Steam Algorithm Case Study

The Steam platform’s algorithm is a powerful engine that powers personalised recommendations, driving game discovery and sales for both indie and AAA titles. This Steam Algorithm Case Study explores the inner workings of this algorithm, shedding light on the collaborative filtering and neural network techniques that shape the user experience.

Key Takeaways:

Collaborative Filtering and Neural Networks form the core of Steam’s personalisation algorithms, utilising user playtime data to predict preferences.
The low barrier to entry on Steam leads to an overwhelming library, making effective recommendations essential for discoverability.
Steam’s data limitations, such as only tracking total playtime per game, can create blind spots in understanding user behaviour.
The Steam Algorithm reacts to sales velocity, revenue, and playtime signals, not just reviews.
Case studies on titles like Team Fortress 2, Forza Horizon 4, and Among Us showcase the algorithm’s impact on indie and AAA games alike.

Collaborative Filtering Mechanics and Biases

At the heart of Steam’s recommendation system lies collaborative filtering, a technique that leverages user-item interactions to predict preferences. The algorithm creates user taste vectors from playtime data, computes the cosine similarity between users, and predicts ratings as a weighted sum of similar users’ ratings. This approach, however, can exhibit biases towards popular games, as new titles struggle to gain traction without external discovery.

To address this, Steam employs L1 normalization to handle extreme playtime differences, such as 12,000 hours for Sid Meier’s Civilization V versus 10 hours for Anno 2070. The platform also utilises user wishlists and other engagement signals to improve the accuracy of its predictions.

Visibility and Discovery Algorithm

The Steam algorithm’s Visibility Score is a crucial factor in determining a game’s discoverability. This score is calculated as a combination of Engagement Rate (wishlists, clicks, video plays divided by impressions) and Conversion Quality. The algorithm reacts to sales velocity, revenue, and playtime signals, prioritizing games that are gaining traction quickly over those with high review scores but low engagement.

This approach is exemplified by titles like FIFA and NBA 2K, which sell millions of copies despite Overwhelmingly Negative reviews, as the algorithm recognises their revenue potential. In contrast, niche games like Wargroove may find success by targeting specific content hubs within the Steam platform.

Case Studies and Real-World Examples

The impact of the Steam algorithm can be seen in various case studies and real-world examples. The free-to-play transition of Team Fortress 2 drove significant user growth, while Forza Horizon 4 saw 100,000 copies sold shortly after its Steam port, capitalising on its prior following.

Indie titles like Lethal Company have also found success by leveraging content-friendly mechanics, while the sharp sales jump of Among Us demonstrates the algorithm’s ability to react quickly to emerging trends. Ultimately, the Steam algorithm operates in a lifecycle-agnostic manner, adapting to the unique characteristics of each game.

Conclusion

The Steam Algorithm Case Study reveals the intricate workings of the platform’s recommendation system, which utilises collaborative filtering and neural networks to personalise the user experience. With a low barrier to entry leading to an overwhelming library of games, effective recommendations are essential for discoverability and sales.

By understanding the algorithm’s visibility signals, developers can optimise their game’s performance and leverage the platform’s immense reach. The case studies and real-world examples showcased in this article demonstrate the algorithm’s ability to adapt to a diverse range of games, from indie to AAA titles, driving growth and engagement across the Steam ecosystem.

Sources:
Collaborative Filtering and Innovation: A Case Study from Steam Recommendations
An Analysis of Steam’s Recommendation Algorithm
Steam Visibility Playbook: The Algorithm Formula for Organic…
The ‘mysterious’ Steam algorithm – not that scary?
Valve breaks down how Steam visibility works: “Algorithms react to…
How the Steam Algorithm Works! (Explained by Valve themselves)
Steam algorithm: what factors directly affect its operation…