random.bml

[nibot]

In a (usually futile) effort to break out of the usual LJ-clique and find new and interesting content*, I have a toolbar button that takes me to a "random livejournal" (http://www.livejournal.com/random.bml).

It just took me to frinkle_twinkle, whose current (from September) entry is all about how it seems she gets a disproportionate number of visitors via the random journal selector: http://frinkle-twinkle.livejournal.com/

Is the random-livejournal selector broken, or is this just observational bias? (Anyone could notice some random visitors, put up a note about them, and then suddenly have lots of people saying that they, too, got there via the random journal button, and isn't that funny.) I suppose the way to test this, aside from inspecting the source code to random.bml, would be to make a new journal with a post that says "Isn't it funny, so many people seem to be getting here via the random button" post, and see how many "me too" comments accrue.

* actually, just to waste time.

Comments

[codetoad.livejournal.com]

Obviously ORDER BY rand() is going to be expensive if run over all the users but what we might look at here is user.timeupdate (or something). Grab the 200,000 users who've updated in the past day and pick 5000 from that randomly. That can't be that expensive to do on a per day basis? (I just ran a SELECT .. ORDER BY random() limit 5000; statement on a 200,000 row table and it returned in about a second)