Random comic algorithm

Joined: 06 Jan 2006 Posts: 26 Location: DC

Just wondering, when someone pushes the 'random comic' button, is the result weighted by the number of transcribed episodes? I mean, say you've got 2 comics and one only has 10 episodes transcribed and the other has 100 episodes transcribed. Would the 'random comic' button go to comic B ten times as often as comic A?

And which would be the 'fair' way to do it. On one hand you could say that a comic with more episodes transcribed deserves more viewers. On the other hand, it is biased against weekly strips and strips that are newer and therefore have less episodes available for transcription. On one hand it would be great as an artist of a new comic to get 1 out of every 300 or so clicks of the random button, but on the other hand it'd be annoying as a reader to keep getting the same episodes you just read because some comics only have a few transcribed episodes. On the other hand, it'd be just as annoying to keep hitting the random button and getting the same comics over and over again instead of variety because most of the transcriptions in the database are from a handful of the comics.

Joined: 17 Dec 2005 Posts: 120 Location: Haverford, PA

Obviously, it should be configurable!

Posted: Thu Jan 12, 2006 5:57 pm

Right now it's 1 chance per transcription - there's downsides to every model, but this seemed to be at least naively fair. I'm willing to entertain suggestions!

Joined: 17 Dec 2005 Posts: 120 Location: Haverford, PA

I was only half-joking about making it configurable. Just a extra param passed as a query string to random.pl.

The obvious ways that people might be interested in surfing through random comics seem to me to be:

Evenly distributed across transcriptions (current, default)

Evenly distributed across comics, then across transcriptions within a comic (removes weight of having deep archives)

If I was searching for a new comic to read, rather than a laugh, I'd probably use the latter. I mean, once I've landed on--say--Clan of the Cats, I've added it to my trawl or not, but I don't need to keep seeing more pages from it...

Joined: 03 Jun 2006 Posts: 121

jamused wrote:

I mean, once I've landed on--say--Clan of the Cats, I've added it to my trawl or not, but I don't need to keep seeing more pages from it...

For the record, I agree with this: I'd enjoy the feature more if I didn't keep getting the same series over and over. Currently, randomizing by series would mean that for each click the probability of a particular series coming up would be at most 1 in 815, which doesn't sound that bad to me. Also, it gives comic creators something closer to equal exposure.

Joined: 09 Aug 2007 Posts: 7 Location: Germany

As devil's advocate (and owner of one of those said deep archives), I do think that comics with larger archives should have a little more weight in the search engine than one-hit wonders. Not as much as they currently have to be sure, but it would give some value to comics that update regularly over abandoned comics. And I know I've found at least one comic URL that doesn't even exist anymore in the OhNoRobot database.

I practically flunked outta math in college so I can't quote formulas to you, but maybe something like 80% spread equally over each comic and 20% to the depth of the comic archives. Or 90-10 if that's too much for people.

Joined: 03 Jun 2006 Posts: 121

Assuming the database has a "date transription was posted" or "date comic was posted" field, you could give weight to the most recently updated comics. But I wouldn't want to see a handicap given to newly started comics - they're usually the ones who need new readers more than anyone.