With the current implementation of the AF dashboard we only display scores for articles that get at least 10 ratings over 24h. There are few articles meeting this condition on a single day in our 100K sample and, unsurprisingly, these are articles on trending topics for the current day. The number of such articles is actually very similar to the one we had in the 3K sample, which suggests that we were sampling popular articles in a similar way. As a result, what we are observing now is not a selection of the best/worst articles based on ratings over the past 24h from the entire 100K sample but a ranking by quality of the few most viewed articles in that sample. This is not bad per se, it only serves a different function than the one we were considering: tracking these articles may help detect short-term changes in quality for trending topics, i.e. articles that get a very high number of views but possibly not a proportional number of active editors watching the article. To have a more representative list of articles in the dashboard we may want to relax the 10 ratings/day condition.