This would be a terrible interview question and is wasting people's time. The problem should be clarified by HackerRank.

It expects output values that would arise if each player was ranked within continuous operation but handles input/output in batches. The pre-sorted player scores add to the confusion. The solver then needs to ignore available information (later player scores) and provide "wrong" or "stale" answers.

We cannot practice eliciting and solving for requirements if they are not presented in the question and there is no interviewer.