>This effect would disappear almost completely if they instead of estimating their rank they estimated their score

In the article, in the section "The effect is in the noise", just before the graphs, it says they estimated their score. Where did you get that they estimated their rank?