I dont understand. The additional experiment data is fairly convincing, but the random data example doesnt seem to disprove the effect in itself. With random data you are going to get a predicted score of 50 for every group, which is what is shown, but this seems to still indicate that, if this is really what people predicted, that low skill people are overestimating their ability. Obviously random data would exhibit the effect; why should it not?

Edit: i think i get it. The random data doesnt show that the low performers dont underestimate and the high performers dont overestimate on average, but this is the natural result if everyone has no idea how they performed. Thus my question above is exactly what they are trying to say; if everyone predicts randomly (everyone equally bad at predicting) the effect arises. So there might be no relashionship between performance prediction and performance