

1. Adam Di Lizia, University of Warwick | Are More Reviews Better? Evidence from a Policy Change on the Steam Store
2. PC Mag | Steam Accounts Just Got a Lot More Private