Make the score speak

1 · Measure the effect of the question

Beginner

After Steam's question arrived, what happened to reviews written just past three hours of play?

They became about 4 times as numerous as just below three hours They fell, as players tired of the question They doubled, as at every playing time

2 · Follow a bad month

Pass-through is the share of the monthly movement in reviews written in the first two hours that shows up in another group of reviews. Just above the three-hour threshold, it goes from 0.73 before the launch to 0.18 after.

In a given month, the score of the reviews from the first two hours drops by 10 points.

Intermediate

How much does the score of the group just above the threshold then drop, before and after the launch?

By 0.73 point, then by 0.18 point By about 7 points, then by under 2 points By 10 points in both cases

3 · Explain the preserved information

Each added review responds about half as much as a voluntary one to a bad month. Yet, at the volumes actually observed, the stream of reviews above the threshold lost no detectable information compared with the stream below (its agreement with the first-two-hours reviews changed by −0.03 relative to the stream below, 95% interval −0.13 to 0.06).

Expert

What explains it?

Added reviews are longer and more detailed Prompted players see the displayed score and follow it Their number makes up for what each review lacks

4 · Find what stays visible

Expert

Which bad news still clearly pulls the score down after the question arrived?

None: the score has gone deaf Failures acknowledged by developers, which run through complaints Drops seen in reviews from the first two hours

Score decoded

—