Most people guess low when r is above about 0.7. Dense scatter reads as noise even when the underlying relationship is tight, and the visual impression tracks something closer to r squared than to r. An r of 0.7 means about half the variance in y is explained, which looks far messier than "0.7 out of 1" sounds.
Every round is generated from a single seed shown in the address bar. Share the URL including the seed and every student sees the identical ten plots, so their scores are comparable without any account or login. Reload without a seed for a fresh random round.
