SteamSpy's new algorithm, introduced after Valve's privacy changes in April 2018, isn’t as accurate as its original version, but its reliability varies depending on the context. The original algorithm relied on publicly available Steam user profile data to estimate game ownership, often achieving accuracy within a 10% margin of error for games with significant sales (typically over 30,000 copies). However, when Valve made user libraries private by default, SteamSpy’s creator, Sergey Galyonkin, had to pivot to a machine learning-based approach using "coincidental data" from outside Steam, like achievement stats and other public metrics.