Earlier today, there were some fails when fetching the TLX and CodeChef ratings of several users. This issue has been fixed.

Additionally, an issue was also found when calculating the external participations' contributions on the users' ratings — instead of one-based performances (1, 3, 6, ...), we used zero-based indices instead (resulting in 0, 1, 3, ...). Due to this, some users' ratings were higher than intended. This issue has also been fixed and I am sorry for the inconvenience.