On Mon, Aug 17, 2009 at 23:02, Alexey Zakhlestin<[email protected]> wrote:
>
> Formula used is:
>
> $karma = ($yesterday * 0.999)
> + 0.1*($avg_d_week - $avg_d_last_week)
> + 0.1*($avg_c_week - $avg_c_last_week)
> + 0.1*($avg_v_week - $avg_v_last_week);
>
> d = downloads
> c = comments
> v = rating-votes
Right, since this is "app karma" (rather than "user karma" or "news
karma") I'd include 't' = time since last release (this information is
in the midgard db for an app). I'd also have it change slightly more
frequently, so that a new release of a popular app quickly floats to
the top, but so that the top 5 is changing relatively recently
(there's no point showing the same 5 apps all the time).
Perhaps (and this is off the top of my head):
$karma = ($yesterday * 0.99)
+ 0.1*($avg_d_week - $avg_d_last_week)^1.5
+ 0.1*($avg_c_week - $avg_c_last_week)^1.4
+ 0.1*($avg_v_week - $avg_v_last_week)^1.6
+ 2*sqrt(gmtime() - $last_update_time);
Is the raw data available to download for experimentation? If not, how
does the formula above affect things (I suppose it's going to be
difficult to tell if no-one's updated their product in the last 19
days)...
Cheers,
Andrew
--
Andrew Flegg -- mailto:[email protected] | http://www.bleb.org/
Maemo Community Council chair
_______________________________________________
maemo-community mailing list
[email protected]
https://lists.maemo.org/mailman/listinfo/maemo-community
