Earlier this month, the popular third-party tool Steam Spy was essentially left dead in the water after Valve announced new privacy settings for Steam users, including hiding libraries and profile data from public view by default. This meant Steam Spy, which tracked stats like games’ sales numbers and play time, wasn’t able to access the data it needed, and was taken offline. But now its creator has brought it back, although with less reliable data.
Steam Spy creator Sergey Galyonkin has revealed that the site will now use a machine learning model to measure sales numbers, based on “coincidental data.” In his blog post, he shared that his algorithm had predicted 252,000 sales of the new game Frostpunk, with the developers revealing themselves that it had sold 250,000 copies.
A number of Steam Spy’s original features are still unavailable, but the key stats like sales numbers and playtime can be found on the site. However, Galyonkin admits that this new version of Steam Spy is much less accurate than before. After testing the algorithm on 70 games, he found 90% of the predictions were within a 10% margin of error, adding that he also “saw some crazy outliers, where the difference between the estimates and the real data could be fivefold.”
Galyonkin says he was convinced to bring Steam Spy back after getting comments from over 200 game developers sharing how the tool had a positive impact on their operations. He believes that having some data is better than no data at all, as long as he makes its limitations clear.
In the meantime, he’s going to continue working to improve the tool’s predictions and bring back some of the missing features. “I do believe in giving everyone the access to the essential information, but until I fix everything, Steam Spy will be semi-closed to the general audience.”
SOURCE Steam Spy blog