in reply to Rep counting
in thread How about

*Total* rep might not be desirable. A thread with 50 nested replies each with 1 or 2 rep could rack up 75 in total rep and have little useful content. I wonder if there's a useful metric we could apply here..? Average? Max? Some formula that comes up with a totally different "score" for ranking threads?

Replies are listed 'Best First'.
RE: RE: Rep counting
by Blue (Hermit) on Oct 18, 2000 at 21:53 UTC
    Fastolfe, I agree that total rep can be misleading with lots of replies with a minimal rep each. However, here's my thinking - if a thread can generate a much larger number of replies then average, and they are good enough to get at least a few votes apiece, then the thread as a whole is probably rather interesting.

    A "Max" rep works well, but right now it's the equivilent of "Best Nodes" and just reading the whole thread.

    I am not so fond of average, since I think a good thread and the same thread with one extra neutral reply shouldn't slip significantly. And a good thread with one extra bad reply could kill it.

    Now, if I was a statistician, perhaps an average of reputations that only counts nodes that are within a standard deviation of the average reputation of the thread. That would mean that one single reply doesn't make a mediochre thread great (and that single node could hit the best list anyway if it's good enough). It also means that a good thread with a few off-topic or wrong posts doesn't get obliterated.

    But that sounds computationally expensive with possibly not enough return. Hmmm.

    =Blue
    ...you might be eaten by a grue...

RE: RE: Rep counting
by footpad (Abbot) on Oct 18, 2000 at 21:38 UTC
    How about total votes (excepting nulls) next to the reputation?