Valve has filed a new patent describing a feature that might potentially improve Steam recommendations. The proposed technology is supposed to create video compilations for users based on their preferences.
The patent, which was spotted by Game Rant, describes multiple machine-learning models generating special video clips based on multiple parameters (e.g. time window).
According to the patent’s description, this model analyzes a player’s preferences and selects certain titles based on, for example, their gameplay history or the amount of time played. Based on this information, the algorithm is supposed to create custom compilations containing video footage from games that this user might like.
Although this technology sounds like a great way to create even more personalized recommendations, it is worth noting that it might never be implemented by Valve. Game companies file tons of patents every year, sometimes almost on the daily basis, but the overall rate of approved filings remains really low.